Quantifying Global Ocean Microplastics

In this study we determined abundances and mass of microplastics starting at the lowest size of 0.33 mm, which is a commonly used lower limit for pelagic microplastics [31] (link). The prefixes micro, meso and macro in relation to plastic pollution are poorly defined. Generally accepted microplastic boundaries are based on typical neuston net mesh size (0.33 mm) and an upper boundary of approximately 5.0 mm [31] (link). We have used 4.75 mm as our upper boundary for microplastic because this is a size for standard sieves used for sample analysis in most of the expeditions contributing data to this manuscript. Mesoplastic has a lower limit of 4.75 mm, and no defined upper limit. In this current study we set the upper boundary of mesoplastic at 200 mm, which represents a typical plastic water bottle, chosen because of its ubiquity in the ocean. Macroplastic has no established lower boundary, though we set it at 200 mm, while the upper boundary is unlimited. There is a clear need for consistent measures in the field [31] (link), and herein we followed a practical approach using commonly employed boundaries and logistic considerations (net and sieve sizes) in order to integrate an extensive dataset that covers the entire global ocean, including areas that have never been sampled before.
Of the 1571 field locations that contributed count data (Fig. 1), a total of 1333 stations also had weight data (Fig. S4). All these data were used to calibrate the numerical model prediction of plastic count and weight density [28] . For the comparison, we fit the model results to measured data by a linear system of equations of the form:
where y_i is the logarithm of a measured value of plastic count density (pieces km⁻²) or weight density (g km⁻²) for each of the N number of samples. K is the number of model output cases with s_ij a dimensionless model solution at the location of sample y_i. β_k and ε_N are the computed weighting coefficients and the error terms for a particular dimensionless model solution s_ij. This method can be used to fit an arbitrary number of model output cases to any number of measured data points producing a weighting coefficient and error term for each case.
In the model we used a set of three model results (K = 3), corresponding to different input scenarios [28] : urban development within watersheds, coastal population and shipping traffic. Values of β and ε are determined for both the concentration distribution (pieces km⁻²) and the weight distribution (g km⁻²) of each of the four size classes based on the linear system of equations. To compare the model results directly to the measured data, the weighting coefficient β_k computed above is used to scale the model output for each of the output scenarios.