Abstract: Spatiotemporal dynamics of atmospheric CO₂ concentrations are crucial to understanding global carbon cycle and predicting regional climate change. Given that China spans multiple climatic zones and features complex and diverse ecosystems, elucidating the patterns of CO₂ differentiation across climatic zones are essential for formulating refined and differentiated reduction strategies. Here, we investigated the spatiotemporal variations of atmospheric CO₂ concentrations in China by using the satellite-retrieved XCO₂ datasets from 2003 to 2023 and the Köppen-Geiger climate classification. The results showed that the average of CO₂ concentrations ranked as: tempe-rate zone>tropical zone>cold zone≈arid zone>polar zone, while its growth rate ranked as: tropical zone>temperate zone>polar zone>cold zone≈arid zone. CO₂ concentrations had increased synchronously across climatic zones, yet the interregional CO₂ concentration differences had further widened, revealing an intensifying trend of spatial differentiation. Nearly half of the regions exhibited significant spatial clustering of atmospheric CO₂ concentration. The high-high clusters were primarily concentrated in the temperate zone, which was densely populated and economically developed. The low-low clusters were mainly distributed in the cold and polar zones, where ecological conservation was relatively robust. The spatial clustering was weakest in 2006 (Moran’s I=0.68, Z=70.00) and peaked in 2017 (Moran’s I=0.88, Z=90.33). While downward shortwave radiation was negatively correlated with CO₂ concentration (r=-0.542), population density (r=0.992), carbon emissions (r=0.970), leaf area index (r=0.845), temperature (r=0.565), and precipitation (r=0.481) all exhibited significant positive correlations with CO₂ concentration. Soil water content (r=0.072) showed a non-significant positive correlation with CO₂ concentration. According to the optimal parameters-based geographical detector, population density exerted the strongest influence on the spatial pattern of CO₂ concentration (q=0.55), followed by temperature (q=0.43). The interaction types between factors were predominantly characterized by bi-linear or nonlinear enhancement.

Key words: remote sensing, atmospheric CO₂ concentration, Köppen-Geiger climate classification, spatial autocorrelation, optimal parameters-based geographical detector