Objective This study aims to enhance the accuracy of Net Primary Productivity (NPP) simulations by optimizing parameter estimation in the Biome-BGC model. It also seeks to analyze the pathways through which key sensitive parameters influence NPP.

Method The research focuses on the Korean pine-broadleaved forests in northeastern China. In the Biome-BGC model, 28 physiological and ecological parameters for conifers and broadleaved trees were optimized separately using the Parameter Estimation System (PEST), based on leaf area index data obtained from the Moderate-resolution Imaging Spectroradiometer (MODIS). A sensitivity analysis was conducted to identify parameters with sensitivity indices above 0.2. Simulation accuracy before and after optimization was evaluated using linear regression models. Additionally, structural equation models were employed to quantify the response of NPP to changes in highly sensitive parameters (sensitivity index > 0.2).

Result (1) Parameter optimization significantly improved the goodness of fit for NPP simulations (p < 0.01), with the coefficient of determination (R²) increasing from 0.15 to 0.31 and the root mean squared error (RMSE) decreasing by 59%. (2) The highly sensitive parameters shared by conifers and broadleaf trees include four key parameters: the annual fire mortality fraction, maximum stomatal conductance, boundary layer conductance, and the fraction of leaf nitrogen content in Rubisco. (3) Structural equation modeling indicated that leaf nitrogen content in Rubisco and maximum stomatal conductance are key parameters affecting simulated NPP. These parameters regulate photosynthetic capacity through carboxylation limitation and carbon dioxide diffusion. However, the regulatory effect of stomatal conductance on photosynthesis may vary with environmental conditions, potentially promoting or inhibiting photosynthetic activity.

Conclusion Integrating observational data with the PEST system, while focusing on critical parameters such as leaf nitrogen content in Rubisco and maximum stomatal conductance, can significantly improve the accuracy and efficiency of NPP simulations in the Biome-BGC model. This study provides a theoretical foundation and methodological support for vegetation parameterization and carbon cycle simulations in the Korean pine-broadleaved forest ecosystems in northeastern China.