Abstract: Based on data obtained from 572 branches of 45 Chinese fir trees in Jiangle Forest Farm, Fujian Province, southeast China, prediction models for branch, foliage biomass and total branch and foliage biomass of individual tree were developed by linear mixed effects (LME) method, and tested by independent samples. The results showed that the LME models provided better performance than the multiple linear regression models for the branch, foliage and total biomass prediction of Chinese fir plantation. The LME models with different combinations of the random effects parameters had different fitting precisions. The LME models including variance structures could effectively remove the heteroscedasticity in the data and improved the precision. The LME model with the exponential function as the variance structure had better fitting precisions for the total biomass and foliage biomass models, and that with the constant plus power function as the variance structure had better performance for the branch biomass model. Model validation confirmed that the LME models with the random effects and heteroscedasticity structure could significantly improve the precision of prediction, compared to the multiple linear regression models.