Abstract: Measured 8 individual standard wood carbon density of Pinus kesiya var. langbianensis, the variation had been analyzed between difference individual, height percent and parts of disk. And the individual wood carbon density models had been constructed based on the mixed-effects models technology. Results showed that there were significant differences in wood carbon density between different individuals. There is significantly different (P < 0.001) with the parts, and the wood densities increase first, then flat or slightly decrease from pith to outer of disk. It was verified that the estimation performance of mixed-effects modelsare better than ordinary models because of lower AIC and BIC values. And there was the lowest AIC and BIC, and the highest Loglik value in the two-levels mixed model with random effects of individual tree and parts, and the highest R² and RMSE in three-levels mixed models. The prediction precision of all mixed models were 95% above, and the two-levels model with random effects from individual tree and tree parts reached to 96.1%. So the mixed effects models had better performance than ordinary model for describing and predicting individual wood carbon density.