[Effect of competition on the prediction accuracy of individual tree biomass model for natural Larix gmelinii forests].

Quantifying the impact of competition on individual tree biomass and its distribution pattern can provide a basis for improving the prediction accuracy of forest biomass models. To accurately quantify the effects of competition factors on individual biomass and its distribution, we constructed three different individual biomass models by using nonlinear coupling equations based on the biomass survey data of 50 Larix gmelinii from 18 plots of Pangu Forest Farm in Daxing'an Mountains. M-1 was a traditional singly additive biomass model. M-2 and M-3 were models taking the distance dependent simple competition index (CI) and distance independent relative diameter (R_d) into account, respectively. Those models were used to reveal the influence of competition factors on the prediction accuracy and distribution pattern of single tree biomass model of L. gmelinii. The results showed that the adjusted R² of three additive models ranged from 0.694 to 0.974, mean prediction errors ranged from -0.017 to 0.021, and mean absolute errors ranged from 0.152 to 0.357. The introduction of R_d could improve the fitting degree and prediction accuracy of most biomass models, but CI did not affect the model fitting effect and prediction ability. Among the three models, M-3 model had the best performance, with good fitting degree and prediction accuracy of the biomass of each part, which could accurately estimate the single tree biomass of L. gmelinii. Further simulation results showed that the variation of biomass with DBH was mainly affected by CI and R_d grade, and the influence of R_d was stronger than CI. CI had greater influence on root and dry biomass, but less influence on branch and leaf biomass. R_d had a more significant effect on biomass of branch and leaf than on that of root and trunk.