Evaluation of Subsampling Methods for Micro-Spectroscopy Analysis of Microplastics

Full micro-spectroscopy analysis of micron-size microplastics (MPs) in environmental filter samples can be very time-consuming. Subsampling is necessary but leads to the introduction of uncertainty in the number, mass and type of MPs that have been collected on a filter. Commonly used random and stratified random subsampling makes minimal use of information about the sample and are compared here to subsampling approaches that make use of the particle number and size distribution of collected MPs obtained with fast pilot scans. These informed subsampling approaches bias the subsampling by proximity to mean count (PMC), proximity to mean size (PMS), proximity to areas with large counts (PPC), and proximity to areas with large size (PPS). These informed subsampling methods were benchmarked against Random and Stratified methods by Monte Carlo Simulations that concerned three metrics: (1) count estimation, (2) size profiling, and (3) identification profiling. For count estimation, the PMC method was marginally better than random methods. For size and identification profiling, the PMS outperformed all other methods. To reaffirm the practicality of the proposed methods, a demonstration was presented relying on PMC and PMS to execute the subsampling and estimate the overall count, size and identification of MPs. The results of this study make strong arguments for applying subsampling methods that make use of particle number and size distribution rather than random subsampling.