Drivers of soil microplastic contamination and machine learning-based abundance standardization: A global meta-analysis

Soil microplastics (MPs) pose escalating environmental threats, yet their global occurrence and drivers remain poorly understood due to fragmented monitoring data and methodological inconsistencies across sampling, processing, and analytical procedures. This study conducted a global meta-analysis of soil MPs contamination and machine learning-based abundance standardization by integrating 1247 monitoring data from 153 studies. The results revealed that methodological factors (51.75 %) dominated the variation of MPs abundance, surpassing environmental (35.18 %) and socioeconomic (13.07 %) influences. The color and polymer compositions of soil MPs were governed by land-use type. The proportion of small-sized MPs (SMPs, <0.1 mm) and MPs shape were determined by detection size limit and identification method, respectively. High-altitude and low-anthropogenic regions exhibited lower MPs abundance but a higher percentage of SMPs. The random forest model produced harmonized abundance estimates with a mean of 1503.20 items/kg (90 % prediction interval: 269.70-3866.52 items/kg) under scenario 1, representing the most common methodology. This can serve as the baseline for global soil MPs exposure levels. This study quantifies key drivers of soil MPs contamination and provides a machine learning-based approach for standardizing abundance data. Beyond abundance, advancing the comparability of MPs characteristics should be a priority for future research to enable holistic assessments.