The global biological microplastic particle sink

For this study we use the University of Victoria Earth System Climate Model (UVic ESCM) version 2.9^40,41,42. The UVic ESCM is an intermediate-complexity earth system model with a resolution of 1.8(^circ ) latitude by 3.6(^circ ) longitude and 19 ocean depth levels. The surface ocean level is 50 m deep. The model contains a two dimensional energy moisture-balance model of the atmosphere, as well as representations of sea ice, ocean circulation and sediments, and terrestrial carbon. The particular biogeochemical version used here includes three phytoplankton functional types, namely diazotrophs (DZ), mixed phytoplankton (PH), and small phytoplankton and calcifiers (CO)⁴³. The model pre-industrial climate has been previously described⁴³, as has its response to business-as-usual atmospheric (hbox{CO}_2) forcing⁴⁴. The following sections describe the MP model. A model schematic is presented in Fig. 6.

Model description

The base model⁴³ was modified in order to quantify the roles of two of the three theorised biological export pathways on MP (aggregation in marine snow and zooplankton ingestion; for now, we neglect an explicit representation of biofouling²⁹). We distinguish between detritus that becomes faecal pellets, and the physical aggregation of marine snow, by introducing a new faecal pellet tracer to divert 50% of zooplankton particulate losses into a separate detrital pool²⁷. For simplicity, this new pellet detrital pool has the same sinking parameterisation as the original detritus. Using the same sinking rates for both detrital classes produces ocean biogeochemistry that is identical to the previously published versions of the model. In the model, plastic particles only interact passively with marine snow (they do not, for example, modify aggregate sinking rates), but they interact actively with zooplankton grazing (described below). Plastic particles have been observed to both increase and decrease the sinking rates of marine snow^21,22 and decrease the sinking rates of faecal pellets^20,38, but for simplicity and as a first approximation we neglect these effects in our model.

Three MP compartments are introduced; “free” (unattached) microplastic (MP), microplastic aggregated in marine snow ((hbox{MP}_A)), and microplastic in zooplankton faecal pellets ((hbox{MP}_Z)). All MP are considered to represent particles within a biologically active size range, but this size (and the particles’ composition) is never made explicit. These assumptions could bias modelled MP towards polymer types favoured by generic zooplankton³⁴ and particle sizes in the lower end^33,34 of the defined range of microplastic size. However, our parameter sensitivity testing in the Supplemental Information tests various fractional uptake rates that implicitly consider size and particle composition in how biologically “active” the MP pool is. As with all model ocean tracers, microplastic concentrations (MP) vary according to:

With T including all transport terms and S representing all source minus sink terms. The source and sink terms for free microplastic are:

Microplastic is emitted to the ocean (Emis) along coastlines and major shipping routes using a scaling against regional (hbox{CO}_2) emissions (a dataset provided with the standard UVic ESCM version 2.9 package download), in order to approximate degree of industrialisation and population density in this first version of this model. The rate of emission is a proportion of the total annual plastic waste generation ((F_T))⁴⁵. For now, abiotic degradation of macroplastics as a source of microplastics to the ocean is neglected to keep the model simple and focus on biological transport. The MP then exchanges with the marine snow ((hbox{MP}_A)) and zooplankton faecal pellet ((hbox{MP}_Z)) pools. A fast particle rising rate ((w_p)) of 1.9 cm per second⁴⁶ is prescribed to a fraction ((F_R)) of the free MP in each grid cell below the surface level as an approximation of positive buoyancy. An alternative approach would be to assign a uniform rise rate to all MP particles, and to subject the value of the rise rate to sensitivity testing. However, a weakness of this alternative approach is that the many types of plastic in the ocean have different characteristic buoyancies, which could produce unique particle pathways¹⁸. In this alternative approach it would be more appropriate to explicitly simulate multiple MP types in the model (which we sought to avoid in this first modelling effort for the sake of simplicity). Nevertheless, we conducted a sensitivity test using several different rise rates, and the effect of reducing the mean rise rate was similar to reducing the fraction assigned a rise rate.

In the current model version there are no abiotic breakdown rates (i.e., photo-degradation³⁹) or respiration losses⁴⁷ removing MP from circulation.

MP is modelled to aggregate in marine snow as:

MP particles are taken up ((A_{upt})) via a Monod function applied to the rate of marine snow formation (sources of detritus; (D_A) in nitrogen units, multiplied by an aggregation fraction, (F_A)) in order to approximate an increased likelihood of MP/marine snow encounter with increasing MP concentrations that approaches a level of saturation at high MP concentrations:

The uptake constant ((k_P)) is subjected to sensitivity testing, as is the fraction of marine snow aggregation ((F_A)). In this parameterisation, the aggregation of MP in marine snow represents the net uptake of MP into aggregates by both aggregation and biofouling processes. Biofouling occurs mostly in the upper 50 m³⁵, which is the entire surface layer grid cell in our model. The entrainment-release cycle of biofouling is implicit in our parameterisation via the microbial loop, which is temperature-dependent. Sensitivity testing of the (k_P) and (F_A) parameters therefore represent testing of the net aggregation due to non-zooplankton biological aggregation effects. MP is released ((A_{rel})) from marine snow at the rate of detrital remineralisation ((mu _D)). This rate is temperature-dependent and results in higher rates of release in the low latitudes.

A particle sinking term ((w_D)) applies to marine snow-associated MP, and has the same value as sinking detritus. The base unit of all MP tracers is number of plastic particles. As a first approximation we assume that all marine snow aggregates forming from free detritus have the characteristic of diatom aggregates (8.8 (upmu )g C per aggregate⁴⁸). Model detritus in mmol N is converted to mmol C using Redfield stoichiometry, which is then converted to (upmu )g C to calculate the maximum number of aggregates. The maximum number of aggregates is then multiplied by the aggregation fraction (F_A), to calculate (hbox{MP}_A) source and sink rates. MP is conserved for all MP tracers when surface flux balances sedimentary loss rate. What fraction of MP particles reaching the seafloor via aggregate and faecal pellet ballasting are returned to the water column ((F_B)) is tested. For simplicity and as a first approximation, detritus ballasted by calcite, and calcite⁴³, are assumed to not aggregate with microplastic.

Similarly, for MP associated with zooplankton, sources and sinks are:

The calculation of MP particle ingestion rate ((P_{upt})) is the same as for other food sources³⁷. A grazing preference ((psi _{MP})) for MP is subjected to sensitivity testing. This sensitivity testing implicitly examines effects such as biofouling altering the grazing preference of zooplankton for MP. It is assumed that 100% of ingested MP will be egested as faecal pellets and released ((P_{rel})) to the “free” MP pool at the rate of faecal pellet remineralisation, with no plastic remaining in the gut and no plastic being metabolised by the zooplankton. Ingesting MP also results in a reduced zooplankton carbon uptake rate¹⁹, with implications for primary and export production (although, Redfield ratios are conserved). Pellet-bound (hbox{MP}_Z) is considered to sink at the rate of faecal pellets ((w_D)).

Plastic is eaten by zooplankton in this model. The Holling II grazing formulation³⁷ is extended to include MP. Grazing of MP ((G_{MP})) is calculated as:

The maximum potential grazing rate ((mu _Z^{max})) is scaled by zooplankton population (Z) and MP availability (MP), and weighted by a food preference ((psi _{MP})), total prey (CO, PH, DZ, (hbox{Detr}_{{tot}})), and Z representing the food sources described in⁴⁴ and a half saturation constant for zooplankton ingestion ((k_z)). Grazing preferences must always sum to 1 in the model, so sensitivity testing of (psi _{MP}) requires that all grazing preferences must also be adjusted. This is done by varying (psi _{MP}) but requiring (psi _{DZ}) always be set to 0.1 (on the basis that diazotrophs are a poor food source, and to minimize disruption to the nitrogen cycle). The remaining allowance is equally split by the other (psi ) terms. The calculation occurs in N units, so MP is first converted to grams of MP using the MP particle-to-mass conversion of 236E3 tonnes MP = 51.2E12 particles MP ((R_{M:P}))⁴. It is assumed that 1 g MP will roughly replace 1 g of food (at Redfield ratios; (R_{N:F}) is the conversion from mol Food to mol N) in the zooplankton’s diet, and MP is thus converted to mmol N for the grazing calculation. However, we subject this ratio ((R_{F:MP})) to sensitivity testing. Zooplankton uptake of plastic is therefore:

MP particles are released from faecal pellets via remineralisation, which occurs at the same rate as the remineralisation of aggregates:

Model forcing

The model was integrated at year 1765 boundary conditions (including agricultural greenhouse forcing and land ice) for more than 10,000 years until equilibration was achieved. From year 1765 to 1950, historical (hbox{CO}_2) concentration forcing, and geostrophically adjusted wind anomalies are applied. From 1950 to 2100 the model is forced with a combination of historical (hbox{CO}_2) concentration forcing (to 2000) and a business-as-usual high atmospheric (hbox{CO}_2) concentration projection RCP8.5^49,50. MP emissions start from 2 million metric tonnes in year 1950 (a total plastic waste generation estimate⁴⁵), increasing at a rate of 8.4% per year. (hbox{CO}_2) and MP forcing is summarized in Fig. 7. It has been estimated that about 4% of total plastic waste generated enters the ocean³⁰, but that the microplastic mass found at the sea surface represents only about 1% of the annual plastic input to the ocean⁴. We test a range of input fractions (see Table 2), after applying a mass conversion from tonnes to number of MP particles⁴. Using a considerable over-estimation of MP pollution rate also implicitly accounts for abiotic degradation of larger plastics.

Experimental setup

A 700-member Latin Hypercube⁵¹ was used to test the microplastic parameter space of the model using the forcing described in the previous section. While biological model parameters might also influence microplastic uptake and transport, we limited our initial tests to the new parameters introduced above. A range of values was prescribed to the parameters listed in Table 2, in which the parameter space was randomly sampled with a normal distribution. The objective was to see what can be learned about plastic accumulation in the ocean, when very little is known about plastic/particle interactions and basic processes are still poorly understood. An analysis of the full Latin Hypercube parameter search is provided as Supplemental Information.

We adopted an incremental approach to increasing model complexity. We started with a control Hypercube where biology was not allowed to take up plastic, in order to first test the physical parameters ((F_T) and (F_R), the fraction of total annual plastic produced entering the ocean as MP, and the fraction assigned a rise rate, respectively). One hundred simulations were performed in this configuration, with the results analysed in the Supplemental Information. We next included passive plastic aggregation in marine snow (MP plus the (hbox{MP}_A) tracer) in a 300 simulation Hypercube, spread across the (k_P) (marine snow uptake coefficient) parameter space (0–1, 1–100, 100–1000 particles (hbox{m}^{-3}), each with 100 Hypercube simulations) in a normal distribution. These 300 simulations explored the 5 relevant MP model parameters: (F_T), (F_R), (F_A) (marine snow aggregation fraction), (k_P), and (F_B) (fractional return to ocean at the seafloor). These results are also provided in the Supplemental Information. Finally, we added active zooplankton-associated plastic (MP, plus (hbox{MP}_A) and (hbox{MP}_Z) tracers) as a third 300-individual Hypercube set. This third Hypercube is similarly split across the (k_P) parameter space in a normal distribution, but with the addition of grazing parameters (psi _{MP}) (MP grazing preference) and (R_{F:MP}) (the food to MP substitution ratio; 7 parameters in total).

Source: Ecology - nature.com