Abstract: Objective  To establish stand biomass growth models and estimate the forest biomass on the regional scale so as to provide supports and references for carbon measurement and accounting. Method  The forest inventory data of five phases in Guangdong province (1997, 2002, 2007, 2012, and 2017) were used to select thirty fixed plots (with more than thirty reserved sample trees for five phases in each plot, totally 1 412 trees). Based on Richards theoretical growth function, the basic biomass growth model, stand-character-based biomass growth model, and stand-character-based and site-condition-based biomass growth model were established respectively using dummy variables. The biomass growth models established for stand level were evaluated and compared in this study. Result  The basic biomass growth model performed the worst with the lowest Ra2 of 0.475. When the stand density was added to the basic biomass growth model, the performance got improved greatly with the R2 of 0.836. When the dummy variable were added to the stand-character-based biomass growth function, the performance got best with the highest R2 of 0.961. Conclusion  Based on the stand characters, the biomass growth function can reflect the relationship between biomass growth and stand density to some extents and got better goodness-of-fit. The introduction of the dummy variable to the stand-character-based biomass growth functions can improve the fitting accuracy further, indicating the difference of stand condition and production among different stand types.