江西省吉安地区火炬松地上生物量分配特征及模型研建

孟盛旺; 谷振军; 肖平江; 刘子荣; 于健; 彭小兵; 周光

doi:10.12171/j.1000-1522.20210400

江西省吉安地区火炬松地上生物量分配特征及模型研建

Characteristics of aboveground biomass allocation and model construction for loblolly pine in Ji’an Region, Jiangxi Province of eastern China

摘要

摘要:
目的探讨火炬松地上生物量分配模式随树木大小的变化规律，建立地上各组分及总量的可加性生物量模型系统，为江西省吉安地区火炬松人工林生物量和碳汇能力准确评估提供有效手段。
方法以安福县武功山林场的火炬松人工林为研究对象，分不同径阶利用收获方法共采集了35株火炬松的干材（去皮树干）、树皮、树枝和树叶生物量，通过计算各组分生物量占地上总量的比例，分析火炬松地上生物量的分配模式及其随树木大小的变化趋势。以胸径（D）、树高（H）为预测变量，探讨各组分的最优生物量模型形式，采用似乎不相关模型建立可加性生物量模型系统，并通过赋予特定的权函数消除模型异方差，利用留一交叉法对模型的预测效果进行验证。
结果随着林木个体的逐渐长大，干材和枝条生物量比例呈上升趋势，而树皮和树叶生物量比例呈下降趋势。D是各组分生物量模型最重要的预测变量，模型系数极显著且拟合效果良好（ R_\textadj^2 为0.91 ~ 0.97），H变量的加入有助于提高干材和树皮生物量模型的拟合优度，但不利于树枝和树叶模型的改进，且H作为独立变量时模型系数不显著。与国外火炬松生物量模型相比，本文模型预测效果良好（ R_\textadj^2 > 0.9），误差较小，总相对误差（TRE）基本在±1%以内。
结论林木个体大小是地上生物量分配特征的主要影响因素，地上各组分生物量占比大小依次为树干 > 树枝 > 树叶。干材和树皮生物量模型以D²H为自变量拟合效果最优，而树枝和树叶模型基于单一变量D即可获得最优效果，基于各组分最优预测变量建立的可加性生物量模型系统预测精度较高，可以在吉安地区火炬松人工林生物量测算中推广使用。

Abstract:
Objective This study aimed to analyze the aboveground biomass allocation patterns varying with tree size, and to establish additive allometric biomass equations for accurate estimation of biomass and carbon sequestration of loblolly pine plantations in Ji’an Region, Jiangxi Province of eastern China.
Method A total of 35 trees covering different diameter classes were harvested and measured for wood (inside bark), bark, branch, and foliage biomass in the Wugongshan Forest Farm of Anfu County of Jiangxi Province. The share of biomass allocated to different components and its variation trend with tree size were assessed by calculating the biomass fractions. The best biomass model for each component was determined by testing the DBH (D) and height (H) as predictors. A seemingly unrelated regression method with a unique weighting function for each equation was applied to establish additive systems of biomass models and to overcome the model heteroscedasticity. The predictive ability was validated by the leave-one-out cross-validation method.
Result With the increase of tree size, stem and branch biomass ratios increased and the ratios for bark and foliage decreased. Diameter proved to be the most important predictor for each biomass component, the model coefficients were extremely significant with good fitting results ( R_\textadj^2 ranged from 0.91 to 0.97). Adding tree height can help to improve the model fit and performance for wood and bark, but it is not conducive to the improvement of branch and leaf models, and the coefficient for tree height is non-significant. Compared with the previously published biomass models of loblolly pine, the models developed in this study performed well and had a small bias with total relative error (TRE) basically within ±1%.
Conclusion Tree size is the main factor affecting biomass allocation, and trunk accounted for most of the aboveground biomass, followed by branch and foliage. Wood and bark biomass models are best fitted with D²H, while models with D alone perform best for branch and foliage biomass. The developed additive biomass models based on the most suitable predictors show high prediction accuracy, indicating that the models can be applied in biomass estimation of loblolly pine plantations in Ji’an Region, Jiangxi Province of eastern China.

HTML全文

参考文献(43)

施引文献

资源附件(0)