Unraveling large-scale spatial patterns and key drivers of heavy metals uptake by plants

Heavy metals (HMs) contamination of soils is a critical environmental issue of global concern. HMs transfer from soil to plants is a key pathway for human exposure to soil HMs. Quantifying soil HMs uptake by plants, represented by the soil-to-root transfer coefficient (TC), is crucial for evaluating soil HMs risk and designing phytoremediation strategies. However, large-scale evaluations using laboratory-based techniques are impractical owing to their time-consuming and labor-intensive nature. In this study, a random forest model achieved the highest accuracy among seven machine learning algorithms using a compiled global dataset and enabling the first-ever spatial prediction of soil HMs uptake by plant roots. Extensive reliability and uncertainty analyses were performed to ensure model robustness across diverse environmental settings. Soil pH (35.0 % ± 2.7 %), HMs content (16.1 % ± 1.8 %), and organic carbon (OC, 9.4 % ± 0.9 %) were identified as key drivers of TC. Low pH significantly increased HMs uptake, while high OC levels enhanced bioavailability. Spatial prediction distribution of HMs uptake by plant roots across European soils revealed hotspots in Italy and Austria that were co-driven by HMs content and TC values. These results offer important insights into the spatial patterns of HMs uptake and their underlying drivers, facilitating the identification of contamination hotspots and guiding effective strategies for health risk mitigation and phytoremediation. • Develop an ecological indicator for soil-to-root heavy metals transfer using large-scale data. • Apply random forest model to predict plant uptake of heavy metals with improved accuracy. • Identify key drivers (soil pH, organic carbon, and heavy metals content) of uptake variability. • Map European hotspots of heavy metals uptake to support targeted risk assessment. • Provide an indicator-based framework for phytoremediation and soil management.