In the context of global warming, it is important to accurately estimate the atmospheric CO2 content and its changes. CO2 over the ocean is the vital component of global atmospheric CO2, and it is necessary to monitor it with a high-frequency and full-coverage technology. The atmospheric CO2 content over the South China Sea is affected by continental, oceanic and atmospheric factors. In this paper, a random-forest-based model of atmospheric CO2 column concentration over the South China Sea was built with chlorophyll-a concentration, instantaneous photosynthetically active radiation, particulate inorganic carbon, particulate organic carbon, sea surface temperature, wind speed and wind direction, which were from multisource satellites remote sensing data. The accuracy of the model was verified for the year 2020, with Bias of 0.27ppm, R2 of 0.59 and the RMSE is 1.00 ppm, showing the accuracy is satisfactory. The results show that the atmospheric CO2 column concentration in the South China Sea presents obvious seasonal characteristics, which are spring>summer>winter>autumn. Moreover, the main influencing factors causing the seasonal differences of atmospheric CO2 column concentration in the South China Sea vary with time. In January and April, the main influencing factors are wind direction. In July, wind speed and wind direction are the two most influential factors. In October, sea surface temperature is the most influential factor. This method can realize the high-frequency and full-coverage monitoring of atmospheric CO2 column concentration in the South China Sea. The relevant results can provide help for the realization of carbon neutrality goal.