Abstract
Abstract
Subsampling strategies are commonly employed in microplastic research to reduce the analytical burden associated with time-intensive techniques such as microscopy and Fourier-transform infrared (FTIR) spectroscopy. However, these strategies are often applied without prior validation. This study combined a systematic literature review and numerical simulations to evaluate the effectiveness of subsampling strategies in FTIR-based microplastic analysis. The review considered subtidal marine studies published between 2019 and 2024, revealing 46% applied subsampling. Of these, 50.8% used a constrained-quota random selection (CQRS) approach to select a random subset of putative microplastics for FTIR confirmation, with one-third of the studies analysing fewer than 25% of items. Notably, no standardized method was applied across studies, not even the minimum-percentage approach for subset selection, thereby limiting comparability and robust data interpretation. In addition, terminology used to describe subsampling approaches was inconsistent, further hindering cross-study comparisons. To evaluate how CQRS influences data representativeness, numerical simulations were conducted using a fully characterised (FTIR) real-world dataset comprising 2,137 putative microplastics from eight subtidal matrices (surface water, mid-column water, sediment, fish, coral, sponge, sea squirt, and sea cucumber), applying subsampling thresholds of 25%, 50%, and 75% across 1,000 iterations per matrix. While polymer representativeness improved with increased subsample size, the relationship was non-linear across all matrices and, even at the 75% threshold, reliable representativeness was rarely achieved. Only 15% of polymer types met the effectiveness criterion in at least one of the 1,000 iterations, yet these were not consistently the most abundant polymers found in the original dataset. These findings expose the limitations of current subsampling practices and underscore the need to exercise caution when extrapolating subsampled data to full populations. By demonstrating the risks associated with subsampling, this study highlights the vulnerability of highly heterogeneous samples to misrepresentation and mischaracterization. If subsampling is unavoidable, selection of at least 50% of items represents a practical minimum, and subsampling outcomes and any extrapolations must be reported explicitly and clearly justified. Finally, methodological and technological innovation is urgently needed to improve sample clarification and polymer identification, ensuring the long-term reliability of data used to inform monitoring, mitigation, and regulatory decisions.