CORTEXA
← Browse
semantic_scholarWater Research2026-08-15

Microplastics deplete soil available nutrients: a global meta-analysis with machine learning reveals critical thresholds and interactive controls.

Yangzhou Xiang, J. Peñuelas, M. Rillig, Xuqiang Luo, Luca Nizzetto, J. Akkanen, Ying Liu, Yang Luo, Bin Yao, Yuan Li

Microplastics (MPs) contamination threatens soil nutrient bioavailability, yet quantitative understanding of effect magnitudes and controlling factors remains fragmented across diverse environmental contexts. Here, we synthesized 206 studies (2853 observations) spanning 12 countries and diverse agricultural systems (soil pH 4.5-8.5; organic matter 5-80 g kg-1) to quantify MPs impacts on soil available nitrogen (AN), phosphorus (AP), and potassium (AK). We integrated random-effects meta-analysis using log response ratios (lnRR) with eXtreme Gradient Boosting (XGBoost) machine learning interpreted through SHapley Additive exPlanations (SHAP) to identify non-linear relationships and interactive controls. The dataset comprised predominantly controlled experiments (pot: 49%; incubation: 50%; field: 1%), with geographical concentration in East Asia (92%). Across all experimental setups, MPs significantly depleted AN by 5.3%, AP by 9.4%, and AK by 7.0% relative to controls. However, effects were context-dependent: pot/incubation studies showed significant depletion, while field studies reported no significant changes. XGBoost models (R2 = 0.44-0.62) indicated hierarchical control structures, with soil pH dominating AN responses (relative importance: 18.4%) and MPs concentration governing AP (15.2%) and AK (14.8%). SHAP analysis identified critical thresholds: particle size effects plateaued above 100 μm, concentration impacts saturated beyond 10 g kg-1, and pH-dependent reversals shifted from depletion in acidic soils to enrichment under alkaline conditions. Biodegradable plastics (e.g., PLA, PBAT) caused equal or greater short-term nutrient immobilization than conventional polymers, though mechanisms likely differ: biodegradable MPs may stimulate microbial biomass growth that temporarily sequesters nutrients, whereas conventional plastics primarily act through surface sorption. Partial dependence analysis indicated synergistic interactions: small particles (<50 μm) combined with high concentrations (>30 g kg-1) induced depletion 1.5-2.3-fold beyond additive expectations. These findings establish quantitative thresholds for risk assessment and demonstrate that MPs-induced nutrient limitation, observed primarily in controlled settings, could affect crop yields, necessitating interventions to prevent irreversible soil fertility degradation, though field-scale validation remains essential.

Related papers

semantic_scholarWater Research2026-08-15

Comparing AI-driven approaches for predicting river water quality: a systematic review of water quality indices and remote sensing methods.

Gomal Amin, O. Pourret, Victor Dupin, S. Guérin-Rechdaoui, Arnaud Dujany

TL;DR: Overall, this review identifies inconsistent preprocessing protocols, and the absence of standardised benchmarking as critical barriers to the transferability and operational reliability of current RWQ prediction frameworks.

River water quality (RWQ) monitoring faces growing challenges from geographical data gaps, methodological inconsistencies, and limited uncertainty quantification in predictive modelling frameworks. This systematic review aims to critically synthesise, compare, and identify knowle…

semantic_scholarWater Research2026-08-15

Towards scalable deep learning for automated microscopy in harmful algal bloom monitoring: Data-centric workflow and multi-region generalisation.

Negar Taheriashtiani, Glenn B. McGregor, N. Crosbie, Peter Hobson, Elloise Trotta, A. Lintern, et al.

TL;DR: A multi-region microscopy dataset comprising 105 microalgae/cyanobacteria taxa collected from three Australian regions is introduced, capturing diversity and challenges of operational monitoring, demonstrating that a structured data-centric workflow can substantially enhance automated microscopy, yet domain generalisation remains a critical bottleneck for deployment.

Harmful algal blooms (HABs) pose growing risks to drinking-water supplies and ecosystems, yet routine monitoring remains dependent on manual light microscopy. Deep learning offers a potential rapid alternative for automating microscopy, but progress has been constrained by ideali…