Matthias C. Rillig (@mrillig.bsky.social)

Paper just online: Xiang et al. Global Meta-Analysis Integrated with Machine Learning Assesses Context-Dependent Microplastic Effects on Soil Microbial Biomass Carbon and Nitrogen | Environmental Science & Technology pubs.acs.org/doi/10.1021/...

loading . . .

Global Meta-Analysis Integrated with Machine Learning Assesses Context-Dependent Microplastic Effects on Soil Microbial Biomass Carbon and Nitrogen Microplastics (MPs) in soil can paradoxically stimulate microbial biomass in a highly context-dependent manner, potentially inducing decomposition and affecting carbon and nitrogen cycles. We conducted a global meta-analysis with 90 studies (710 observations of microbial biomass carbon (MBC), 354 of microbial biomass nitrogen (MBN)) integrated with machine learning to quantify MPs effects on soil microbial biomass. Field studies showed no significant effects, contrasting with controlled experiments where MPs increased MBC by 9.6% (95% CI: 7.2–11.9%) and MBN by 10.4% (6.8–14.0%). Biodegradable plastics (PBAT, PLA) induced stronger effects (36.1–67.6%) than conventional polymers (PE, PP, PS, PVC). Temperature emerged as the dominant factor, with a contrasting MPs effect on MBC (positive) and MBN (negative) at higher temperatures, suggesting potential decoupling of carbon and nitrogen cycles under warming conditions. Machine learning models (XGBoost, R2 = 0.62) significantly outperformed linear regressions (R2 = 0.02–0.05), revealing nonlinear responses and threshold effects. Stimulatory effects were most significant for medium-sized MPs (30–90 μm), at high concentrations (>10 g kg–1), and in soils with intermediate fertility, highlighting context-dependent risks to soil carbon and nitrogen cycling. https://pubs.acs.org/doi/10.1021/acs.est.5c12883