Measurement and analysis of CO2 concentration at the ecosystem scale is the basis for studying ecosystem feedback to global climate change, and it is particularly useful for understanding the processes and mechanisms of ecosystem C exchange. Through observation of CO2 concentration at different heights, this study examined whether a CO2 lake phenomenon exists in the Larix gmelinii ecosystem of the Greater Khingan Mountains (China), and how it might be changed and what might represent its driving factors if such a phenomenon were found to exist. Plants and soils were sampled regularly to determine δ13C and to quantify the proportion of C released by each component of the ecosystem. The main path of C release and the main source of CO2 lake formation were investigated. Statistical analysis revealed that a CO2 lake phenomenon does exist in the L. gmelinii ecosystem. Comparative analysis showed that on the daily scale, when the ecosystem was a C source, the CO2 lake phenomenon often occurred. On the scale of the growing season, the strongest CO2 lake was accompanied by emergence of the peak respiratory flux. Stepwise regression analysis showed that environmental factors could explain 74.87% of the CO2 lake phenomenon. The occurrence and strength of the CO2 lakes were found to mainly respond to changes in temperature. Linear model analysis revealed that the rate of C release from autotrophic respiration in the forest was 51.18%; the rate of C release from heterotrophic respiration during litter decomposition was 51.78%. Therefore, the C release of the L. gmelinii ecosystem is mainly from autotrophic respiration. The CO2 released during decomposition of litter represented the main source for the formation of CO2 lakes. The CO2 lake effect has substantial impact on the net C flux and plays an important role in the C source/sink effect of the ecosystem.