Objective To improve the prediction accuracy of growth dynamics in oak stands, this study integrates tree-ring data and Bayesian calibration to parameterize the 3-PG model for Quercus aliena var. acuteserrata, and evaluates its applicability in predicting stand growth in both pure and mixed forests.

Method The study was conducted in 20 pure forest plots of Q. aliena var. acuteserrata and 5 mixed forest plots of Q. aliena var. acuteserrata and Q. mongolica. Tree-ring samples were collected from standard trees in each plot. Based on these samples, biomass and volume equations from different provinces were applied to estimate stem, bark, branch, foliage, and root biomass, as well as stand volume. The 3-PG model for Q. aliena var. acuteserrata was parameterized using Morris global sensitivity analysis and Bayesian calibration. Model validation was conducted using both pure and mixed forest plots. Regression analysis and accuracy evaluation were used to assess the consistency between model predictions and observations.

Result Sensitivity analysis indicated that parameters related to self-thinning, canopy characteristics, and root biomass allocation (such as wSx1000, alphaCx, and pRn) were highly sensitive. Bayesian calibration results showed that the simulated values of DBH, tree height, stem biomass, and root biomass were close to the observed values. After calibration, the coefficients of determination (R²) between predicted and observed values for DBH, stem biomass, root biomass, total biomass, and volume were all larger than 0.90, and the percentage bias (Pbias) was less than 6%. Validation results showed that stand density had the highest prediction accuracy (R² = 0.93, RMSE = 62.18, Pbias = −6.09%, MAPE = 6.09%) in pure forests. In mixed forests, stand density (R² = 0.99, RMSE = 78.60, Pbias = −4.73%, MAPE = 4.15%), DBH (R² = 0.90, RMSE = 1.66, Pbias = −5.45%, MAPE = 8.71%), and total biomass (R² = 0.87, RMSE = 25.42, Pbias = −10.90%, MAPE = 16.98%) showed relatively high prediction accuracy, whereas root biomass showed comparatively lower accuracy (R² = 0.66, RMSE = 5.34, Pbias = 10.05%, MAPE = 18.03%).

Conclusion The 3-PG model parameterized using tree-ring data and Bayesian methods can effectively simulate the stand growth dynamics of pure forests of Q. aliena var. acuteserrata, and also demonstrates good predictive performance under mixed forest conditions. The model provides a scientific basis for the management of oak stands.