Transport and Fate of All-Time Released Plastics in the Global Ocean

Mismanaged plastics accumulate in oceans and threaten marine life. About 40 million tonnes of plastics have reached the oceans, where their fate remains unclear. To track the sources, sinks, sizes, and age of all-time released plastics, we developed a new mechanistic model and 15 synthesized decades of measurements. We find that Asian plastics are the largest contributor (76%) to marine plastics by mass but only affect the North Pacific and the Indian Ocean, whereas plastics from fishing and shipping activities contribute 24% by mass but cover 60% of the ocean surface. Using the model, we demonstrate that biologically productive nearshore (63%) or upper ocean (25%) ecosystems trap 88% of the marine plastic. This study provides a 20 model framework to assess the potential effect of future mitigation strategies.


Introduction
Plastics are durable, versatile, and ubiquitous in modern life. While their production has increased from 1.7 million tonnes in 1950 to 300 million tonnes per year today, plastic waste management has not kept up, resulting in >1 million tonnes of plastics entering the oceans every 25 year (Lebreton et al., 2017). Plastics are causing physical problems for wildlife, and the harmful chemical additives mixed into plastic polymers for improved performance (e.g. bisphenol A, lead) are leaching into the environment (Mato et al., 2001, Teuten et al., 2009, Gregory, 2009. After seven decades of production, they are found in coastal areas, subtropical gyres (Eriksen et al., 2014, Cozar et al., 2014, and remote areas such as the polar regions and deep seafloors Simulated concentrations near the discharge points along the coastlines of South Asia and Europe are also relatively high (up to 22 kg km -2 ). High concentrations are modeled in the Atlantic sector of the Arctic Ocean (0.83 kg km -2 ) due to intense fishing and ship traffic (Lusher et al., 2015b).
The model considers the transport and fate of common plastics: Polyethylene (PE), 25 Polypropylene (PP), Polyvinyl chloride (PVC), and lumped others with smaller discharges and similar density to seawater. All types are found in ocean basins but show different spatial distribution patterns. PE and PP have an initial density lower than the seawater while PVC and the others are denser or similar to seawater. Thus PE and PP float in the ocean surface for even seventy years and are easily transported by surface ocean currents and winds. They travel far 30 from source regions (river mouth or the open ocean where shipping/fishing activities occur) and accumulate in subtropical gyres (Fig. 1D). In contrast, PVC tends to sink near sources and does not converge in open ocean gyres (Fig. 1E).
We estimate that 9.9×10 4 tonnes (ranging from 7.9×10 4 -1.7×10 5 tonnes) plastics are present in the global surface ocean, of which ~45% are microplastics (defined as diameter < 5 mm) and 35 ~55% are macroplastics (defined as diameter > 5 mm) (Fig. 2, Fig. S10). This mass represents a small fraction (0.25%) of all the plastics that entered the oceans between 1950-2018. The model estimates that there are 2.0×1 0 12 catchable plastic particles (i.e. > 0.33 mm that is the mesh size of the towing net often used for plastic measurements), weighting 9.5×10 4 tonnes afloat on sea surface. The modeled spatial distribution of plastic concentrations is generally consistent with 40 previous measurements (Fig. 1A). The observed mass concentrations span more than five orders of magnitudes, while our model simulates a narrower range and the modeled number concentrations are also lower (Fig. S1). This inconsistency may come from the uncertainty of the size distribution of riverine discharge inventories and the model parameterizations (see Uncertainty Analysis). 3 The previous estimate for the plastic mass in the global surface ocean spans a wide range. Our estimate of total afloat mass is one order of magnitude higher than the result of .5×10 4 tonnes), but they considered only the surface convergence zones (Cozar et al., 2014). In addition, we model the change of the density of plastic particles influenced by phytoplankton attachment/detachment (aka biofouling/defouling) . The 5 upward movement of plastics (e.g. the effect of upwelling and the re-rise due to reduction in density caused by defouling) in the three-dimension model also increases the mass of plastics in the surface ocean. Indeed, Eriksen et al. and Koelmans et al. brought the estimate higher (2.7×10 5 tonnes and 3.1×10 5 tonnes, respectively) by considering the vertical distribution of plastics in the seawater or modeling more processes like sedimentation and 10 fragmentation (Eriksen et al., 2014, Koelmans et al., 2017, which is close to our high-end estimate. Other estimations that are higher than ours are due to the differences in parameterizing the plastic transport and transformation processes (Lebreton et al., 2019, Van Sebille et al., 2015. The model estimates that most of the total plastics entering the oceans are held by beaches 15 (42%) and benthic sediment (31%) with the remaining 27% in the water column (Fig. 2). Among the plastics in the water column, ~1% is suggested to be in the surface ocean with the remaining 99% are simulated in the water column below the surface (Fig. 2). The model suggests a cascade of plastics from macro-to micro-, and eventually to uncatchable ones by fragmentation/abrasion, and 3.5% of the plastics in the surface ocean become uncatchable due to continuous 20 fragmentation and abrasion. These masses make up the "missing" plastics from the surface ocean. Moreover, removal by the ingestion of marine organisms (Kühn et al., 2015), and interception by sea ice (Obbard et al., 2014), that are not included in our model, could also contain the "missing" plastics. But their importance may be much smaller than the beaching and sedimentation processes (Hardesty et al., 2017;van Sebille et al., 2020). The budget analysis also The model suggests that 92% of the plastics in seawater stay in the top 500 m (Fig. 2), as 60% of discharged plastics are light ones like PE and PP (Andrady, 2011). The modeled plastic 10 concentrations decrease with depth in all ocean basins due to the dilution of a large volume of waters below the surface, however, there are also variabilities among basins (Fig. 1B). These profiles also have important implications for the large-scale transport of plastics by deep ocean currents given their long lifetime. This thus calls for sampling the subsurface waters as most of the current observations concentrate in the upper one-meter depth. 15 Relatively high fluxes of plastics buried by sediments are simulated near coasts (10 0 -10 6 kg km -2 ), where plastic waste is discharged, and denser plastics sink rapidly ( Fig. 1C and Fig. S3B). Contrary to the hypothesis that deep seafloor is the major sink of plastics (Kane et al., 2020, Woodall et al., 2014, the model simulates that seafloors shallower than 100 m receive 68% of the plastics in sediments (Fig. 1C). The biofouled low-density plastics (e.g. PE) are also most 5 frequently buried in the shallow seafloor near the coasts, where the abundance of nutrients favors the bloom of phytoplankton that facilitates biofouling (Ivan Valiela et al., 1997). Another important sink of plastics is the interception by sandy beaches (Fig. 2), which are distributed along the coasts (Fig. S3A). In all, the model indicates that 88% of the plastics are trapped in the upper ocean or coastal regions (Fig. 2), posing large ecological risk for pelagic and benthic 5 ecosystems that are shallower than 500 m. However, the abundance of observational data over these regions so far has not been commensurate with this portion, and more data is needed to evaluate our projections.
The Arctic Ocean appears to be a dead-end for plastic debris due to the poleward transport from sub-polar North Atlantic Ocean (Fig. 1A). The Arctic Ocean seafloor (e.g. Barents Seas) is 10 thus an important sink of marine plastics (Fig. S3B) (Cózar et al., 2017). The fluxes of plastics buried by seafloors are high in high-latitude North Atlantic Ocean and around Greenland and Barents Seas (10 3 -10 6 g km -2 ) (Fig. S3B), which is associated with deep water formation over this region (Cózar et al., 2017). Plastic sedimentation in this area accounts for 12% of the global total sedimentation fluxes. This is alarming to the vulnerable ecosystems in this region (Overland 15 et al., 2014). Age of Plastics. By numerically tagging the discharges in different times and regions (see Methods), we find distinct spatial patterns for plastics with different ages and origins (Fig. 3). The plastics discharged before 1970 represent only 2.3% of the all-time releases and thus contribute little to today's marine plastics burden (Fig. 3A). Almost all of the oldest plastics are modeled to end up at the center of gyres ( Fig. S4A-B), and the spatial pattern keeps almost 5 unchanged in the last several decades of simulation. However, they are not the dominant contributor in these regions due to emission in 2000s and 2010s (elaborated in next section). Plastics discharged in the 1970s are modeled to dominate (about 34%) near Antarctica. It indicates that about 40 years are required for plastics to transport across the global oceans as ~90% of the plastics discharged in the 1970s were from the northern hemisphere. However, we 10 need to emphasize that the modeled magnitudes of plastics in this remote region are extremely low due to continuously beaching, sinking, fragmentation, and abrasion ( Fig. 1). The 40-year-old (i.e. discharged in the 1980s) plastics are still on their way to Antarctica and dominate in the high-latitude waters of the southern hemisphere. The plastics discharged in the 1990s and 2000s dominate in the lower latitudes than the 40-year-old ones. The youngest (10-  The age of dominant plastics in the convergence zones in both hemispheres is relatively young compared to the length of the history of plastics (~70 years). The 10-and 20-year-old plastics account for 60% and 25% of the floating plastics there, respectively (Fig. 3A). This is consistent 5 with the relatively young age found in the great pacific garbage patch by both recent studies (Lebreton et al., 2018) and earlier ones (Wong et al., 1974). Even though the 10-year old plastics begin to dominate in some of the 'garbage patches' (black boxes in Fig. 1C), the model suggests that only ~15% of the surface plastics discharged in the 2010s have reached the center of gyres and remote areas in the open ocean (Fig. S4G). It takes at least 10~20 years for most of the 10 riverine discharged plastics to reach and accumulate to a significant level of concentrations in the convergence zones, longer than the travel time revealed by Lagrangian models (Lebreton et al., 2018, Maximenko et al., 2012. As discharges during the 2010s account for 36% of all-time releases, which is much higher than the previous decades ( Fig. S7), we thus expect a substantial further increase of plastic amount in these 'garbage patches' in the coming decades. It implies 15 that improving solid waste management and curbing the discharge today will have immediate results in coastal regions, but may not take effect in these patches in 10-20 years.
The Origin of Plastics. The model indicates a widespread distribution of plastics dumped by shipping and fishing activities. This ocean dumping is the dominating source in 60% of ocean surface areas, although they only contribute 24% of surface plastic mass (Fig. 3B). They are the 20 most widely distributed plastics and dominate in the Arctic, the Southern, the North Atlantic, and the South Pacific Ocean. The dumped plastics can reach areas far away from where shipping and fishing activities occur (Fig. S5A) because these plastics are directly discharged into the open ocean. Compared to the terrestrial sources which discharge at the shallow edge of the seas and oceans, debris in the open ocean is more easily transported and scattered. This calls for 25 stricter observance to the international treaty MARPOL (International Convention for the Prevention of Pollution from Ships) that banned ships from dumping plastic waste into oceans.
The model allows us to directly relate the marine plastics to their region of release, which provides important source information to control the mass of floating plastics in different ocean basins. Excluding ocean dumping sources, Asian sources contribute 76% to the global plastic 30 mass in the surface ocean, due to the large plastic discharge resulting from increasing population and rapid industrialization in recent decades. The most impacted regions by Asian sources are the North Pacific and Indian Ocean (Fig. 3B). North American sources are the largest contributor to the South Pacific (38% in mass) and Arctic Ocean (37% in mass) (Fig. S6), which is associated with its peak discharge in the 1980s and 1990s. Their contributions in the eastern 35 North Pacific are smaller than that of Asian sources. Discharge from Europe gathers in the Arctic Ocean but its contribution to the Arctic Ocean (25% in mass) is smaller than North America. Sources from South America dominate 83% of the areas in the Southern Ocean due to its geographical proximity, but the absolute concentrations remain low (Fig. S5C). The African sources dominate in the South Atlantic Ocean (87% in mass) but their contributions in the Indian 40 Ocean are surpassed by sources from Asia (Fig. S5B). Sources from Oceania gather in the surrounding western South Pacific Ocean and contribute 70% of the surface plastic mass there.
Uncertainty Analysis. Our study is subjected to uncertainty in the estimate of plastic emissions to the oceans ( Fig. 4 and Table S4). The reported estimates of global riverine input of plastics vary by more than factors of two between the lower (Exp3 in Fig. 4) and upper (Exp2) bounds 8 (Lebreton et al., 2017). Other studies also suggest even higher mass of plastic entering the marine environment via rivers, fishing/shipping activities, and other pathways such as coastal erosion and groundwater discharge (Kershaw and Rochman, 2015). The historical discharge also has uncertainty as we assume the ratio of discharge to total use keeps constant (Geyer et al., 2017). Similarly, the model results are also sensitive to the emissions from marine activities and 5 its temporal trends (especially after MARPOL).
The size distribution of the plastic input also matters. The Lebreton inventory does not consider the plastic larger than 2.5 cm and may underestimate the mass of total discharge (Van Emmerik et al., 2019, Vriend et al., 2020. When using the plastic discharge fluxes estimated by Schmidt et al. (Schmidt et al., 2017) (Exp1), which assumes 94% of the global discharged as 10 microplastics, the simulated macroplastics mass in the surface ocean account for less than 15% by present day (Table S4).  (Lebreton et al., 2017), the degradation rate of all plastics is 1% month -1 , and the beaching rate is 10% day -1 . Exp1, the riverine plastic discharge dataset is from Schmidt et al. (Schmidt et al., 2017). Exp2: the riverine plastic discharge dataset is from upper estimate of Lebreton et al. 20 (Lebreton et al., 2017). Exp3: the riverine plastic discharge dataset is from lower estimate of Lebreton et al. (Lebreton et al., 2017). Exp4: the fragmentation/abrasion rate of all plastics is 10% month -1 . Exp5: the fragmentation/abrasion rate of all plastics is 0.1% month -1 . Exp6: the fragmentation/abrasion rate of microplastics is 10% month -1 and 1% month -1 for macroplastics. Exp7: the beaching rate is 25% day -1 . Exp8: the beaching rate is 1% day -1 . 25 The spatial and temporal variability of marine plastic sources are also important factors. In the Lebreton inventory, only the large rivers within 60° North and South are considered, and Asian sources account for 67% of the global discharge. While in another study, the United States contributes the most in 2016 (Law et al., 2020). The riverine discharge fluxes estimated by Lebreton et al. (Lebreton et al., 2017) also unaccounted for deep layers, which could deliver high 30 amounts of non-buoyant PVC and biofouled PE/PP to the seafloor (Morritt et al., 2013 , this is a non-peer reviewed preprint submitted to EarthArXiv 9 Pierdomenico et al., 2019). We find the uncertainty of emissions can be fully propagated to our estimate of plastic mass in the ocean (Fig. 4 and Table S4). Although we cover only selected possibilities in the uncertainty analysis, we expect an overall linear response for the environmental loads to further improved estimate of emissions.
Another source of uncertainty is the model representations of plastic transport and 5 transformation processes. We choose the fragmentation/abrasion and beaching rates as an example to test the sensitivity of model results to these parameters, as a full evaluation of all schemes and parameters is computationally prohibitive. The fragmentation/abrasion rates are a function of environmental factors including light, temperature, oxygen and plankton biomass (Andrady, 2011). Even though perturbing this rate by a factor of 10 (Exp4 and 5) causes less 10 than ~10% change for the total plastic mass in the surface ocean as the plastics are generally persistent in the ocean, the fraction of uncatchable plastics in surface plastics varies between 0.79% and 19% (Table S4). However, stronger fragmentation/abrasion generates more smallersize particles. In another sensitivity simulation (Exp6), we specifically increase the fragmentation/abrasion rate of microplastics by 10 times (but keep that for macroplastics 15 unchanged), as small plastics degrade faster than the large ones due to higher surface-area-tovolume ratio (Eriksen et al., 2014). The fraction of uncatchable plastics in the surface ocean also increases to 14%.
We find the beaching rate a major source of uncertainty, as beaches are a major reservoir for marine plastics (Lebreton et al., 2012) but the rate is largely unconstrained. The rates of both 20 beaching and resuspension and the eventual residence time of plastics on beaches depend on the characteristics of plastics, the coastal morphological features, wind, and wave conditions (Zhang, 2017, Kaandorp et al., 2020, Hinata et al., 2017, Lebreton et al., 2019. Our calculated overall beaching ratio (42%) is consistent with a previous study for the Mediterranean (49-63%) (Kaandorp et al., 2020). Perturbating the beaching rate from 1% to 25% day -1 (Exp8 and 7, 25 respectively) results that the beached plastics account for 8-62% of total discharge (Table S4). The simulated total mass in the surface ocean also varies form 5.7×10 4 to 2.0×10 5 tonnes, indicating a high sensitivity.
These uncertainty ranges reflect our current understanding of these processes (Hardesty et al., 2017. The general agreement between the model results and the 30 observations over the surface ocean may result from the cancellation of different uncertainties. Our assessment is thus designed to be illustrative, calling for further studies to improve the riverine discharge estimate and the rates of important plastic processes (e.g. fragmentation/abrasion and beaching) (Hardesty et al., 2017). Therefore, the model result should not be taken as a comprehensive simulation of the fate of plastics. However, the relative 35 temporal and spatial patterns and trends are more robust than the absolute amounts. As scientific knowledge evolves, our mechanical model framework could be improved using similar methodology. One drawback of this study is not considering the bi-directional exchange between beaches and the coastal waters as well as the resuspension of sunk plastics in the sediments. This also indicates the necessity of an Earth System modeling approach to incorporate the sediment, 40 beaches, water columns, and the ecosystems therein.
Policy Implications. Ocean plastics' research has focused on convergence zones that tend to be oligotrophic regions with relatively low productivity (Westberry et al., 2008). Our work reveals more complex spatial patterns for the distributions of plastics with different ages and source origins. We identify potentially new garbage patches in regions that are also important habitats for commercial catch fishes (Watson, 2017), such as the coastal and downstream regions of Asia and Western Europe. The increasing plastic waste reaching the oceans poses a growing threat to the survival of marine organisms (Jambeck et al., 2015). In addition, plastics that remain in the seawater become increasingly threatening, as they have been detected to be able to break into nanoplastics (Halle et al., 2017), whose trophic transfer and ecological toxicity have been 5 described in experimental studies (Tommy Cedervall et al., 2012). More measurements in these regions are in urge need to confirm our projections.
The ubiquity of marine plastics and their long-range transport potential revealed by this study indicate the necessity of global controlling efforts, such as the market-based instruments, policies, regulations, and legislation to reduce marine debris, as proposed by the Honolulu 10 Strategy (UNEP 2012). Based on the spatial pattern of plastics in the global ocean, the North Pacific and the Indian Ocean are the "hard-hit" regions that are most impacted by Asian sources. The newly discharged plastic waste remains close to coastlines. And most have not yet broken down to smaller sizes, which can be easier to remove from the sea by ordinary capture methods. In the coming years, efforts to control plastic discharge in these countries, such as banning 15 single-use plastic products, improving the waste collection system and infrastructure, and controlling illegal dumping and landfills are essential steps to mitigate the global plastic problem.

Methods
General description of the model. We simulate the fate and transport of plastics in the MITgcm 20 model framework (Marshall et al. 1997). This model is Euler-based and simulates the emission, transport, diffusion, sinking, and transformation (including biofouling, fragmentation, and abrasion) of plastics in each model grid cell. The model has a resolution of 2°×2.5° horizontally with 22 vertical levels, and a time step is of 4 hours. The ocean circulation data is from the Integrated Global Systems Model (IGSM) (Dutkiewicz et al. 2005). The ocean boundary layer 25 physics is modeled based on Large et al. (1994), and the effects of mesoscale eddies on isopycnal mixing are parameterized following Gent and McWilliams (1990). We run the model from 1950 to 2018. The model has a relatively coarse resolution to resolve the currents over coastal regions and western boundary currents such as Kuroshio and Gulf Stream but perform better over the open ocean (Dutkiewicz et al. 2005). The lower computational costs compared to high-resolution 30 models also allow us to perform long-term and multiple-scenario runs. We test different time steps and the results keep relatively robust due to the low stiffness in simulating these processes. The model includes a total of 54 plastic tracers. Plastics are divided into four categories according to the chemical composition and the density of each category is pre-determined: PE (950 kg m -3 ), PP (900 kg m -3 ), PVC (1410 kg m -3 ), and others (1050 kg m -3 ). The plastic's 35 density is increased when biofouled. Each category has six size bins: four for microplastics: <0.0781 mm, 0.0781-0.3125 mm, 0.3125-1.25 mm, and 1.25-5 mm, and two for macroplastics: 5-50 mm and >50 mm.
We develop a universal framework for the transport of these tracers in the global ocean. Tracking the three-dimensional motion of microplastics is distinct from estimating other trace 40 components in ocean models. Since plastics particles have non-negligible volume and different density with the seawater. Obtaining closed expressions describing the hydrodynamic forces experienced by rigid particles embedded in various flows has been a subject of active research for a long time (Magnaudet 1997). Equations could be selected and simplified, based on facts, to help the simulation. Most of the time in the global ocean, light particles (PP and PE) float and drift in quasi-two-dimensional motion relative to the sea surface. Besides, light particles would sink after becoming heavier than seawater by biofouling, and rise after defouling. These particles would also rise, once got immersed in water by convection and turbulence. Other particles that have a higher density than the seawater (e.g. PVC), would instantly sink after dumped. Sinking or rising particles make an approximate one-dimensional motion relative to the water column. 5 Thus, in our model, the motion of plastic particles is resolved into seawater transportation (threedimensional), sinking/rising (one-dimensional), and drifting (two-dimensional). We calculate particles' sinking, rising, and drifting velocity by dynamics and experiential equations. Without loss of generality, plastics particles in our model are treated as smooth rigid spheres. These processes are elaborated on in the sections below. 10 Plastic sources. We use the riverine plastic discharge dataset from Lebreton et al. considering the seasonality, spatial variability, and size distribution of local sources (Lebreton et al. 2017). This dataset estimates plastic discharge based on waste management, population density, and hydrological information. The total discharge is divided into micro-and macroplastics according to the ratio of micro and macroplastics in sampled rivers. We also consider direct ocean 15 emissions from marine activities such as shipping and fishing, which is 25% of riverine discharge (Faris and Hart 1994;Macfadyen, Huntington, and Cappell 2009). The discharge from marine activities is allocated spatially according to the global footprint of fisheries (Kroodsma et al. 2018). The historical trend of plastic discharge during 1950-2018 is from Isobe et al. (Isobe et al. 2019) and Plastics Europe (https://www.plasticseurope.org/en/resources/market-data). The 20 historical trend of plastic discharge from each continent is assumed to follow that of per-capita GDP (https://data.worldbank.org/). The discharge of each type of plastics is calculated based on its proportion in global consumption (China Plastic 2014).
Sinking and rising. The sinking or rising speed of plastic particles depends on its density, diameter, shape, and the seawater state. In a steady state, plastic particles have the same 25 horizontal velocity as the seawater around, and three balanced vertical forces acting on the particles: (1) where FD is vertical dragging force, Fg is gravity, and Fb is buoyancy. Vp is the volume of the particle, while Vs is the volume of particle that is submerged in seawater (in this case, Vp = Vs). CD is the coefficient of dragging, which is a function of the Reynolds number (Re) of a certain motion of a fluid. Ap is the horizontal sectional area of a particle, ρs is the density of seawater, d 35 is Stokes diameter of a particle, ρp is the mean density of a particle, u is the velocity of the particle relative to the seawater, and g is the gravity acceleration.
Based on Equation (1) where µs is the dynamic viscosity of seawater. By substituting Eq. (6) and (7) into Eq. (5), with few techniques, we can solve u (Flemmer and Banks, 1986). In this way, we get the rising or sinking speed of the particles, by which we simulate the vertical transport of the plastics particles in the seawater columns (More details see Table S2).
where subscript p is for particle, a is for air (or wind), and s is for seawater. u, us and ua are the velocity of plastic particles, seawater, and wind, respectively. Aa and As are the vertical sectional 20 areas of particles exposed to the air and seawater, respectively. ρa and Rea are the density and the Reynolds number of air, respectively. µa is the dynamic viscosity of air. f is the Coriolis parameter.
We assume the vertical forces act on the particle reach a balance: 25 with which we can solve the Vs (in this case, Vs < Vp) and subsequently Aa and As based on geometry. CD(Rea) and CD(Res) are calculated as functions of Reynolds number of air (Eq. 11) and seawater (Eq. 7), respectively, based on Eq. 6.
We solve u by assuming the horizontal forces Fa and Fs reach a balance while Fc is neglected due to a much smaller magnitude: 0 += as F F By substituting eq. (8) and eq. (9) into eq. (13), we solve u numerically by a gradient descent method (Table S3). This leads to a constant drifting speed u given ua and us. The ua is from 5 NCEP/NCAR reanalysis (Kalnay et al., 1996). The random walk of plastic particles caused by oceanic eddy turbulence with a horizontal scale that is smaller than the grid size is simulated as an isopycnal diffusion process in the model. The stokes drift process of plastic particles is also not included due to the lack of ocean wave conditions in the model. In our model, drifting velocity is treated as a correction of ocean tracer.
10 Fragmentation/Abrasion. Fragmentation represents the process during which large plastics break up into smaller ones. Abrasion refers to the process in which tiny plastics peel off from the surface of larger ones, usually caused by mechanical shearing (White and Turnbull 1994). In the model, we assume 0.9% of plastic in each size bin breaks into particles one size bin smaller per month, while an additional 0.1% is transferred to the smallest size bin (Niaounakis 2017 The surface downward shortwave radiation is taken from the CMIP5 project.

Beaching.
A plastic particle is 'beached' when it arrives at a beach-adjacent cell. The beaching 20 rate is geographically diverse depending on the coastal morphological features, wind, and wave conditions. Previous studies suggest that beached plastics can be eroded back to the ocean, and bi-directional exchanges occur between the beach and coastal seawaters (Lebreton, Egger, and Slat 2019). Atwood et al. found <10-94% of released plastics are beached and the majority of beaching occurs within the first 3 days (Atwood et al. 2019). Ocean drifter studies reveal that the 25 timescales of the beaching and resuspension processes range from several weeks to months under different conditions (Samaras et al. 2014;Hinata et al. 2017;Stanev et al. 2019). We consider a net beaching rate (i.e. beaching -resuspension) in this study due to the large uncertainty of the two processes. We assume that the plastics in grid cells immediately adjacent to sandy beaches are partially removed from the seawater and the mass of beached plastics 30 is proportional to the length of the beach in the cell. We assume a conservative beaching rate of 10% per day. The global sandy beaches dataset is from Luijendijk et al. (Luijendijk et al. 2018).
Biofouling and defouling. Biofouling of light plastic types (PE and PP) is modeled following Kooi et al. (Kooi et al. 2017) but adjusted for more realistic scenarios. Three stages with different degrees of biological attachment are considered: clean, balanced, and sinking. Clean 35 plastics have no biomass attached and stay floating on the sea surface with the same density as the original plastic materials. Balanced plastics are the same density as seawater and suspend in water columns. Sinking plastics are heavier than seawater and sink to the subsurface ocean. The biomass of phytoplankton in global oceans is used as a proxy to scale the overall biofouling potential from phytoplankton, zooplankton, marine snow, and the ingestion and inclusion in feces. The phytoplankton biomass data is taken from the MITgcm Darwin ecosystem model (Follows et al. 2007). Using phytoplankton as a proxy might introduce uncertainty as the community structure varies drastically in different ocean biogeoprovinces. The Darwin model also does not consider the vertical migrations of zooplankton. But it is relatively robust as it is 5 constrained by satellite remote sensing data (Dutkiewicz, Follows, and Bragg 2009). Especially, the phytoplankton distribution suggests stronger biofouling in productive coastal waters than the open ocean, consistent with empirical studies (Cole et al. 2011).
The volume of biomass on plastic particles (Vbf) depends on the algae volume Va, the number density of attached algae per unit area A (# m -2 ) and the surface area of the plastic 10 particle θp: where βa is the encounter kernel rate (m 3 s -1 ), Aa is the ambient algae concentration (# m -3 ), and 15 ma is the mortality rate (s -1 ). The encounter kernel rate βa is the sum of Brownian motion and advective shear collision frequencies (m 3 s -1 ): where Dp and Da are the diffusivity of plastics and the individual algae cells (m 2 s -1 ), respectively, rp and ra are the radius of plastics and the individual algae cells, respectively, γ is 25 the shear rate (s -1 ), κ is the Boltzmann constant (m 2 kg s -2 K -1 ), T is the temperature (℃), and µsw is the dynamic water viscosity (kg m -1 s -1 ) (More details see Table S4).
The transformation rate between different types of plastics τtrans (s -1 ) is calculated as: where δ is an adjustable coefficient tuned to match the result of Kooi et al. (2017), and ΔV is the deviation between the volumes of two plastic types, e.g.: ( , ) 5 All data are available in the group web site: https://www.ebmg.online/model_plastics.html.

Code availability
All model code is available at the research group website: https://www.ebmg.online/ model_plastics.html.