摘要
【目的】探究地基激光雷达(Terrestrial laser scanning,TLS)对长白落叶松各相对高度处直径的提取精度,分析黑龙江省南部不同林分条件下长白落叶松的根径、材积之间的关系,验证不同建模方法在材积预测中的效果,为当林木仅剩伐根时预测单木材积提供科学依据。【方法】以黑龙江省孟家岗林场的8块长白落叶松Larix olgensis人工林样地为研究对象,利用地基激光雷达提取的根径(dst)与胸径(DBH)数据代入一元立木材积公式,计算单木材积(V)。在此基础上,采用非线性最小二乘法构建根径-材积基础回归模型;进一步引入林分密度(株数密度N)和林木年龄(AGE)作为协变量,构建广义线性模型(GLM);随后以样地为随机效应因子,构建线性混合效应模型(LME)。通过比较各模型的拟合精度与预测性能,并优选最优模型,最后采用分径阶检验法对模型的稳定性与适用性进行验证。【结果】TLS在根径处的提取精度为93.42%,对比胸径的提取精度(P=98.98%)下降了5.56%,40株解析木材积(V解)与一元材积式求算材积的相关性较高(R^(2)=0.9887),相对误差(RE)的平均值为9.19%;基础模型中,Berkhart模型的均方根误差最小(RMSE为0.1949),拟合效果最好(R^(2)=0.8452、R^(2)_(a)=0.8448);相关性分析中,林木材积与根径、林木年龄呈显著正相关,与株数密度呈显著负相关,林木年龄对林木材积的影响略大于株数密度;广义模型的R^(2)_(a)相比基础模型提高了7.98%,RMSE从0.1955下降到0.1949;混合效应模型中,随机效应在b1和b2位置表现最优,相比广义模型Ra 2提高了1.66%,RMSE从0.1955下降到0.1942;分径阶检验中,当11 cm≤dst<43 cm时,Berkhart模型、广义模型和混合效应模型的平均绝对误差(MAE)随着根径的增大而增大,且MAE的增长速率由大到小依次为基础模型>广义模型>混合效应模型。【结论】TLS提取的根径及胸径可作为建模数据,解析木材积与单木材积的RE均值在合格范围内(±10%),单木材积可作为建模数据;基础模型的拟合优度由大到小依次为Berkhart模型>Krenn模型>Meyer模型>Gehrardt模型;广义模型由于株数密度和树木年龄的加入显著提高了模型预测精度;混合效应模型评价指标R^(2)_(a)和RMSE的表现均优于广义模型。
【Objective】To explore the extraction accuracy of ground-based laser radar(TLS)for each relative high diameter of Larix olgensis,and the relationship between stump diameter and wood volume of Larix olgensis under different stand conditions in southern Heilongjiang Province,verify the effect of different modeling methods in the prediction of wood volume,and provide a scientific basis for predicting single wood volume when only stumpwere left.【Method】Eight samples of Larix olgensis plantation plots were used as the research objects,and stump diameter(dst)and diameter at breast height(DBH)extracted by TLS were substituted into One variable volume formula to obtain volume(V).The basic model of stump diameter and volume was constructed by nonlinear least square method,and the generalized model was constructed by adding trees per hectare(N)and tree AGE(AGE).Then,the sample plot was added as a random effect to construct a mixed effect model,and the prediction effects of different models were compared.Finally,the model was tested by the method of diameter order comparison.【Result】The extraction accuracy of TLS at stump diameter is 93.42%,which is 5.56%lower than that of DBH(96.37%).There is a high correlation between the analytical wood volume(V解)of 40 trees and the calculated wood volume(V)of One variable volume formula(R^(2)=0.9862),and the average relative error(RE)is 9.83%.Among the basic models,the Berkhart model has the smallest stump mean square error(RMSE=0.1949)and the best fitting effect(R2=0.8452,R^(2)_(a)=0.8448).In the correlation analysis,V is positively correlated with dst and AGE,and negatively correlated with N.Among them,the influence of AGE on V is slightly larger than that of N.The Ra2 of the generalized model is 7.98%higher than that of the basic model,and the RMSE is reduced from 0.1955 to 0.1949.In the mixed effects model,random effects performed best in b1 and b2 positions,R^(2)_(a) increased by about 1.66%compared with the generalized model,and RMSE decreased from 0.1955 to 0.1942.When 11 cm≤dst<43 cm,the Mean absolute error(MAE)of the Berkhart model,the generalized model and the mixed effects model increased with the increase of the stump diameter(dst)in the diameter order test.And the growth rate of MAE is from large to small:basic model>generalized model>mixed effect model.【Conclusion】Stump diameter(dst)and DBH extracted by TLS can be used as modeling data.The mean value of RE with V is within the qualified range(±10%),and V can be used as modeling data.The goodness of fit of the basic model from large to small is:Berkhart model>Krenn model>Meyer model>Gehrardt model.The addition of N and AGE in the generalized model significantly improved the prediction accuracy of the model.The mixed effects model performed better than the generalized model in terms of evaluation metrics Ra2 and RMSE.
作者
刘陈妤
贾炜玮
王帆
郭昊天
李丹丹
LIU Chenyu;JIA Weiwei;WANG Fan;GUO Haotian;LI Dandan(College of Forestry,Northeast Forestry University,Harbin 150040,Heilongjiang,China)
出处
《中南林业科技大学学报》
北大核心
2025年第9期82-93,共12页
Journal of Central South University of Forestry & Technology
基金
国家重点研发计划项目(2023YFD2200802)。
关键词
长白落叶松
根径-材积模型
地基激光雷达
混合效应模型
Larix olgensis
stump diameter-volume model
terrestrial laser scanning
mixed effects model
作者简介
第一作者:刘陈妤(liuchenyu816@163.com),硕士研究生;通信作者:贾炜玮(jiaww2002@163.com),教授,博士。
