A mechanistic framework to quantify sampling uncertainty of soil microplastics: integrating spatial heterogeneity and subsampling effects.

Despite the rapid growth of studies and datasets, the statistical representativeness of reported soil microplastic (MP) concentrations remains poorly understood, limiting the interpretation of environmental surveys and risk assessments. Here, we develop a mechanistic framework that integrates (i) a multinomial patch model describing spatial heterogeneity during field sampling and (ii) a Poisson model describing random counting uncertainty during laboratory subsampling. Using realistic ranges of contamination (0.022-0.22 g PET kg-1), particle size (1-1000 μm), patch volume (1-1000 mm3), and sampled mass (10-5000 g), we quantify total sampling uncertainty and the probability of false negatives. Fine particles (≤100 μm) reach acceptable uncertainty (CV <20%) with ≥100 g of soil, whereas coarse particles (≥500 μm) can exhibit extreme variability (CV >100%) under low abundance and unfavourable sampling conditions. Under such conditions, false-negative probabilities may become very high (up to ∼80%) when sample volumes are limited. Laboratory subsampling introduces a second statistical constraint: for large particles, typical aliquots of only a few grams often contain too few items to substantially reduce Poisson uncertainty under low-to-moderate contamination. Our results define quantitative relationships linking particle size, sampled mass, and spatial structure, and explain the systematic under-representation of large MPs in soil datasets. This framework provides operational guidance for designing more representative soil MP sampling strategies.