文献类型：期刊文献

中文题名：基于分位数回归的杉木人工林地位级划分方法研究

英文题名：Site Classes Grouping Method Based on Quantile Regression of Chinese Fir Plantations

作者：张博[1] 陈科屹[2] 周来[1] Sajjad Saeed[1] 张雅馨[3] 孙玉军[1]

第一作者：张博

机构：[1]北京林业大学森林资源和环境管理国家林业和草原局重点实验室,北京100083;[2]中国林业科学研究院林业科技信息研究所,北京100091;[3]中国林业科学研究院森林生态环境与保护研究所,北京100091

年份：2021

卷号：34

期号：4

起止页码：103-110

中文期刊名：林业科学研究

外文期刊名：Forest Research

收录：CSTPCD;;Scopus;北大核心:【北大核心2020】；CSCD:【CSCD2021_2022】；

基金：国家自然科学基金“基于树木生长过程的长白落叶松树冠模型”(No.31870620);林业科学技术推广项目“基于分水岭算法的森林植被碳储量监测技术成果推广应用”([2019]06)。

语种：中文

中文关键词：分位数回归模型;杉木人工林;立地质量;地位级

外文关键词：quantile regression model;Cunninghamia lanceolata plantation;site quality;site class

分类号：S758.57

摘要：[目的]优化地位级划分策略的效率并提高立地质量分级的准确性,提出基于分位数回归模型的地位级模型和立地质量评价方法。[方法]分别采用标准差调整法和分位数回归方法分级并评价了福建省三明市将乐国有林场418个杉木纯林样地的立地质量,并将结果进行比较。具体为:以林分树高生长量趋于稳定和杉木达到成熟龄为标准确定基准年龄(A0)。采用标准差调整法并按标准年龄时树高值和地位级距构建地位级曲线簇,并划分了8个地位级。采用分位数回归法以导向曲线为基础,根据数据分布特点,指定8个分位点(0.01,0.05,0.15,0.30,0.70,0.85,0.95,0.99)构建分位数回归模型,并以分位数曲线划分地位级。[结果]根据林分平均高与各分级曲线树高预测值的差值平方和(或差值绝对值)最小的原则,分位数回归模型在林地地位级划分方面虽与传统方法无显著差异,却显著提高了划分效率,并准确地实现了对杉木人工纯林地位级的评价。[结论]分位数回归模型可描述、分类、预测和验证林分生长与地位级之间的关联性,基于导向生长模型的分位数回归曲线簇能够更直观地反映出不同地位级下杉木树高的变化规律,从而全面准确地预测杉木人工林的生产力水平。
[Objective]To optimize the efficiency of the site classes grouping strategy and improve the accuracy of site classification,site classes grouping model and to propose a site quality evaluation method based on quantile regression model.[Method]The traditional methods(standard deviation adjustment method)and quantile regression method were used to classify and evaluate the site quality of 418 pure Chinese fir(Cunninghamia lanceolata)forests at Jiangle Forest Farm in Sanming City,Fujian Province,and the results were compared.The baseline age(A0)was determined based on the high growth of the stand tree and the maturity of the Chinese fir plantations.Using standard deviation adjustment method and according to the standard age tree height value and exponential interval,the site classes curve cluster was constructed and divided into 8 levels.The quantile regression method was based on the guiding curve.According to the data distribution characteristics,8 quantile points(0.01,0.05,0.15,0.30,0.70,0.85,0.95,and 0.99)were specified to construct the quantile regression model,and the quantile curves were used to divide the site classes.[Result]The results showed that the quantile regression model could quickly and accurately determine the site type of the pure Chinese fir plantation,based on the principle that the sum of squares(or the absolute value of the difference)between the average stand height and the prediction stand height of each site class curves is the smallest.The evaluation results of the site quality were not significantly different from the traditional methods.[Conclusion]The quantile regression model describes,classifies,regresses,predicts and verifies the correlation between stand growth and site quality from the perspective of data.The quantile regression curve cluster,based on the guided growth model,intuitively reflect the stand high changing under the different site class,so as to comprehensively and accurately predict the productivity of Chinese fir plantations.