Enhancing\nSpace-Based Tracking of Fossil Fuel CO<sub>2</sub> Emissions via Synergistic\nIntegration of OCO-2, OCO-3, and\nTROPOMI Measurements

Top-down\nestimates of fossil fuel CO<sub>2</sub> (FFCO<sub>2</sub>) emissions\nare crucial for tracking emissions and evaluating mitigation\nstrategies. However, their practical application is hindered by limited\ndata coverage and overreliance on NOx-to-CO<sub>2</sub> emission ratios\nfrom emission inventories. We developed the Machine Learning-Driven\nMapping Satellite-based XCO<sub>2en</sub> (ML-MSXE) model using the\ncolumn-averaged dry-air mole fraction of CO<sub>2</sub> enhancement\n(XCO<sub>2en</sub>) derived from OCO-2 and OCO-3 measurements to reconstruct\nthe XCO<sub>2en</sub> distribution for monitoring FFCO<sub>2</sub> emissions. Compared to the previous Machine Learning-Driven Deriving\nXCO<sub>2en</sub> from Mapped XCO<sub>2</sub> (ML-DXEMX) model, ML-MSXE\nenhances the utilization of TROPOMI NO<sub>2</sub> measurements, increasing\ntheir relative contribution from 4.3 to 21.7%, thereby improving XCO<sub>2en</sub> reconstruction accuracy and enhancing the ability to track\nemissions. Despite the COVID-19 lockdown, XCO<sub>2en</sub> levels\nin China rose from 1.33 ± 1.06 in 2019 to 1.39 ± 1.01 ppm\nin 2021. In February 2020, while the national average rate of XCO<sub>2en</sub> decline (16.3%) aligned with the reduction in FFCO<sub>2</sub> emissions estimated by inventories, XCO<sub>2en</sub> further\nrevealed varying rates of decline between cities. Furthermore, the\nspatial distribution of XCO<sub>2en</sub> identified hotspots where\nFFCO<sub>2</sub> emissions might be underestimated by inventories.\nThis study presents a space-based approach for monitoring FFCO<sub>2</sub> emissions, offering valuable insights for assessing carbon\nneutrality progress and informing policy.