Identification of methodological biases to assess global levels of microplastic pollution in rivers

Studies on microplastic (MPs) pollution in aquatic environments have been conducted for over 20 years, developing and implementing a wide array of sampling and analytical methods to capture their complexity and variability in waters. However, the selection of different sampling equipment, minimum particle size considered, and sample size (sampling volume), have led to a series of potential methodological biases. This could hinder comparison of results and, therefore, understanding actual environmental variability among studies and geographical regions. In the case of different minimum particle size and size range reported, which directly impact the particle counts in the sample (e.g., smaller particles are numerous), correction factors were developed to align the concentrations found to a predefine microplastic size range (Koelmans et al., 2020; Xue et al., 2024). However, other factors, such as the potential bias caused by sampling volume on concentration is not well understood. 

In this work, we developed a statistical model to isolate the effect of sample size in freshwater samples. Literature mining allowed the study of 7506 samples, from 363 studies, collected by the most common sampling equipment: Nets, pumps and bottles (grab sampling). Each of these methods showed a wide range of mesh/filter sizes used and sampling volumes. We first corrected concentrations to a default size range (0.02 mm – 5 mm) using the correction factor by Koelmans et al. (2020) to remove bias caused by size. Second, we identified that the sampling volume and the size-corrected concentration remained negatively correlated (rho = -0.7, p-value < 0.01), indicating an additional methodological bias. Third, we modelled this correlation through a regression analysis, adjusting the parameters to allow a secondary correction due to sampling volume. The residual term of the regression model was interpreted as the actual environmental variability (i.e., spatiotemporal variation in concentration) to preserve the actual differences between sampled sites. Finally, a ‘normalized concentration to a standard volume’ was obtained, minimising both methodological biases.

This normalized concentration allows assessing microplastic contamination level in each sample, irrespective of the method used. The mapping of the corrected concentrations reveals a new regional distribution in the intensity (by orders of magnitude) of the microplastic contamination in global rivers, where those areas oversampled using higher sampling volumes show higher levels of pollution than previously thought and vice versa.

References

Koelmans, A. A., et al. (2020). Solving the nonalignment of methods and approaches used in microplastic research to consistently characterize risk. Environmental science & technology, 54(19), 12307-12315.

Xue, Y., et al. (2024). Standardization of monitoring data reassesses spatial distribution of aquatic microplastics concentrations worldwide. Water Research, 254, 121356.