In recent years, the carbon cycle has been severely affected by global change and human being in semi-arid grassland ecosystems. The simulation of carbon flux mainly focus on the regions where observation data are available. However, few studies paid attention to the simulation of carbon flux in the regions which paucity of observation data. Herein, Hulun Buir grassland, which is lack of observation data, was selected as the main study area. Additionally, the vegetation photosynthesis and respiration model (VPRM) was used to simulate the spatial and temporal distribution of net ecosystem exchange (NEE) in 2016. The results showed that:(1) In 2016, Hulun Buir grassland was a weak carbon source under the extreme drought conditions (annual NEE value 47.27 g C/m²), and the trend of daily NEE of t was similar to those of precipitation and temperature. (2) The spatial distribution of NEE gradually decreased from the grassland to the meadow and forest areas. Based on vegetation distribution, The region in different vegetation type were ordered by carbon emission as follow:Stipa krylovii and Stipa grandis steppes > Leymus meadows > forb meadows,(dominated by plants such as Filifolium sibiricum), (3) Drought stress was one of the main reason for the region being a carbon source. There was a significant quadratic correlation between precipitation and NEE (R²=0.938, P < 0.001). The results implied that the grassland ecosystem tends to shift from a carbon source to a carbon sink with the increase in precipitation under drought conditions. (4) Aboveground biomass (AGB) showed a significant positive correlation with gross primary production (GPP) and ecosystem respiration (R_eco) (R²=0.89 and 0.9, respectively, P < 0.01), and a significant negative correlation (R²=0.68, P < 0.01) with NEE, indicating that above-ground biomass (AGB) growth can effectively reduce carbon dioxide emissions in grassland ecosystems.