Main

Marine plastic pollution is a global issue7, with 9–14 million metric tons of plastics entering the ocean annually8. Microplastics (1 µm–5 mm)1 dominate measured plastic counts and pose serious global threats to ocean health9,10. Most studies focus on surface waters, collecting microplastics from the upper 15–50 cm using net tows5. However, microplastics exist in various forms with complex properties that affect their interactions with the environment11,12 (Box 1), ultimately shaping their three-dimensional distribution13. Consequently, microplastics are found throughout the water column, from the coastal to open ocean, across all latitudes3,4,14,15,16. Despite this, a synthesis of current knowledge and research priorities regarding microplastics across ocean depths is lacking.

The water column of the ocean, providing the most voluminous habitat on Earth, has a vital role in biogeochemical cycling17. The ocean contributes approximately 50% of global net primary production, serves as a major sink for anthropogenic carbon dioxide18, and facilitates particle transport and distribution17,19. As marine particles and microplastics transit the water column, their interactions with biogeochemical processes influence their behaviour and impact marine ecosystems17,19,20, including the ocean’s carbon cycle21,22. Therefore, understanding the distribution and potential impacts of microplastics throughout the water column is essential.

To advance understanding of the distribution of microplastics, here we synthesize depth-profile data from 1,885 stations collected between 2014 and 2024 (Fig. 1a), describing distribution patterns by size and polymer type, and assessing potential transport mechanisms. We examine existing microplastics distribution models and highlight research challenges and priorities (Table 1) for advancing knowledge of subsurface microplastics distributions and impacts in the ocean.

Fig. 1: Observations of subsurface microplastics in the ocean.
figure 1

a, Global observation stations of marine microplastics collected through surface net tows (orange dots; data from refs. 2,5,89,124), at a single depth beneath the surface using various techniques (light blue dots), from multiple water depths (dark blue dots), and from sediment traps (red triangles). The solid yellow lines represent the predicted boundaries of the offshore convergence zones, where floating plastic debris accumulates125. b, Latitudinal trends in the adjusted large microplastics, residing within the 1–60-m depth bin (light blue line, subsurface microplastic data collected at a single depth in refs. 55,77,78,79,80,81,82,83,84,85,86) and those floating microplastics at the sea surface (0–50 cm, orange line; data from ref. 5). Only the trends of coloured lines are considered here (for exact values, refer to Supplementary Fig. 1). The coloured shaded areas denote ±1 s.d. from 1,000 Monte Carlo runs that used varying predictors for the Generalized Additive Models (GAMs). c,d, Depth profiles of microplastic abundances (particles per m3) observed in nearshore (c; data from refs. 33,67,68,69,70,72,103,126,127,128,129,130) and offshore (d; data from refs. 3,4,6,15,16,25,60,72,83,131,132) waters with the log–log regression fits (dashed lines) between microplastics abundances and water depth of individual studies. The exponent (ξ), also referred to as the slope, provides information on the rate of abundance change along the depth gradients. The lower particle size limit of each study in c and d is presented in Supplementary Table 3. The corresponding linear regression equations are shown in Supplementary Table 4.

Full size image
Table 1 Current understanding, process insights and research priorities for subsurface microplastics
Full size table

Uncertainties in microplastic measurement methods

The reported abundance of water column microplastics spans eight orders of magnitude (Supplementary Table 1). It needs to be noted that some of this variation arises from inconsistencies in collection and analysis methods (Extended Data Table 1 and Supplementary Tables 2 and 3). The physicochemical and biological characteristics of aged microplastics further complicate measurements (Box 1), as each plastic category demands customized sampling and analysis approaches23. For instance, the use of 19 different pore sizes of meshes and filters (Supplementary Table 3) may contribute to pronounced differences in reported abundances; smaller pore sizes capture more particles, as microplastics counts typically increase with decreasing size, following an approximate power law24. This effect is evident in parallel sampling efforts, where samples collected at the same site and depth but using 10-µm versus 500-µm meshes yielded 5–6 orders of magnitude differences in microplastic abundances6,25. Analytical techniques also drive discrepancies. Microscopy-aided preselection followed by chemical identification—a method applied in 70% of studies—relies heavily on the investigators’ experience and becomes unreliable for small microplastics26. For instance, a study employing micro-Fourier transform infrared (µ-FTIR) imaging, capable of identifying polymers particles down to 11 µm (ref. 27), revealed 2–3 orders of magnitude more microplastics in Arctic sea-ice cores than earlier light-microscopy-based estimates28. Variability may also result from subsampling rather than analysing the entire sample. Extrapolating results from subsamples can introduce biases of up to +600% (refs. 29,30). These methodological uncertainties currently hinder accurate quantification of the distribution of microplastics in the environment. Research efforts aimed at enhancing observation resolution in the water column, while employing consistent methods, are crucial in refining our understanding of the marine microplastics standing stock, whether measured by count or by mass.

Microplastics permeating ocean waters

Our synthesis reveals that subsurface microplastics sampling is concentrated in the Atlantic and Atlantic–Arctic oceans (Fig. 1a). In waters between 1 m and 60 m, large microplastics abundances, determined at a single depth and represented by light blue dots in Fig. 1a, can reach up to 800 particles per m3 (ref. 31), with a median of 0.49 particles per m3 (Extended Data Fig. 1). Comparing microplastic abundances in comparable size fractions (>200 µm; Methods), reveals consistent latitudinal patterns between subsurface microplastics (1–60 m) and floating microplastics (median: 0.02 particles per m3; upper 0.5 m; ref. 6), both peaking in subtropical zones (Fig. 1b). However, subsurface microplastic counts at 1–60 m depth are approximately significantly higher than those at the surface (P < 0.05, Extended Data Fig. 1).

Despite uncertainties, measurements from studies collecting samples at multiple depths per station (median 205 particles per m3; range 10−4 to 104 particles per m3, represented by dark blue dots in Fig. 1a) suggest that the water column represents a major reservoir of microplastics. High abundances have been consistently observed in deep waters, including over 1,100 particles per m3 at 100–270 m in a North–South Atlantic transect16, 600 particles per m3 at 2,000 m in the North Pacific Subtropical Gyre25, 200 particles per m3 at 2,500 m in the Arctic4, and 13,500 particles per m3 at 6,800 m in the Mariana Trench32. Along the Korean coast alone, an estimated 3.13 trillion microplastics sized 0.33–4.75 mm are present in the water column33. This accounts for a small but notable fraction of the 171 trillion floating plastics in the same size range observed globally, most of which are microplastics2. In the Atlantic Ocean, microplastics of the 32–651-μm size category in the top 200 m average 2,200 particles per m3, with an estimated mass of 11.6–21.1 million metric tons. This mass is comparable to the total input of plastic in the >300-μm size category (17–47 million metric tons) into Atlantic waters and sediments from 1950–2015 estimated in ref. 16. However, these estimates are subject to substantial uncertainty owing to coarse observation resolution, simplified ocean physics and methodological inconsistencies.

Subsurface microplastic distribution patterns

Sampling locations were categorized as ‘nearshore’ (within 200 nautical miles from shore) and ‘offshore’ (beyond 200 nautical miles)31. Microplastic abundances ranged from 10−3 to 104 particles per m3 nearshore and from 10−4 to 104 particles per m3 offshore. Despite potential underestimation in nearshore studies using microscopy-aided methods, the median nearshore abundance (500 particles per m3) is over 30 times higher than the median offshore abundance (16 particles per m3) where advanced methods were often used (Supplementary Tables 2 and 3). This nearshore median aligns with values from offshore plastic accumulation zones confirmed via µ-FTIR imaging (for example, 400 particles per m3 in the North Pacific Subtropical Gyre25 and 250 particles per m3 in the South Atlantic Subtropical Gyre3). High nearshore abundances match previous predictions and field observations34,35,36, suggesting that coastal regions may act as plastic accumulation zones akin to offshore gyres37. Factors such as proximity to terrestrial sources38 and shallow nearshore waters where turbulence draws plastics downwards32,36 likely contribute to this observation.

Microplastic abundances generally decrease with water depth (Fig. 1c,d). Nearshore waters show a much steeper decrease in abundance compared with offshore waters (P = 0.029; Extended Data Fig. 2 and Supplementary Table 4), with abundances declining by up to a 1,000-fold (Fig. 1c). This sharp decline probably results from the high mineral and biological productivity in coastal waters, which enhances the sinking rates of aggregated microplastics39. Diatoms, dominant in coastal ecosystems, contribute siliceous frustules (density 2.6 g cm−3) frequently found on the surface of microplastics, thereby increasing their ballast effect40,41,42. Similarly, calcite precipitates (density >2.63 g cm−3) associated with biofouled microplastics add weight, accelerating their sinking43. Together, these factors promote efficient vertical transport of microplastics in nearshore waters. These findings agree with models predicting rapid sinking of microplastics in biological productive coastal zones with strong downwelling, such as in East Asia34,35,36. Conversely, low productivity in offshore waters may cause less efficient vertical transport44,45,46, contributing to the observed differences in abundance decay rates. In addition, owing to its proximity to coastal plastic sources, the dilution effects on microplastic levels (stemming from coastal and riverine inputs) in coastal waters38 could also cause this rapid concentration decline. Direct measurements of microplastic vertical flux are scarce and methodological variations hinder clear insights22,47,48,49,50,51,52,53 (Supplementary Table 5). Future research should explore vertical transport mechanisms in ecosystems with varying biological productivity, combining concentration measurements with flux-profile analysis to improve understanding of the processes that transfer microplastics downwards, analogous to the study of particulate organic carbon54.

Abundance profiles of offshore microplastics smaller than 100 µm show a gradual decrease (within one order of magnitude) with depth (Fig. 1d), implying a relatively even distribution throughout the water column3,55,56. Models also suggest a uniform dispersion of small microplastics (1–100 µm) in the water column, behaving differently from large microparticles35,55,56, with sinking speeds ranging from 10−3 m s−1 to 10−6 m s−1, regardless of buoyant57 or dense58,59 plastics. Even biofouling minimally alters the settling velocities of small microplastics owing to their restricted surface area42. A two-orders-of-magnitude decrease in the abundance of large microplastics (100 µm to 5,000 µm) with depth in offshore waters also supports this size-dependent distribution6,60 (Fig. 1d). Models and observations show that, compared with small microplastics, large microplastics tend to either remain at the sea surface or quickly reach the sea floor, leading to a rapid decline in abundances along depth gradients25,35,61.

Long-term observations are crucial to understanding microplastic abundance profiles and vertical fluxes. Capturing both discrete (such as polymers and colour) and continuous (such as size, shape and density) characteristics of microplastics is essential to elucidate physical and biological redistribution mechanisms24,62. Physical processes, such as seasonal stratification and eddy-driven subduction, differ in temporal and spatial scales13 and remain poorly understood. Biological factors, such as the physical structure (for example, thickness, biomass and density63,64) of microplastic-attached biofilms across depths have rarely been studied. How microplastics interact with marine aggregates (composed of microbes, organic and inorganic matter19) during their transit through the water column, including factors such as sinking velocities and the cohesive strength of plastic-laden aggregates, also warrants future research.

Results are generally reported as counts, so measurements of the microplastic mass in the water column are limited. Microplastic mass concentrations vary by up to five orders of magnitude (Extended Data Fig. 3), partially attributed to methodological differences. In the North Pacific Subtropical Gyre, small and large microplastics concurrently sampled have comparable mass concentrations at shallow depths6,25 (Extended Data Fig. 3), but the ratio of small to large microplastic mass increases with depth25, highlighting the growing significance of small microplastics at greater depths.

Density stratification retaining large microplastics

Vertical density stratification in the ocean, driven by temperature and/or salinity gradients, creates pycnoclines, where the water density changes sharply65. These pycnoclines are common in the ocean and are often associated with intense biological activities and elevated concentrations of particles, such as the accumulation of marine snow66. Observations in both nearshore56,67,68,69,70,71 and offshore waters3,72 show increased microplastics within pycnocline layers compared with those above. Our synthesis suggests that the elevated abundances in stratified layers are mostly linked to large microplastics rather than small microplastics3,33,56 (Fig. 2a). Stratification affects particle settling through a combination of buoyancy, diffusion and viscosity, which slows sinking by increasing drag and suppressing vertical motion, directing particles along isopycnals73,74. Whether stratification impacts a particle’s motion depends on the stratification length scale (L), the distance over which water density changes, which ranges from 100 μm to 1 mm. Particles larger than L, such as large microplastics, are significantly affected, whereas smaller microplastics are largely unaffected73. Irregular-shaped particles undergoing stratification-induced reorientation experience further deceleration75. As a result, large and irregular microplastics experience prolonged settling times and concentrate at pycnoclines. The lack of small microplastic accumulation at density interfaces was also observed in a freshwater system76. Our findings also agree with a model study that identifies a pronounced accumulation of large microplastics at pycnoclines, compared with small microplastics, in the open ocean46. Further investigation into microplastics in pycnoclines and other distinct layers, such as the deep chlorophyll maximum and benthic nepheloid layers, is essential to understand not only transport but also the potential exposure of marine life.

Fig. 2: Vertical distribution of microplastics along water-depth gradients as a function of particle size, density stratification and sampling location.
figure 2

a, Relationship between microplastic abundance within pycnocline layers and those above the pycnocline layers as a function of microplastics size category: small microplastics (gold triangles; refs. 3,33,56) and large microplastics (brown dots; refs. 3,56,67,68,69,70,71,72). Dashed lines represent ratios of microplastic abundances within pycnocline layers to those above the pycnocline layers. Refer to Methods for large and small microplastics classification. b, Box-and-whisker plots for the measured abundances of small and large microplastics within and outside the well-known accumulation zones in the top 100 m of the water column (Methods). The datasets are from refs. 3,6,16,25,60, which sampled microplastics at multiple depths at each station (represented by dark blue dots in Fig. 1a). Bold black horizontal lines within the coloured boxes represent the boxplot medians, while the top and bottom of the boxes indicate the 25th and 75th percentiles. The whiskers indicate the largest and smallest measured values within 1.5 interquartile ranges from the box. The asterisk indicates a statistically significant difference between small microplastics and large microplastics (P = 0.03). ‘NS’ indicates no difference (P = 0.11).

Full size image

Vertical extension of microplastic accumulation zones

Large microplastics, both floating at the sea surface (0–50-cm depth) and present in subsurface waters (data from refs. 55,77,78,79,80,81,82,83,84,85,86 collected at a single depth between 1 m and 60 m), peak in abundances at mid-latitudes (generalized additive models (GAMs), P ≪ 0.01; Fig. 1b and Extended Data Fig. 1). However, their distribution varies along the latitudinal gradients. Elevated abundances are also observed in the 1–60-m depth range above 55° N and 60° S. The increased abundances at high latitudes align with modelling results, suggesting that subsurface currents carry microplastics to the polar regions87. Observations indicate that the Atlantic is the largest source of subsurface microplastics collected in Arctic waters86. Besides the input by oceanic flows, other sources such as atmospheric delivery and rivers (for example, Eurasian river inflows15,86) from lower latitudes88 could also contribute to the pronounced microplastic abundances at high northern latitudes.

The accumulation of large microplastics at 1–60-m depths at mid-latitudes matches plastic convergence zones observed by surface net tows89,90,91 and global models92,93,94, suggesting that these zones extend deeper. Three-dimensional simulations of microplastic transport in the global ocean also affirms the persistence of subtropical microplastic convergence zones at certain water depths, showing that the structure of these accumulation zones in the gyres remains discernible down to 16 m and disappears at 60 m (ref. 87). Other models suggest that buoyant microplastics in the top 10 m of the water column follow the surface pattern, with high concentrations in mid-ocean gyres35. Our analysis of large microplastics measured at multiple depths at each station in the subtropical gyres also supports this finding. Within the top 100 m, large microplastic abundances in accumulation zones are significantly higher than outside (P = 0.03; Fig. 2b). However, this difference is not found below 100 m (Extended Data Fig. 4). The existence of large microplastic patches in the near-surface waters can be mainly explained by a combination of wind-driven Ekman currents and prolonged residence of large microplastics in the upper water column35,95. Taken together, accumulation zones of large microplastics extend into the oceanic water column, primarily constrained to the near-surface waters; a pattern not evident with small microplastics. Further research efforts are essential to constrain the subsurface boundaries of accumulation zones, characterize microplastics in each layer and assess their ecological impacts.

Plastic-carbon entering marine particulate carbon pool

Plastic debris represents a source of allochthonous carbon (C) to marine ecosystems. Subsurface microplastic-C has been estimated by calculating particle mass and multiplying it by the C% in the chemical formulas of common polymers22,25. Our analysis shows that the ratio of microplastic-C to total particulate organic carbon (POC) increases with depth in subtropical gyres (P = 0.002; Extended Data Fig. 5). This is because 75% of particulate organic matter is remineralized in the upper 500 m (ref. 96), compared with persistent plastic-C. The microplastic-C:POC ratio can reach up to 5% at 2,000 m, reflecting shifts in the overall particulate composition. With the increasing plastic leakage into the ocean, substantial long-term addition of microplastic-C to the marine POC pool is anticipated. Microplastic-C, utilized by marine microbes alongside metabolites and organic debris released by microbial processes97,98, can influence biogeochemical cycles, such as nitrification and denitrification98. However, plastic-C altering the marine POC pool has received little attention, necessitating further exploration, as marine POC is central to long-timescale carbon sequestration and biogeochemical cycling19. In addition, 14C-depleted plastic-C can interfere with 14C-based age determinations by reducing the fraction of radioactive 14C (ref. 99). A 5% contribution of plastic-C could make marine POC samples appear approximately 420 years older than their true apparent age. Plastic-C introduces additional complexity to the interpretation of already scarce deep-ocean 14C data, complicating our understanding of ocean circulation, carbon cycling and past climate conditions100. This calls for more plastic-C data and adjustments to models that predict or interpret deep-ocean 14C levels.

A comparison of models with observations

Simulations increasingly explore the transport and fate of microplastics at different depths utilizing a variety of frameworks and parameters34,35,36,45,46,61,87,95,101 (Fig. 3a). Understanding how well these models reflect observed water-column microplastic patterns is key to elucidating dispersal mechanisms and identifying sources of error and uncertainty in subsurface microplastic estimates.

Fig. 3: Comparison between the specifications of modelling studies and polymer compositions as functions of sampling locations.
figure 3

a, The available model parameters for simulating the transport, distribution and fate of microplastics in the three-dimensional ocean from refs. 34,35,36,45,46,61,87,95,101. Various model frameworks, including Lagrangian and Eulerian ocean models, and coupled Earth system models, have been used in research with different aims. These models incorporate different parameters influencing microplastic transport, such as plastic properties (for example, size, shape and density), oceanic physical processes (for example, mixing, advection and diffusion), biological factors (for example, biofouling, aggregation and faecal pellets), as well as scenarios of plastic inputs and spatial scales. b, Proportions of buoyant and dense microplastics in nearshore and offshore waters. Data were compiled from 18 peer-reviewed papers on water-column microplastics that provided polymer compositions3,4,6,15,16,25,33,60,67,68,72,83,126,127,129,130,131,132. Polymers with densities lower than natural seawater (ρ = 1.025 g cm−3) are defined as buoyant polymers whereas dense polymers have a density greater than natural seawater. c, Proportions of polyethylene, polypropylene, polyester and polyamide microplastics in the water column in nearshore and offshore. Bold black horizontal lines within the coloured boxes represent the boxplot medians, while the top and bottom of the boxes indicate the 25th and 75th percentiles. The whiskers indicate the largest and smallest measured values within 1.5 interquartile ranges from the box. The asterisk indicates the statistically significant differences among different polymer categories (P < 0.05).

Full size image

In most models, vertical microplastic transport is governed by wind-induced mixing and large-scale three-dimensional advection processes (Fig. 3a). However, submesoscale (1–10 km) dynamics, such as eddy-driven advection—which are critical for the downwards transport of small oceanic particles19—are rarely parameterized35,36. Including submesoscale dynamics in future models could improve our understanding of small microplastic vertical transport and its variability.

Similarly, biological processes, including biofouling, marine snow formation and faecal pellet production—which enhance the gravitational sinking of microplastics20—are rarely taken into account34,45,46. When included, these processes are often modelled using temperature-dependent remineralization rates and constant grazing rates. However, regional variations in ecosystem structure102, which significantly influence rates, are generally omitted. In addition, biogenic minerals (for example, calcium carbonate, hydrated silica and celestite), known to be critical ballasting factors driving the absolute flux of sinking particles43,102, are overlooked in all models. Models commonly assume that microplastics are spherical and buoyant particles (Fig. 3a), despite observations of both buoyant and dense plastics of varying shapes with large differences in surface area to volume ratios. Such parameterizations contribute to discrepancies between model outputs and observations. For example, although observations reveal subsurface maxima of microplastics from epipelagic to abyssal depths3,4,25,68,72,103, some models suggest that most microplastics might only reach depths ranging from 150 m to 1,000 m on a global scale36,95 or within specific marine regimes46. Similarly, models indicate limited or even an absence of transport of microplastics to polar regions34,95.

The sparse and uncertain measurements (discussed below) of subsurface microplastics makes it difficult to accurately evaluate the model performance and other parameterizations. Despite these challenges, models consistently emphasize the role of microplastic size in influencing their vertical distribution, aligning with our analysis35,45,46,61 (Fig. 1b,c). For example, in the eastern North Pacific Ocean, microplastic sizes decrease towards the gyre centre, a trend predicted by models and confirmed by observations25,35. Similarly, in the Mediterranean, models predict that small water-column microplastics tend to reach the open ocean, primarily owing to the fast removal of large microplastics from coastal environments61.

These results underscore the need for diverse empirical data on microplastics. Such data should include, but not be limited to, reliable measurements of abundance and characteristics, environmentally relevant sinking rates, biofouling rates, incorporation-detachment rates from aggregates, and the structural integrity and degradation rates of plastic-laden aggregates. Data on zooplankton microplastic particle selection and ingestion, categorized by grazing strategy, would also be invaluable to improving estimates of faecal pellet transport.

Polymer composition of subsurface microplastics

Over 56 polymer types have been detected in our synthesized dataset (Supplementary Table 6). Buoyant polymers, constituting half of global plastic production104, dominate the subsurface microplastics synthesized in this study (Fig. 3b). But subsurface microplastics denser than seawater are more abundant offshore than nearshore (P = 8 × 10−4; Fig. 3b), probably owing to their higher fragmentation susceptibility25. Dense polymers typically have higher glass transition temperatures (Tg) than buoyant plastics. Tg is a critical parameter below which plastic becomes rigid and brittle, and above which it turns rubbery or viscous. This property increases the likelihood of surface erosion in dense polymers, leading to their fragmentation into smaller particles105. Dense microplastics were found to be skewed towards smaller size fractions compared with buoyant ones in the North Pacific Subtropical Gyre25. Considering this, certain processes could further expedite the fragmentation of dense plastics. Land-based dense plastic containers (for example, PET bottles), which can take years to reach offshore gyres92,106,107, undergo extended weathering and degradation, intensifying their fragmentation. A study in the North Pacific Subtropical Gyre found that almost half of the plastics with identified production dates dated back to the twentieth century, showing decades-long persistence108 and potential for microplastic release. A significant rise in dense plastics such as nylon and polyester, often linked to fishing activity4, is noted offshore (P = 0.001 and P = 0.004; Fig. 3c). The atmosphere–ocean influx of microplastics, estimated to range from 0.013 million metric tons to 25 million metric tons annually109, may also contribute, with polyester comprising a significant fraction of airborne microplastics110.

Another notable finding is the apparent decline in the portions of buoyant polypropylene submerged in the water column compared with the apparent increase in polyethylene offshore (P = 2 × 10−4 and P = 1.9 × 10−7; Fig. 3c), indicating faster polypropylene removal. Ultraviolet stability may contribute to polyethylene fragmenting less and remaining longer at sea, whereas the tertiary carbon in the polypropylene backbone makes it more vulnerable to abiotic degradation111. Photodegradation experiments have documented that polypropylene fragments and produces dissolved organic carbon faster than polyethylene in seawater112,113. These findings highlight the varying environmental degradation potential of different marine plastics.

Uncertainty sources and improvement strategies

Uncertainties in quantifying subsurface microplastics in the ocean water column stem from several factors. First, observational data on subsurface microplastics are modest and unevenly distributed owing to sampling challenges and time-consuming analysis. Compared with floating plastics2, data on subsurface plastics remain limited (Fig. 1a). Variations in microplastic concentrations with depth underline the importance of obtaining higher-resolution depth-resolved samples. However, collecting microplastics at different depths requires specialized tools that are not always accessible to the research community (Extended Data Table 1). In addition, ship time is required for sufficient sampling with depth, creating a trade-off between maximizing depth versus horizontal sample resolution. Collectively, these factors contribute to observational data paucity, causing the uncertainties in estimating global patterns. Second, uncertainties arise from inconsistent analysis methodologies, as highlighted in Extended Data Table 1, and Supplementary Tables 2 and 3. Differences in sample collection, purification, particle size limit, polymer identification approaches and corresponding extrapolation can lead to divergent results. Moreover, particle loss or contamination during sample preprocessing and transport further compromises result robustness. Finally, data processing and reporting influence the accuracy and comparability of microplastic determinations. Variations in spectral collection approaches and libraries significantly affect the accuracy of microplastic identification26,114. The main challenge in data reporting is the limited availability of original datasets detailing microplastic abundance, size and polymer type, hindering data standardization for meaningful comparison.

Enhancing sampling efforts and harmonizing methodologies are crucial steps to mitigate the existing uncertainties. Interdisciplinary collaboration and resource-sharing are essential, given the sampling challenges. Joint research cruises, multi-institutional collaborations and coordinated, targeted sampling campaigns can improve spatial coverage and depth resolution. Recent programmes on plastic pollution in coral-reef115 and freshwater116 ecosystems have yielded consistent datasets through global sample collection and standardized methods23. Collaboration between scientists focused on marine particles research to standardize sampling strategies across multiple regions can greatly benefit water-column microplastic studies. Established marine particulate research protocols117, such as trace-metal clean protocols, offer efficient procedures for sample collection and preservation. Sharing both new and archived particulate samples can enhance sample resolution for microplastic analysis. In addition, developing camera/optical-sensor-based techniques on autonomous platforms would enable continuous, high-resolution monitoring. For harmonized microplastic analysis, we recommend employing chemical imaging (for example, μ-FTIR imaging)3,118 followed by semi-automated data analysis with open-source reference spectrum libraries119,120 to identify particles smaller than 100 μm in each sample without subsampling. This minimizes human bias and allows for effective identification and quantification of microplastics down to 10 μm. For large microplastics, μ-Raman/FTIR combined with visual microscopic inspection can be applied for single-particle analysis, requiring no specific sample preparation or advanced skills. Besides improving observational coverage, refining current models and incorporating new parameterizations are necessary to enhance confidence in global projections of microplastic distributions across depths.

Conclusion

Microplastics in the ocean are mostly irretrievable and persistent1,9. Our synthesis of subsurface microplastic observations over a 10-year period generates a global benchmark (Table 1). Despite observational uncertainties, the substantial presence of subsurface microplastics in both nearshore and offshore waters underscores the ocean water column as a critical yet uncharacterized reservoir of plastics. Small microplastics show a gradual decrease in abundance with depth, suggesting a more uniform distribution and longer lifespan in the water column, whereas large microplastics are more effectively trapped by stratification. Offshore surface accumulation zones37 extend into subsurface waters but are primarily composed of large microplastics. Such size-dependent transport patterns align with existing models35,45,46,61. The prolonged presence and accumulation of microplastics pose risks to the marine biome, where substantial knowledge gaps persist1, and may impact biogeochemical cycles and isotopic measurements. This analysis highlights the urgent need for consistent methodologies, finer-scale investigations and broader international coordination to establish comprehensive, long-term monitoring and more accurate model projections. These efforts will improve understanding of microplastic dispersion, fate and impacts, aiding in formulating effective policies and management strategies.

Methods

Categorization

Small and large microplastics

Our analysis categorizes microplastics into two size class, small (<100 μm) and large (>100 μm), based on previous laboratory59,73 and field3,4,16,25 analyses, as well as modelling results35,45,46,61. Experimental evidence and mathematical models indicate that the relationship between settling velocity and particle size is described by quadratic linear regression, despite the influence of particle shape and density58,59,73. Once marine sinking particles and microplastics decrease below 100 μm, the settling velocities of different plastic polymers converge, typically differing by less than one order of magnitude (see Fig. 1 in ref. 39, Fig. 2 in ref. 59 and Fig. 3 in ref. 58). In addition, particles under 100 μm are less affected by the omnipresent density stratification, which generally hinders particle vertical motion73. Numerous modelling studies on the vertical transport of microplastics have demonstrated distinct behaviours for particles smaller than 100 μm compared with large particles35,45,46,61. Finally, field observations, using advanced identification techniques such as chemical imaging, further confirm that microplastics smaller than 100 μm dominate marine plastic debris globally, accounting for up to 80% of total plastic particles3,4,25,55.

Using ‘100 m’ as the reference depth for large-microplastics accumulation zone at depth

We identify subsurface microplastic accumulation in the 1–60-m depth range, where measurements from one reference were collected at a single depth (shown as light blue dots in Fig. 1a,b). Elevated subsurface microplastic abundances at 1–60-m depth at mid-latitudes corresponds to well-documented surface convergence zones at the sea surface within subtropical gyres92,93,94. This surface accumulation zone is reflected in our study, where floating microplastics of a comparable size fraction were collected at 0–50-cm depth with surface net tows (Fig. 1a,b). Data from the 0.5–1-m range are absent owing to discontinuities in sample collections along the water depth in the literature. This indicates that plastic accumulation zones stretch from the surface into the ocean water column. The existence of patches of large microplastics in the near-surface waters is mainly attributed to a combination of wind-driven Ekman currents and prolonged residence in the upper water layers enabled by the strong buoyancy of large-sized plastics35,95.

To investigate whether datasets from studies collecting subsurface microplastics from multiple depths per station (dark blue dots in Fig. 1a) could support our finding, we compare the abundances of microplastics across two categories: (1) large versus small microplastics; and (2) abundances above or below the water depth of 100 m. We chose 100 m as the lower boundary for the following reasons: (1) the upper 100-m layer is where the majority of wind-driven mixing occurs36,87; and (2) the lower boundary of the sunlit euphotic zone is at 100 m, a depth typically used as a reference for assessing the POC flux via the ocean’s biological carbon pump. Below 100 m, the consumption of organic-matter-associated particles varies significantly between oceanic provinces134. As biological processes are one of the main mechanisms for the vertical transport of microplastics, using 100 m as the boundary allows us to constrain variations in the biological factors when comparing microplastic abundances.

Search term

We acknowledge the possibility of inadvertent omissions. To ensure thoroughness in retrieving scientific articles containing quantitative subsurface microplastics data into the oceanic water column, the following search pattern was constructed for Web of Science (www.webofscience.com):

TS = (microplastic$ or microlitter or “micro plastic$” or “plastic particle$” or “plastic fragments” or “resin pellet$” or “plastic particle$”) and TS = (subsurface$ or subsurface$ or vertical$ or water column$) and ab = (marine or ocean$ or sea or seawater or coast$).

Inclusion and exclusion criteria for literature studies on subsurface microplastics

The estimates of microplastics in the water column show considerable variability, primarily stemming from inconsistent methodologies and human errors. To construct a robust dataset, this study focuses exclusively on research targeting subsurface microplastics with clearly defined sampling depths.

Spectroscopy identification methods, such as Raman or FTIR, are widely used and offer a robust approach for microplastic analysis. Visual differentiation of particles smaller than 300 μm using optical microscopy alone is considered unreliable26. Studies relying solely on this method without spectroscopy (for example, Raman or FTIR) for plastics smaller than 100 μm were excluded. Studies with incomplete methodological descriptions—such as missing details on identification techniques, filter pore size or mesh specifications—were followed up with email enquiries. Articles were excluded if essential details were not provided in the response.

Detailed methodologies are available in Extended Data Table 1, and Supplementary Tables 2 and 3.

Data collection

For each study meeting the inclusion criteria, details such as mesh size, microplastic concentration, particle size distribution, sampling coordinates and sampling time were extracted into a spreadsheet. When this information was not directly available, data were obtained from maps and graphics within the articles using the ‘WebPlotDigitizer’ tool (https://github.com/ankitrohatgi/WebPlotDigitizer) to address data gaps. If no maps or graphics provided the required information, an email was sent to the corresponding author. Articles were excluded if no response was received with the necessary data. The complete dataset is publicly available in the ‘Source data’.

Curating data to compare microplastics at depths of 1–60 m with those floating at the sea surface

For robust comparisons between microplastics in subsurface waters (1–60 m) and those at the sea surface within the upper 0.3-m layer (dataset from ref. 5), our analysis exclusively focuses on large microplastics. Floating microplastics at the sea surface are primarily sampled using surface net trawls with mesh sizes ranging from 200 μm to 300 μm (ref. 5). Therefore, microplastic concentrations in size fractions above 200 μm collected in near-surface waters were extracted from reviewed articles based on plastic size distribution data to facilitate the global comparisons. This refined dataset of microplastics in near-surface waters (n = 1257) comprises three size fractions: 5.3% (n = 67) of particles >200 μm, 78.7% (n = 989) of particles >250 μm, and 16.0% (n = 201) of particles >300 μm. The detailed information on the size fraction of microplastics in near-surface waters (1–60 m) is summarized in Supplementary Table 2. This approach also minimizes inconsistencies in the synthesized data collected from different projects, as some include fibrous microplastics in their estimates whereas others do not.

Building microplastic abundance distribution models

We fitted two GAMs to analyse the adjusted microplastic abundances in near-surface waters (1–60 m) and the floating microplastic abundances along the latitude gradient. To quantify prediction uncertainty, we employed a Monte Carlo simulation with 1,000 iterations. In each iteration, we resampled the data with replacement, fitted a GAM and predicted density values for 1,000 equally spaced latitude points. This process was repeated 1,000 times, generating a distribution of predictions. For each latitude, we calculated the mean and standard deviation of the predicted values to estimate central tendencies and uncertainties. This approach ensures robust predictions with uncertainty estimates across various resampling scenarios.

Power-law function for subsurface microplastic abundance profiles

A simple vertical model was built to replicate the observed plastic debris profiles in the water column of each study. This model is based on the measured relationship between microplastic abundances and water depth (Fig. 1c,d). In this approach, microplastic abundance is calculated as a function of water depth using the following equation: AbundanceMP = \(1{0}^{{a}^{-1}{\times [\log }_{10}({\rm{W}}{\rm{a}}{\rm{t}}{\rm{e}}{\rm{r}}{\rm{d}}{\rm{e}}{\rm{p}}{\rm{t}}{\rm{h}})-b]}\), where a and b represent the slope and intercept of the regression line obtained from the log–log plot of observed microplastics abundances against water depth.

Comparison of microplastic-C to total POC

The ratios of microplastic-C to total POC were measured in both the North Atlantic and North Pacific Subtropical gyres25. In the North Pacific, microplastic-C was estimated from in situ pump samples. The estimation process involved calculating particle mass based on polymer density and size (identified via μ-FTIR). This mass was then multiplied by the carbon content percentage (C%) of each polymer’s chemical formula. Total POC data came from three stations in the North Pacific. At two of these stations, samples were filtered using 0.5-μm glass fibre filters at 6 depths (40–5,300 m) in August 2017 using in situ pumps, with POC calculated as the difference between total carbon and particulate inorganic carbon135. In addition, the total POC at ALOHA station was calculated by multiplying the total particulate carbon (TC) by the empirical POC:TC ratio (about 90%), which was measured at Station ALOHA136. The total particulate carbon data at Station ALOHA were from bottle samples at 10 different depths (from 5 m to 350 m) collected on 16 November 2018 during the cruise KM 18–21 (http://hahana.soest.hawaii.edu/hot/hot_jgofs.html). In the North Atlantic, microplastic-C was calculated using data from drifting sediment traps. Plastic mass in the traps was determined via pyrolysis gas chromatography–mass spectrometry, with carbon content based on each polymer’s chemical formula. POC was measured from aliquots filtered onto combusted GF/F filters, which were exposed to fuming hydrochloric acid to remove carbonates. The dried filters were then analysed with an elemental analyser using an acetanilide standard. For more details on the method, the reader is referred to studies22,25. Finally, a log–log regression model is constructed to predict the relationship between microplastic-C and total POC as a function of water depth.

Statistical analysis

As the datasets were not normally distributed (Kolmogorov–Smirnov test) and lacked homogeneity of variances (Levene’s test), the Kruskal–Wallis test, a non-parametric method, was used for multiple comparisons. When significant, pairwise comparisons were conducted with the Mann–Whitney–Wilcoxon test. Statistical significance was determined at P < 0.05. In addition, the mgcv, MASS and boot packages were used to fit the GAM, run the Monte Carlo simulations and do the bootstrap analysis. All statistical analyses and visualizations were performed using R software (v.3.4.3, R Development Core Team).