文献类型：期刊文献

中文题名：基于胸高处边材面积、胸径和冠基部直径的杉木单木叶生物量预测模型

英文题名：Tree Leaf Biomass Models of Chinese fir Plantations Based on Sapwood Area and Diameter at Breast Height and Diameter at Crown Base

作者：屈彦成[1,2] 江怡航[1] 姜彦妍[1] 张建国[1] 罗安利[3] 张雄清[1,2]

第一作者：屈彦成

机构：[1]中国林业科学研究院林业研究所,国家林业和草原局林木培育重点实验室,北京100091;[2]南京林业大学,南方现代林业协,同创新中心,南京210037;[3]浏阳市森林资源监测事务中心,湖南省浏阳市林业局,浏阳410300

年份：2023

卷号：59

期号：7

起止页码：106-114

中文期刊名：林业科学

外文期刊名：Scientia Silvae Sinicae

收录：CSTPCD;;EI(收录号：20234014834486);Scopus;北大核心:【北大核心2020】；CSCD:【CSCD2023_2024】；

基金：国家自然科学基金面上项目“不同发育阶段杉木人工林林分叶生物量变化及其峰值与生长和地力维护的关系”(31971645)。

语种：中文

中文关键词：杉木;叶生物量;管道模型理论;冠基部直径;胸高处边材面积;胸径

外文关键词：Chinese fir;leaf biomass;pipe model theory;diameter at crown base;sapwood area at breast height;diameter at breast height

分类号：S758

摘要：【目的】基于多个变量分别构建杉木单木叶生物量预测模型,并选择出预测效果最佳的模型,为杉木叶生物量的精准预测提供参考。【方法】以21块不同林龄样地共63株解析木为例,分别基于胸高处边材面积、胸径和冠基部直径3个变量,考虑其他与叶生物量相关的单木和林分因子,以样地为随机效应因子构建非线性混合模型,采用指数函数、幂函数和常数加幂函数消除数据间的异方差性。根据模型评价指标赤池信息准则(AIC)、贝叶斯信息准则(BIC)和对数似然值(Log Likelihood)选择最佳模型,并对不同参数的混合模型进行似然比检验。采用留一交叉验证法,计算模型决定系数(R^(2))、总相对误差(TRE)和平均绝对误差(MAE),对模型预测效果进行检验。【结果】基于3个变量以幂函数为异方差结构构建的混合模型效果最好,混合模型均优于基础模型,且基于冠基部直径构建的模型预测效果最佳。【结论】以基于冠基部直径构建的非线性混合效应模型(模型16)作为预测杉木单木叶生物量的最佳模型,符合管道模型理论。各变量均具有一定生物学和统计学意义,野外调查较易获取(非破坏性)。模型具有一定实用性,且预测精度较高(R2=0.8051)。本研究结果可为其他树种构建单木叶生物量模型提供参考。
【Objective】The tree leaf biomass models of Chinese fir plantations were developed based on multiple variables,and the best model was selected to provide reference for accurate prediction of leaf biomass of Chinese fir.【Method】Besides the three variables of sapwood area at breast height,diameter at breast height and diameter at crown base using 63 trees from 21 plots of different forest ages,the other variables related to leaf biomass were also considered.The nonlinear mixed model was considering the random effect of plot.In addition,exponential function,power function,and constant plus power function were used to eliminate the heteroscedasticity among the data.The best model was selected according to the model evaluation index AIC(Akaike information criterion),BIC(Bayesian information criterion),and Log Likelihood.The mixed model with different parameters was tested by likelihood ratio test.Finally,the leave-one-out cross-validation method was used to calculate coefficient of determination(R^(2)),total relative error(TRE),and mean absolute error(MAE)to test the models.【Result】The mixed models considered power function as heteroscedasticity structure performed the best among the three types of models.In addition,all the mixed models were better than the basic models,and the leaf biomass model developed based on diameter at crown base performed the best.【Conclusion】The nonlinear mixed effect model(Model 16)based on diameter at crown base with R2 values of 0.8051 was used as the final model for individual leaf biomass of Chinese fir plantations,which was consistent with the pipe model theory.All the variables had certain biological and statistical significance and were easy to obtain in the field work(non-destructive).In addition,this study can also provide a reference for other tree species in predicting individual leaf biomass.