文献类型：期刊文献

中文题名：我国尺度兼容和树种分类的材积源森林碳储量模型

英文题名：Scale-Compatible and Tree Species-Classified Forest Carbon Storage Model of Volume-Derived in China

作者：张聪[1] 刘琪[1] 李海奎[1] 刘鹏举[1] 詹思颖[1]

第一作者：张聪

机构：[1]中国林业科学研究院资源信息研究所、国家林业和草原局森林经营与生长模拟重点实验室,北京100091

年份：2025

卷号：61

期号：1

起止页码：57-69

中文期刊名：林业科学

外文期刊名：Scientia Silvae Sinicae

收录：;EI(收录号：20250717889601);北大核心:【北大核心2023】；

基金：国家重点研发计划项目(2021YFD2200404);国家自然科学基金项目(42141004)。

语种：中文

中文关键词：森林碳储量模型;非线性似然无关;哑变量;参数库;尺度兼容;树种分类

外文关键词：forest carbon storage model;nonlinear likelihood independent;dumb variable;parameter database;scale-compatible;tree species-classified

分类号：S757

摘要：【目的】提出一种简单方便的森林碳储量估算方法,构建考虑林分特征的尺度兼容和树种分类的材积源森林碳储量模型,为估算多尺度和多树种森林碳储量提供方法和技术支持。【方法】基于第6~9次全国森林资源清查数据和异速生长方程,分别利用含哑变量的非线性最小二乘法的独立模型和非线性似然无关回归的联立方程组模型,构建考虑起源、龄组2个主要林分特征的尺度兼容和树种分类的森林碳储量模型,通过加权回归消除异方差,采用决定系数(R^(2))、估计值的标准差(SEE)、平均预估误差(MPE)、总相对误差(TRE)和差异百分比(VP)对模型进行评价;同时利用2021年林草综合监测数据,比较不同尺度模型估算全国森林碳储量的差异。【结果】1)共构建2974类尺度兼容的森林碳储量模型,与独立模型相比,联立方程组模型的R^(2)无明显差异。独立模型和联立方程组模型分别为1383和1591类,模型R^(2)的平均值分别为0.9661和0.9652,MPE分别为0.75%和0.78%,联立方程组模型的R^(2)仅下降0.0009,MPE仅上升0.03%。2)共构建2520类树种分类的森林碳储量模型,与尺度兼容模型结果一样,独立模型和联立方程组模型的R^(2)无明显差异。独立模型和联立方程组模型均为1260类,模型R^(2)的平均值分别为0.9443和0.9424,MPE分别为0.48%和0.49%,联立方程组模型的R^(2)仅下降0.0019,MPE仅上升0.01%。3)构建4种不同建模方式(独立-尺度模型、独立-树种模型、联立-尺度模型、联立-树种模型)的森林碳储量模型。相比独立模型,联立方程组模型的参数变动幅度更小。4种不同建模方式共包含参数a和参数b分别为46157和23935个。独立模型和联立方程组模型参数a的平均值分别为0.5965和0.6200,极差分别为2.3186和2.1922,独立模型的参数极差偏高0.1264;参数b的平均值分别为0.9332和0.9318,极差分别为0.6723和0.5065,独立模型的参数极差偏高0.1667。4)不同尺度模型估算全国森林碳储量时,无论何种尺度,独立模型的估算差异均大于联立方程组模型,但总体上各种尺度模型的估算差异均在3%以内。【结论】1)本研究提出的从森林蓄积量直接到森林碳储量的材积源森林碳储量模型数据有效、方法可靠,可用于直接估算森林碳储量。2)基于含哑变量的非线性似然无关的联立方程组方法,可更好地建立尺度兼容和树种分类的森林碳储量模型。3)本研究构建的森林碳储量模型平均R^(2)在0.95以上,MPE在1%以内,可用于林业实践中快速准确估算森林碳储量。4)根据模型的拟合精度以及参数的稳定性,建议使用以联立-尺度(以尺度为建模总体的联立树种分类模型)为建模方式的森林碳储量模型。5)在5%精度要求下,可使用国家尺度考虑林分起源、龄组的树种分类模型估算全国森林碳储量。
【Objective】A simple and convenient method for estimating forest carbon storage was proposed,and a scale-compatible and tree species-classified forest carbon storage model of volume-derived was constructed,which provided a method and technology support for estimating forest carbon storage of multi-scale and multi-tree species.【Method】Based on the data of the 6th-9th national forest resource inventory and allometric growth equation,the scale-compatible and tree species-classified stand carbon storage model considering stand origin and age group was constructed by using the independent model of non-linear least square method and the simultaneous equations model of non-linear seemingly unrelated regressions with dummy variables.Heteroscedasticity was eliminated by weighted regression,and the model was evaluated by determination coefficient(R^(2)),standard error of estimate(SEE),mean prediction error(MPE),total relative error(TRE)and variance percentage(VP).Meanwhile,using the data of the National forest and grass ecological comprehensive monitoring in 2021,the differences of forest carbon storage estimation by different scale models were compared.【Result】1)A total of 2974 scale-compatible forest carbon storage models 2 were constructed.Compared with the independent model,there was no significant difference in R^(2) of the simultaneous equations model.The independent model and the simultaneous equations model were 1383 and 1591 categories,respectively.The average values of the model R^(2) were 0.9661 and 0.9652,MPE were 0.75%and 0.78%,respectively,the R^(2)of the simultaneous equations model only decreased by 0.0009,and the MPE only increased by 0.03%.2)A total of 2520 tree species-classified forest carbon storage models were constructed.As with the results of the scale-compatible model,there was no significant difference in R^(2) between the independent model and the simultaneous equations model.The independent model and the simultaneous equations 2 model were both 1260 categories.The average values of the model R^(2) were 0.9443 and 0.9424,MPE were 0.48%and 0.49%,2 respectively,the R^(2) of the simultaneous equations model only decreased by 0.0019,and the MPE only increased by 0.01%.3)Four forest carbon storage models with different modeling methods(independent-scale model,independent-tree species model,simultaneous-scale model,simultaneous-tree species model)were established.Compared with the independent model,the parameters variation of the simultaneous equations model was smaller.Four forest carbon storage models with different modeling methods contained 46157 and 23935 parameters a and b respectively.The average values of parameter a in the independent model and the simultaneous equations model were 0.5965 and 0.6200,respectively,and the ranges were 2.3186 and 2.1922,respectively.The range of parameters in the independent model was 0.1264 higher.The average values of parameter b were 0.9332 and 0.9318,respectively,and the ranges were 0.6723 and 0.5056,respectively.The range of the independent model was 0.1667 higher.4)When estimating national forest carbon storage by different scale models,regardless of the scale,the estimation difference of the independent model was higher than the simultaneous model.However,in general,the estimation differences at various scales were within 3%.【Conclusion】1)The volume-derived forest carbon storage model proposed in this paper from stand volume to stand carbon storage was effective and reliable,which can be used to directly estimate forest carbon storage.2)Based on the simultaneous equations model of non-linear seemingly unrelated regressions with dummy variables,the scale-2 compatible and tree species-classified forest carbon storage model can be better established.3)The average R of the forest carbon storage model constructed in this paper was above 0.95,and MPE was less than 1%,which can be used to quickly and accurately estimate forest carbon storage in forestry practice.4)According to the fitting accuracy of the model and the stability of the parameters,we recommend using the simultaneous-tree species model.5)Under the accuracy requirement of 5%,the national scale model can be used to estimate the national forest carbon storage.