Future mercury emissions from coal-fired power plants in China under dual-carbon targets: An emission-driven sensitivity assessment integrating machine learning and GEOS-Chem modeling.

Mercury (Hg) is a highly toxic trace metal that poses severe threats to ecosystems and human health. Coal-fired power plants (CFPPs) have historically been a major anthropogenic source of Hg emissions in China, but their long-term evolution under China's dual-carbon targets remains poorly quantified. Here, we develop an integrated framework that combines a machine-learning-based emission prediction model with GEOS-Chem to conduct an emission-driven sensitivity assessment of future CFPP Hg emissions and their atmospheric impacts. A Light Gradient Boosting Machine (LightGBM) model was trained using EDGAR inventory (2000 to 2022) to capture relationships between speciated Hg emissions and air pollutants, and was driven by scenario-based projections (baseline and dual-carbon scenarios) of air pollutant emissions from the Dynamic Projection for Emissions in China (DPEC) database for 2025-2060, which were implemented in the GEOS-Chem model under constant meteorological conditions to simulate changes in CFPP-contributed atmospheric Hg concentrations and deposition. Under the baseline scenario, CFPP Hg emissions decline from 16.1 t in 2025 to 12.9 t in 2060, whereas they decreased sharply to 0.93 t by 2060 (∼ 93% reduction) under the dual-carbon scenario. These reductions lead to substantial nationwide decreases in CFPP-contributed atmospheric Hg concentrations and deposition, with the strongest declines in eastern and southern China. On average, CFPP-contributed atmospheric Hg concentrations and total deposition decreased by 51% and 38% under the baseline scenario, while they decreased by 78% and 54% under the dual-carbon scenario, respectively. Our results suggest that China's dual-carbon strategy has the potential to deliver substantial Hg mitigation co-benefits under the assumed future emission scenarios.