Spatial heterogeneity, source apportionment and ecological risks of trace metal(loid)s in lakes across China.

Anthropogenic activities increasingly influence trace metal(loid)s (TMs) inputs to inland waters, yet the large-scale drivers of their distribution and ecological risks remain insufficiently understood. Here, we quantified 13 TMs in surface waters from 45 lakes across China to investigate spatial patterns, identify pollution sources, evaluate ecological risks, and assess predictive performance using machine learning models. TMs concentrations showed clear regional clustering, with southern lakes enriched in Ag, Cu, Fe, and Mn, while northern and northeastern lakes exhibited higher levels of As, Cr, and Ni. Positive matrix factorization indicated that Cu and Ni originated from combined agricultural, industrial, and natural sources; Cr was predominantly controlled by natural sources; and As reflected mixed contributions from natural, agricultural, and industrial inputs. Species sensitivity distribution (SSD)-based assessment identified Ag and Cu as the primary ecological risk contributors, with moderate risks occurring in 82.22% and 75.56% of lakes, respectively, and high risks observed in up to 24.44% of lakes. Risk quotient (RQ) analysis further revealed localized ecological risk hotspots for Cu, Fe, and Mn in northeastern China rather than widespread elevated risks. To support large-scale monitoring and prediction, machine learning models were developed. Among the tested algorithms, Random Forest achieved the highest predictive accuracy, highlighting the potential of machine learning approaches for as supportive tools for preliminary screening of lakes that may require enhanced monitoring or further pollution-source investigation. These results provide a large-scale cross-regional perspective on TMs contamination and support risk-based management of lake ecosystems.