Abstract: With the flux data of ChinaFLUX and the concurrent satellite remote sensing data in Changbai Mountain, we recombined parameters of four models, i.e., vegetation photosynthesis model (VPM), eddy covariance-light utility efficiency model (EC-LUE), terrestrial ecosystem model (TEM) and Carnegie-Ames-Stanford approach model (CASA) within 3PG model. The most suitable parameters of 3PG model were determined by comparing the root mean square error, coefficient of determination and average error between measured and observed flux values. To improve its accuracy in estimating gross primary productivity (GPP) of broadleaved Korean pine forest in Changbai Mountain, the fitness of the optimal model was validated using the observed flux data. The results showed that when temperature, enhanced vegetation index, and surface water index were used to characterize the temperature limiting factor, photosynthetic active radiation absorption ratio and water limiting factor in the original model to estimate GPP of broadleaved Korean pine forest, the simulation results were the best. The precision of the optimized model (R²=0.948, RMSE=0.035 mol·m^-2·month^-1) was better than that of the original model (R²=0.854, RMSE=0.177 mol·m^-2·month^-1), which could effectively improve the phenomenon of obvious overestimation of the original model in the growing season. Results from the parameter sensitivity analysis showed that the uncertainty of GPP estimation was dominated by temperature, followed by enhanced vegetation index, photosynthetically active radiation and land surface water index, as well as their interactions.