The importance of ensuring representative sample volumes in microplastic monitoring - A predictive methodology

Abstract A large body of literature is available quantifying microplastic contamination in freshwater and marine systems across the globe. “Microplastics” do not represent a single analyte. Rather, they are usually operationally defined based on their size, polymer and shape, dependent on the sample collection method and the analytical range of the measurement technique. In the absence of standardised methods, significant variability and uncertainty remains as to how to compare data from different sources and so consider exposure correctly. To examine the issue, a previously compiled database containing 1603 marine observations and 208 freshwater observations of microplastic concentrations from across the globe between 1971 and 2020 was analysed. Reported concentrations span nine orders of magnitude. Investigating the relationship between sampling methods and reported concentrations, a striking correlation between smaller sample unit volumes and higher microplastic concentrations was observed. Many studies scored poorly in quality scoring protocols according to the sample size taken. It is critical that sufficient particles are measured in a sample to reduce the errors from random chance. Given the inverse relationship with particle size and abundance, the volume required for a representative sample should be calculated case-by-case, based on what size microplastics are under investigation and where they are being measured. Here we have developed the Representative Sample Volume Predictor (RSVP) tool, which standardises statistical prediction of sufficient sample volumes to ensure microplastics are detected with a given level of confidence. Reviewing reports in freshwater, we found ~12% of observations reported sample volumes which would have a false negative error rate >5%. Such sample volumes run the risk of wrongly concluding that microplastics are absent in samples and are not sufficient to be quantitative. The RSVP tool also provides a harmonised Poisson point process estimation of confidence intervals to test whether two observations are likely to be significantly different, even in the absence of replication. In this way, we demonstrate application of the tool to evaluate historic data but also to assist in new study designs to ensure that environmental microplastic exposure data is relevant and reliable. It can also be applied to other data for randomly dispersed events in space or time, and so has potential as a transdisciplinary tool.