Abstract:Individual tree biomass equations have been frequently used in ecological and forestry research over the last 60 years. They represent a powerful tool to understand forest productivity, nutrient cycling, and carbon sequestration, and they are used to estimate other structural and functional characteristics of forest ecosystems. Current attempts to develop above-ground biomass equations for Larix forests in northeast China have been mainly focused on only one species or applied to the genus Larix as a whole. However, generalized above-ground biomass equations for Larix could be used to estimate the average relationship between above-ground biomass and different independent variables and also variations among different Larix species. We developed generalized biomass equations for different Larix species by using Larix olgensis and Larix gmelinii. In this study, a total of nine tree variables that were able to predict above-ground biomass in Larix species were examined using biomass equations. The results show that D, H, and CW contributed significantly to predict above-ground biomass. Therefore, three combinations of these variables, including D alone; D and H; and D, H, and CW, were selected as independent variables to develop univariate, bivariate, and trivariate biomass generalized equations, respectively. The trivariate biomass generalized equations predicted above-ground biomass better than the other two equation types, while the predictive power of the univariate equation was the worse than the rest. Theoretically, the prediction accuracy of trivariate biomass equations could be further increased by adding stands or tree variables; however, including an excessive number of parameters in the biomass equations may hinder computation convergence and reduce the speed required to estimate model parameters. Furthermore, including many stands or tree variables would increase the cost and time required to conduct forest inventories. Therefore, determining the appropriate number of independent variables able to provide the level of accuracy required by forest managers is essential in forest modeling. A parsimonious model with reliable accuracy of prediction has been suggested as a reasonable approach for efficient forest management. For this reason, D, H, and CW were finally selected as independent variables for the generalized biomass equations developed in this study. In general, the biomass of individual trees with the same D would depend on the region studied and the origin of the tree. Thus, the generalized above-ground biomass equations developed for different Larix species in northeast China consider this inter-regional variation by using a dummy variable. To reduce heteroskedasticity in the data, we used weighted least square regressions. The results showed that the predictive precision of the biomass equations could be improved by adding predictor variables. Regardless of the traditional biomass equation used, both generalized equations considering only tree species and those considering tree species, tree origin, and region showed the highest prediction power. In addition, the accuracy for predicting above-ground biomass did not differ among univariate, bivariate, and trivariate equations when tree species, tree origin, and region were considered. On the basis of these results, the trivariate generalized biomass equation that considers tree species, tree origin, and region was believed to be the best option for estimating the biomass of L. olgensis and L. gmelinii.