A multivariate beta mixed-effects approach to modeling stem carbon concentration of Larix olgensis in northeastern China

Accurate carbon stock estimation is essential for supporting forest management and climate-related decision making, yet most regional assessments continue to apply uniform carbon concentration values. To improve estimation accuracy, we analyzed over 5,600+ samples from 194 destructively harvested Changbai larch (Larix olgensis) trees to quantify intra- and inter-component variation across three components: sapwood, heartwood, and bark. Significant differences in carbon concentration were found across components (p < 0.05). And the stem showed significant vertical variation in carbon concentration. Tree diameter at breast height (DBH), Relative DBH (Rd), tree age (Age) and mean DBH of dominant trees (Ddom) exerted significant influences on the carbon concentration across all tree components. A multivariate beta mixed-effects model (MBMM) was developed to simultaneously modeling carbon concentrations of multiple components while accounting for the nested sampling design. Incorporating tree- and stand-level random effects greatly improved model fit (R² increased from 0.5 to 0.8 roughly) and resulting in a reduction in mean absolute prediction error by 15% for sapwood and heartwood, 10% for bark, nearly, when compared to the fixed-effects models. These findings provide empirically based carbon parameters for L. olgensis plantations and highlight the importance of component-specific values in forest carbon accounting. Adoption of refined component-specific carbon concentrations can enhance the accuracy of carbon offset assessments, forest asset valuation, and monitoring frameworks, supporting policy implementation under national carbon neutrality goals.