基于TLS数据的落叶松–水曲柳混交林单木因子提取及树高模型构建研究  

Extraction of Individual Tree Factor and Tree Height Model Construction of Larix olgensis−Fraxinus mandshurica Mixed Forest Based on TLS Data

作　　者：崔译今 贾炜玮[1,2] 王帆 郭昊天 李丹丹 Cui Yijin;Jia Weiwei;Wang Fan;Guo Haotian;Li Dandan(College of Forestry,Northeast Forestry University,Harbin Heilongjiang 150040,China;Key Laboratory of Ministry of Education on Sustainable Forest Ecosystem Management,Harbin Heilongjiang 150040,China)

机构地区：[1]东北林业大学林学院,黑龙江哈尔滨150040 [2]森林生态系统可持续经营教育部重点实验室,黑龙江哈尔滨150040

出　　处：《西南林业大学学报(自然科学)》2025年第2期142-150,共9页Journal of Southwest Forestry University:Natural Sciences

基　　金：国家重点研发计划项目(2023YFD2200802)资助;中央高校基本科研业务费专项资金项目(2572019CP08)资助。

摘　　要：以孟家岗林场1 hm^(2)落叶松水与曲柳混交林样地为研究对象,利用等株径级标准木法把林木分为优势木、平均木、被压木3个等级,然后以人工实测值作为参考值,分别分析利用TLS提取2种树种的3种等级木单木因子的精度,最后采用TLS数据提取的单木因子构建树高模型。筛选出2种树种最优基础树高模型,并进一步评价和比较以林木分级为哑变量构建的树高模型。结果表明:针对本研究选取的水落混交林样地,点云数据与实测数据单木匹配结果中,落叶松匹配精度为92.79%,水曲柳为92.25%;2个树种的胸径提取精度达到97%以上,且胸径提取精度优势木>平均木>被压木,2个树种的树高提取精度达到95%以上,落叶松树高提取精度平均木>优势木>被压木;水曲柳树高提取精度优势木>平均木>被压木。使用TLS数据构建的基础树高模型中,拟合落叶松效果最好的是Logistic模型(R^(2)=0.783 0、RMSE=1.951 6),拟合水曲柳效果最好的是Gompertz模型(R^(2)=0.724 8、RMSE=1.953 6),因此以Logistic模型、Gompertz模型分别为2个树种基于TLS数据构建的最优基础模型,最后2个树种采用以林木分级为哑变量构建的模型R^(2)分别为0.790 7、0.731 2。TLS技术对水落混交林样地单木匹配率很高,单木因子提取精度较好,基于TLS数据所构建的以林木分级为哑变量的模型,在预测树木高度和胸径的生长差异方面表现优于基础模型,具有更好的预测精度和适应性,可以为该地区水落混交林的林业经营提供参考。Taking 1 hm2 mixed stand of Larix olgensis−Fraxinus mandshurica in Mengjiagang Forest Farm as the research object, the forest trees were divided into 3 grades: dominant trees, average trees, and pressed trees, using the equal plant diameter step standard wood method. Then, with the measured values as the reference value, the accuracy of extracting the individual tree factors of 3 grades of 2 species by TLS was analyzed respectively, and finally the tree height model was constructed using the individual tree factors extracted from TLS data. The optimal basic tree height models of the 2 species were screened out, and the tree height models built with the dummy variable of tree classification were further evaluated and compared. The results showed that for the mixed stand of L. olgensis−F. mandshurica selected in this paper, the matching accuracy of the individual tree between the point cloud data and the measured data was 92.79% for L. olgensis and 92.25% for F. mandshurica. The ex-traction accuracy of the diameter at breast height of the 2 species reached more than 97%, and the extraction ac-curacy of the diameter at breast height was dominant tree > average tree > pressed tree. The extraction accuracy of the tree height of the 2 species reached more than 95%, the extraction accuracy of L. olgensis was average tree > dominant tree > pressed tree, and the extraction accuracy of F. mandshurica was dominant tree > average tree > pressed tree. In the basic tree height models built using TLS data, the Logistic model(R^(2)=0.783 0, RMSE=1.951 6) fitted L. olgensis best, and the Gompertz model(R^(2)=0.724 8, RMSE=1.953 6) fitted F. mandshurica best. There-fore, the Logistic model and Gompertz model were the optimal basic models built based on TLS data for the 2 tree species respectively. Finally, the models built with tree classification as dummy variable for the 2 tree species had R^(2) values of 0.790 7 and 0.731 2, respectively. TLS technology has a high individual tree matching rate for the mixed stand and a good

关 键 词：落叶松 水曲柳 混交林 地基激光雷达 树高 哑变量模型 

分 类 号：S757[农业科学—森林经理学]