Carbon dioxide is generally regarded as the most important greenhouse gas affecting global warming. Many researches have been conducted to measure atmospheric CO<sub>2</sub> concentration, analyze CO<sub>2</sub> variation on both seasonal and interannual scales and predict future CO<sub>2</sub> tendencies. Among them, ground-based remote sensing observation of CO<sub>2</sub> is the essential approach to provide validation data for satellite observation owning to its much higher accuracy and column CO<sub>2</sub> measurement capability. Unlike the Fourier Transform Spectrometer, the sun photometer observation system introduced in this paper takes advantages of full-automation, easily portable and mobility to provide a non-supervised automatic field operational CO<sub>2</sub> observation approach. In this study we acquired the CO<sub>2</sub> measurements from 2010 to 2012 in Beijing, translated them to cloud free data, and designed an index related to column CO<sub>2</sub> amount. The diurnal and seasonal variations of atmospheric CO<sub>2</sub> acquired by the system were also analyzed. We compared our observation with the simulation of CarbonTracker model, and good agreements with the model suggested the long-term stability of the system and reasonability of the data processing method. The future work will focus on absolute CO<sub>2</sub> quantity retrieval and accuracy assessment of this observation system.