Abstract: The Shennongjia Forestry District of Hubei Province has a well preserved primeval forest at the mid-latitude region in the world, and it is the only administrative unit named by forest in China. Quantifying the net primary productivity (NPP) of the forest ecosystems plays an important role in the county ecosystem assessment. Based on the CEVSA2 (carbon exchange between vegetation, soil and atmosphere 2) model, we simulated the NPP of forests during 1981-2015, and verified the model performance using the field survey data. Then we analyzed the spatio-temporal variations of NPP and its correlation with climate factors. The results showed that:(1) The average NPP was 628.27 g/m² during 1981-2015, and the spatial distribution had significant spatial heterogeneity, lower in the central part and higher in the surrounding and east area. (2) The mean increasing rate of NPP was 2.58 g/(m²·a) (R²=0.65, P < 0.001) during 1981-2015, and the turning point was found around 1998 using Mann-Kendall method. The NPP showed an upward trend in both periods, but the increasing rate decreased by 7.01% in the latter period. (3) The increasing rate of the middle and northern areas was relatively high (4-6 g/(m²·a)), while the southwestern and eastern parts showed downward trends (-3-0 g/(m²·a)). The inter-annual variation of NPP had a positive correlation with temperature and global radiation, but a negative correlation with precipitation. The inter-annual variation of temperature has the highest explanatory rate of NPP, which is 43% (P < 0.01). In the spatial scale, the inter-annual variation of NPP in about 67.83% of the study area, located in the central and southeastern parts, was controlled by temperature. We can see that temperature was the main environmental factor driving the spatial and temporal variations of the NPP in the Shennongjia Forestry District.