Spatiotemporal bias in OCO-2 XCO2 retrievals relative to TCCON observations

Accurate monitoring of atmospheric CO₂ concentrations is essential for understanding carbon dynamics and informing climate-mitigation strategies. This study examines the seasonal variability of satellite-based XCO2 retrievals from NASA's OCO-2 mission relative to ground-based TCCON XCO2 measurements over ten years (2014-2024) across 17 global TCCON sites. The study distinguished between two satellite observation modes, nadir and glint, and assessed how seasonal variations in satellite-derived environmental parameters – normalized difference vegetation index (NDVI), soil surface moisture (SSM), land-surface temperature (LST), and evapotranspiration (ET) – relate to statistical characteristics of the XCO₂ bias. In addition, we evaluated the temporal dynamics of the bias across the dominant land-cover types surrounding each site. Our results identify a latitude-dependent bias in nadir retrievals across Northern Hemisphere sites, with strong associations between nadir bias and land-cover at TCCON sites located > 45°N (R2 = 0.95) and between 0-45°N (R2 = 0.78). This work provides novel insights into the temporal dynamics of satellite XCO2 bias relative to ground measurements and demonstrates distinct behaviors across observation modes that have not been reported previously. These findings are crucial for improving the accuracy of global CO2 mapping and for strengthening the reliability of carbon flux models.