Abstract: Gross primary productivity (GPP) plays an important role in global carbon cycle. Vegetation maximum light use efficiency (ε_max) is the key parameter for GPP simulation of terrestrial ecosystem. Based on the vegetation photosynthesis model (VPM) and the eddy covariance flux data at 40 stations from FLUXNET (179 site-years of data), we identified the key model parameters influencing the simulation of GPP with VPM through one-at-a-time (OAT) method. The cross validation method was employed to optimize the key model parameters and evaluate the model perfor-mance for global forest ecosystems. The results showed that the prediction of GPP was mostly affec-ted by ε_max, maximum temperature for photosynthesis (T_max), and optimum temperature for photosynthesis (T_opt). There were distinguishable differences for the key optimized parameters among different forest ecosystems. The optimized ε_max ranged from 0.05 to 0.08 μmol CO₂·μmol^-1 PAR (evergreen broad-leaved forest>evergreen coniferous forest>mixed forest>deciduous broad-leaved forest). The optimized T_max ranged from 38 to 48 ℃,while T_opt ranged from 18 to 22 ℃. With the optimized key parameters based on ecosystem types, the VPM was able to simulate the seasonal and inter-annual variations of GPP in four forest ecosystems.