Abstract: Net primary productivity (NPP) of forests is one of the most key indicators for the carbon sequestration in forest ecosystems. This study aims to explore its spatiotemporal changes and driving factors, in order to assess the quality of forestry projects at regional. The forest NPP in Beijing was taken as the research object. The FPAR (fraction of photosynthetically active radiation) parameter was optimized in the Carnegie-Ames-Stanford Approach (CASA) model. The stand types were selected to estimate the forest NPP. The spatiotemporal variation of forest NPP was obtained using Theil-Sen analysis and Mann-Kendall test (Sen + MK analysis). A systematic investigation was then made to determine the driving factors of natural and human activities on the forest NPP. Geographical detector was also combined with the correlation analysis. The results show that: 1) The optimal CASA model was used to more accurately simulate the forest NPP, compared with the measured data. Specifically, the forest NPP was presented an upward trend, with an average annual value of 404.11 g/(m²·a). The forest NPP in the mountainous areas (433.50 g/(m²·a)) was higher than that in the plain areas (366.82 g/(m²·a)). A relative low decline was observed in the forest NPP in the mountainous areas from 2004 to 2006. Subsequently, a fluctuating upward tendency was then found between 2006 and 2019. Similarly, the forest NPP in the plain areas also exhibited a fluctuating upward trend. The growth rate shared a marked increase after 2012. 2) The spatial distribution of the forest NPP showed the pattern of "higher in the northwest and lower in the southeast". Forest NPP increased significantly in 73.70% of the regions, which were distributed mainly in the areas of the Taihang and Yanshan Mountains. In plain areas, the average annual forest NPP fell within the medium-value bracket of 300-500 g/(m²·a) and the low-value bracket below 300 g/(m²·a). In mountainous regions, the average change slope of forest NPP measured 10.43 g/(m²·a), with a coefficient of variation of 0.18. In plain areas, the average change slope of forest NPP was 9.89 g/(m²·a), and the coefficient of variation reached 0.25. 3) The dominant influencing factors of forest NPP in the mountainous areas were the vegetation type (q=0.434), air temperature (q=0.163), elevation (q=0.063), and radiation (q=0.042). Conversely, the dominant influencing factors were the vegetation type (q=0.116), air temperature (q=0.065), precipitation (q=0.047), and population density (q=0.040) in the plain areas. Among them, there was a positive correlation between the natural factors and the forest NPP. In the driving factors of human activity, the population density and GDP showed a negative correlation with the forest NPP. Interaction analysis also demonstrated that there was the most pronounced impact of several combinations on the spatial distribution of the forest NPP in the mountainous areas, such as the vegetation type×air temperature (q=0.487), vegetation type×elevation (q=0.463), vegetation type×precipitation (q=0.455), and vegetation type×population (q=0.453). A greater influence was also observed in the plain areas, such as the vegetation type×air temperature (q=0.259), vegetation type×elevation (q=0.236), vegetation type×slope (q=0.203), and vegetation type×GDP (q=0.193). Furthermore, the combination of the vegetation type×air temperature was the interaction type with the highest explanatory power for the forest NPP. As such, multiple ecological projects were greatly contributed to the gradual decrease in the impact of human activities on the forest NPP. Future recommendations were also proposed to optimize the spatial layout of vegetation in the selection of vegetation types, according to the regional benefits of ecological projects.