Study on CO 2 concentration assimilation by integrating a multi-source satellite fusion dataset based on the WRF-Chem model

Accurate monitoring of meteorological conditions and carbon dioxide (CO2) concentration variations is a key scientific foundation for understanding and controlling greenhouse gas emissions. This study developed a 3DVAR (Three-Dimensional Variational Data Assimilation) carbon assimilation system based on the WRF-Chem (Weather Research and Forecasting Model with Chemistry) model, which simultaneously assimilates meteorological observational data and a fused XCO2 product derived from multi-source satellite remote sensing data, including OCO-2 , OCO-3 , and GOSAT . The system was applied to simulate meteorological conditions and CO2 concentration distributions over central and eastern China from September 2018 to August 2020. In terms of meteorological simulation, relative humidity simulation improved significantly, with RMSE at Shanghai station dropping from 14.92% to 8.60%. The simulation of the wind direction showed modest improvement, with R increasing from 0.58–0.62 to 0.64–0.69 after assimilation. Temperature and atmospheric pressure simulations consistently maintained high accuracy, with R remaining at 0.97 for temperature and 0.99 for pressure both before and after the assimilation process. Regarding CO2 concentration simulation, the assimilation of XCO2 data enhanced the accuracy across all validation sites. At the TCCONstations, the system captured seasonal fluctuations – particularly in spring and summer 2020—and corrected systematic biases such as the underestimation observed in autumn 2019. As a result, the overall Bias improved from 0.376 ppm to 0.296 ppm, RMSE decreased from 3.151 ppm to 1.913 ppm, and R increased from 0.629 to 0.829. Meanwhile, simulations at the WDCGG sites achieved better accuracy, with Bias corrected from −0.344 ppm to −0.028 ppm, RMSE decreased from 2.898 ppm to 2.024 ppm, and R increased from 0.833 to 0.914. The simultaneous assimilation of meteorological and CO2 observations established in this study enables more accurate regional simulations of CO2 concentrations. This assimilation system provides reliable technical support for carbon cycle research.