Currently, analyzing the impact of climate change on terrestrial ecosystem functions based on models is the focus of global change ecology. However, one of the model simulation uncertainties stems from the spatial heterogeneity. Spatial heterogeneity is a function of scale. In this paper, an ecological process- based model Boreal Ecosystem Productivity Simulator(BEPS) was used to simulate the daily Gross Primary Productivity (GPP) in the spatial resolutions of both 1 km and 8 km from 2003 to 2005 at four sites of ChinaFLUX, including Changbaishan (CBS), Qianyanzhou (QYZ), Haibei (HBGC) and Lasadangxiong (LSDX). In terms of Land Cover data and Leaf Area Index (LAI), we try to find how these differences influence the GPP simulation difference influenced by spatial resolutions of model inputs. The results show: (1) the finding that GPP simulations varied with spatial resolutions is mainly due to LAI diversity in the 8-km mixed pixels, the averaged absolute difference values of the LAI between 1 km and 8 km across the four sites are 0.85, 1.60, 0.13 and 0.04, respectively; (2) GPP simulations at the spatial resolution of both 1 km and 8 km could capture the GPP's seasonal dynamics across the four sites, the correlation coefficients (R~2) between the simulated and eddy covariance flux measurements, range from 0.79-0.97 (1 km), and 0.69-0.97 (8 km), and the absolute difference is 11.46-29.65 gC/m~2/month (1 km), and 11.87-24.81 gC/m~2/month(8 km); (3) the averaged monthly GPP absolute differences derived from spatial resolutions in the four sites are 14.43 (CBS), 12.05 (QYZ), 4.79 (HBGC) and 3.22 (LSDX) gC/m~2/month, in which greater differences were found at the forest site than at the grass site, and in growing season than in non- growing season. In conclusion, it is feasible to input coarser spatial resolutions data to improve the large- scales and long-term GPP simulations. Also, we should reduce the simulation differences at the forest sites as well as in the growing seasons.