Wetlands are regarded as one of the largest ‘unknowns’ regarding future carbon (C) dynamics and greenhouse gas fluxes in the context of global change and climate policy-making. To understand the dynamics of carbon cycling of wetland ecosystem, we used an eddy covariance technique to measure the net ecosystem carbon dioxide (CO₂) exchange (NEE, positive or negative values of NEE represent net losses or gains of C, respectively, for the ecosystem), sensible heat flux (Hs) and latent heat flux (LE) between vegetation and the atmosphere at a reed wetland ecosystem in the Yellow River Delta during the periods of two growing seasons in 2009 and 2010. The total amount of rainfall in 2009 (571.4 mm) was higher than the annual average (551 mm). In contrast, precipitation in 2010 (523.5 mm) was significantly lower than average. The results from 2-year eddy tower observations showed that there was a dual peak in diurnal pattern of NEE fluxes for the reed wetland, which occurred at about 11:00 and 16:00, respectively. There were two different temporal patterns for the maximum diurnal uptake values of CO₂ in 2 years. The maximum diurnal uptake values of CO₂ were-0.30 mg CO₂ m^-2s^-1(July, 2009) and-0.37 mg CO₂ m^-2s^-1(June, 2010), respectively. The maximum diurnal emitting values of CO₂ were 0.19 and 0.25 mg CO₂ m^-2s^-1, respectively, and both occurred in September for 2 years. The diurnal patterns of Hs and LE were both single peaks, and their peak values both occurred at noon. The maximum latent heat flux was higher than the sensible heat flux, and the latent heat flux was the primary consumption component for net radiation during both two years. In diurnal scale, the heat fluxes were strongly negatively correlated to NEE fluxes (R²≥0.5, P < 0.0001). On the seasonal scale, the reed wetland was a strong C sink during the growing season. In 2009, the wetland ecosystem fixed 354.63 g CO₂ m^-2 in daytime of the whole growing season, and meanwhile it released 159.24 g CO₂/m² in nighttime. Approximate 651.13 g CO₂/m² was fixed by gross primary production (GPP), and 455.74 g CO₂/m² were released as ecosystem respiration (R_e), which resulted in a strong sink of atmospheric CO₂ with-195.39 g CO₂/m² sequestered in 2009 growing season based on the observed data. Path analysis results showed that the fluctuation of diurnal NEE fluxes was closely related to the photosynthetic active radiation (PAR) (R²=0.46-0.84)。However, soil temperature had the greatest effect on seasonal dynamics of ecosystem CO₂ exchange during the growing seasons at the study site, higher than the contributions of the other environmental factors such as precipitation and PAR. Our results strongly suggested that the combination of temperature, precipitation and PAR, as well as phonological stage of vegetation, control the C dynamics of reed wetland ecosystem. Therefore, an accurate representation of these parameters is extremely valuable for developing accurate and predictive wetland C cycle models and for the success of forecasting carbon budgets of reed wetlands.