基于随机森林的杉木标准树高曲线
2021-11-11赵文纯张再鑫刘检明赖永超
赵文纯 张再鑫 刘检明 赖永超



摘 要: 以湖北省赤壁市国有林场40块杉木人工林实测数据为例,运用随机森林方法,以胸径、优势树高、优势胸径为自变量,建立树高预测模型。首先根据随机森林的置换精度重要性筛选出建模的自变量,并确定决策树的数量和竞争节点变量数,得到决定系数R2为0.945 0,均方误差MSE为2.696 6的随机森林树高预测模型。利用检验数据对随机森林树高预测模型和传统树高预测模型分别进行精度检验。结果表明:随机森林模型的拟合效果与预测效果都优于该传统树高模型,随机森林模型可以作为有效的树高预测技术。
关键词: 杉木;标准树高曲线;随机森林
中图分类号:S757;S791.27 文献标识码:A 文章编号:1004-3020(2021)05-0020-04
Generalized Height-diameter Model for Cunninghamia lanceolata Based on Random Forest
Zhao Wenchun Zhang Zaixing Liu Jianming Lai Yongchao
(Hubei Provincial Forestry Investigation and Planning Institute Wuhan 430079)
Abstract:
Taking the measured data of 40 Cunninghamia lanceolata plantation plots in the national forest farm of Chibi City,Hubei Province as an example,a tree height prediction model was established by using the random forest method and taking the DBH,dominant tree height and dominant DBH as independent variables.First,the independent variable for modeling was selected,then,number of trees and number of predictors sampled for spliting at each node were determined,then,an optimum random forest model was developed,with a determinate coefficient of 0.945 0 and error of mean square of 2.696 6.And then,it was compared with one traditional generalized height-diameter equation,the validation datasets were used to test the models,respectively.The fitting effect and prediction effect of random forest are better than the traditional equation,and random forest model can be used as effective tree height prediction technology.
Key words: Cunninghamia lanceolata;generalized height-diameter model;random forest
树高和胸径是两个重要的林分调查因子,常用于预测林分生产力和树木材积。一般而言,林木胸径的测量简单、方便、结果准确,而在林分中树高的测量较为费时费力,比较复杂。因此在林分调查中,一般通过先测量部分林木的胸径和树高,建立树高胸径的回归方程,然后预测缺失的树高。
树高和胸径的关系会受到立地条件、林分条件、营林措施的影响[1-2],不同样地间的树高和胸径关系通常存在着较大的差异。而仅以胸径为自变量的简单模型不能描述不同林分状况下树高和胸径关系的差异。仅以胸径为自变量的普通树高曲线需为每个林分建立不同的模型,因此,其应用范围非常有限。而加入了林分因子的标准树高曲线可适用于更广的区域[3-5]。……
