Abstract:Abies nephrolepis (Maxim) is an important coniferous tree species in the mixed broadleaved-Korean pine forests in the Changbai Mountain. In this study, the above- and belowground biomass allocation patterns of A. nephrolepis were analyzed using 21 harvested trees with different diameters at the breast height (DBH). At the branch level, allometric models for live branch and needle biomass were developed based on independent variables of branch diameter (BD), branch length (BL), and whorl position (WP). At the whole tree level, independent variables including DBH, tree height (H), tree age (Age), crown length (CL), crown ratio (CR), south-north crown width (CW1), and east-west crown width (CW2) were used to develop allometric models for biomass components of stem wood, bark, live branches, needle, coarse roots, and the whole tree. The best fitting models were identified by stepwise regression method. Results show that majority live branch biomass occurred in the middle and lower canopy layers, while the needle biomass was mostly allocated in the middle layer of the tree crown, with no significant difference between the middle and lower layers in the combined biomass of live branches and needle (P>0.05). The aboveground biomass and belowground biomass were 1.026-506.047 kg/tree and 0.241-112.000 kg/tree, respectively. Relative proportions of coarse roots, live branches, needle, stem wood, bark, and dead branches to total tree biomass were 18.68%, 18.39%, 12.02%, 39.29%, 8.70%, and 2.92%, respectively. There was a significant linear relationship between the aboveground biomass and belowground biomass (P<0.001). Slope of the fitted linear model was 0.23. At the branch level, allometric models of the live branch biomass explained more than 95% of the variations in data and the mean prediction error was less than 30%. Models based on two (i.e. BD and BL) or three variables (i.e. BD, BL, and WP) were better than the single-variable (i.e. BD) model, with the variability explainable increasing by 1.2% and 2.0% and the mean prediction error decreasing by 6.26% and 9.27%, respectively. Needle biomass was more difficult to estimate than the biomass of live branches. Allometric models of needle biomass explained only 82.7% of the data variability, with the mean prediction error reaching 50%. Compared with the allometric model based on a single variable, prediction accuracy improved little when including BL and WP variables for the needle biomass. The live branch biomass was positively related with BD, BL, and WP; whereas needle biomass was positively related with BD and negatively with BL and WP. At the tree level, the biomass allometric models based on DBH explained more than 90% of the data variability. Inclusion of tree height did not always improve biomass estimation. Stem biomass was age-related and biomass estimation without considering tree age could be slightly biased. Crown variables were very important to accurately estimate the biomass of live branches and needle. Considering the variability explainable and the significance of regression coefficient in allometric models, it can be concluded that DBH is a reliable predictor for estimating above- and belowground biomass in A. nephrolepis.