Lake sediments on the Qinghai–Xizang Plateau capture shifting metal(loid) dynamics within an alpine ecosystem uniquely vulnerable to climate change and anthropogenic pressures. However, plateau-wide ecological risk assessments incorporating quantitative source apportionment remain scarce, leaving a critical gap that hinders targeted ecosystem conservation and precise policy interventions. Here we present a plateau-scale analysis of sediment geochemistry across 110 lakes over two decadal periods (2014 and 2024), utilizing a hybrid positive matrix factorization–random forest framework. Arsenic serves as the primary diagnostic sentinel, persistently exceeding baseline thresholds across both decades. Although mean chromium concentrations declined from 85.7 to 68.0 mg kg⁻¹, the broader multi-metal pool exhibits remarkable regional stability, with low baseline ecological risks dominated by cadmium toxicity alongside localized high-risk anomalies in Lakes Tosu, Gahai, and Xiligou. Hybrid source apportionment demonstrates that a mixed footprint of transport emissions, transboundary atmospheric influx, and lithogenic weathering increasingly drives geochemical variance, surging from 44% to 66% over ten years, whereas the independent arsenic source remains invariant at approximately 15%. These patterns reveal that while natural weathering constrains the regional background, localized anthropogenic signatures are rapidly intensifying around emerging infrastructure corridors. This predictive framework establishes a transferable template for tracking contaminant dynamics in vulnerable alpine headwaters, providing a vital baseline for targeted, source-specific mitigation globally.