Moisture effect on the carbon balance of terrestrial ecosystems is a key issue in global change research. It is crucial to accurately analyze the response of terrestrial ecosystem carbon cycle to moisture. However, the carbon flux models responding to environmental factors rarely consider the moisture effects on photosynthesis and respiration simultaneously; meanwhile there are still large uncertainties in model structures and parameters. Thus, this study was designed to (1) choose the optimal carbon flux model with accurate parameters for different ecosystems through model-data fusion approach, reducing the uncertainties of modeled results; (2) systematically analyze the influence of water factors on carbon flux simulation, including gross ecosystem productivity (GEP), ecosystem respiration (RE) and net ecosystem exchange (NEE). To consider the effects of moisture on both GEP and RE, we developed four different NEE models. Then, based on carbon flux and meteorological data during growing season from 2003 to 2009 in Changbaishan temperate mixed forest (CBS) and Qianyanzhou subtropical coniferous plantation (QYZ), Markov Chain Monte Carlo was employed to estimate model parameters, and Bayesian Information Criterion was applied to choose the optimal model for two forest ecosystems. The results showed that (1) the posterior values of model parameters were normally distributed, indicating that the parameters were well constrained by NEE. Photosynthetic and respiratory parameter values of CBS were higher than those of QYZ during the growing season. The model without vapor pressure deficit (VPD) overestimated the value of temperature sensitivity (Q₁₀) and underestimated the value of basal respiration rate (B_R) in QYZ; (2) the model considering VPD only was the optimal model for CBS,but its performance was not improved much. The modeled flux components were similar among the four models; (3) the model considering both VPD and soil water content (S_w) was the optimal model for QYZ, and its performance was improved significantly. The model ignored water factors overestimated 2% (21.85 g C/m²) of the total GEP, and 4.4% (38.02 g C/m²) of the total RE, and therefore, underestimated 7.8% (18.55 g C/m²) of the total measured NEE during the growing season.