基于随机森林和MaxEnt模型的长沙市银杏古树生境适宜性评价  

Habitat suitability evaluation of ancient ginkgo trees in Changsha based on random forest and MaxEnt model

作　　者：邱汉周[1] 陈存友 QIU Hanzhou;CHEN Cunyou(School of Tourism Management,Ningde Normal University,Ningde 352100,Fujian,China;College of Landscape Architecture,Central South University of Forestry&Technology,Changsha 410004,Hunan,China)

机构地区：[1]宁德师范学院旅游管理学院,福建宁德352100 [2]中南林业科技大学风景园林学院,湖南长沙410004

出　　处：《中南林业科技大学学报》2024年第11期87-97,共11页Journal of Central South University of Forestry & Technology

基　　金：湖南省自然科学基金项目(2019JJ50990);湖南省教育厅科学研究重点项目(23A0203)。

摘　　要：【目的】城市中的银杏古树是重要的自然资源和文化遗产,具有重要的生态、历史、景观和经济价值。然而,快速城市化及人类活动导致城市生态系统功能的急剧变化,已经不可避免地造成了城市中银杏古树的生长和保护问题,研究以期实现对银杏古树的胸径、树高以及生境适宜度的评估,为长沙市银杏古树的保护提供参考。【方法】以中国湖南省长沙市范围内树龄100年以上的160株银杏古树为研究对象,以省市级的森林资源清查数据为基础结合实地勘测数据,采集160株银杏古树的树龄、海拔、坡向、坡度、年均降水量、土壤类型和平均冠幅等为建模参数并按照训练∶验证=4∶1的比例对数据分类,采用Pearson相关性筛选及重要性排序对参数处理后,利用多元线性回归、支持向量机回归及随机森林回归方法建立了银杏古树的胸径、树高生长模型。在此基础上,利用最大熵权模型评估了银杏古树的生境适宜性,并进行长沙市银杏种植适宜区制图。【结果】研究表明:银杏的胸径、树高随机森林回归模型拟合效果皆为最优,其中胸径随机森林模型的决定系数R^(2)最高为0.86,均方根误差RMSE为2.67,高于支持向量机回归(R^(2)为0.72)以及多元线性回归(R^(2)为0.79);树高随机森林模型的R^(2)最高为0.82,RMSE为13.09,高于支持向量机回归方法(R^(2)为0.59)以及多元线性回归方法(R^(2)为0.78)。【结论】胸径的生长主要受海拔、坡度及年均降水量的影响,而树高受树龄、海拔及年均降水量的影响较大。该研究结果可为未来长沙市种植银杏、补充银杏名木数量、保持银杏名木可持续性发展提供科学依据。【Objective】Ancient ginkgo trees in cities are important natural resources and cultural heritage with significant ecological,historical,landscape and economic values.However,rapid urbanization and human activities leading to drastic changes in urban ecosystem functions have inevitably caused problems in the growth and conservation of ginkgo trees in cities.The study was conducted with a view to realizing the assessment of the diameter at breast height(DBH),height of trees and habitat suitability of ginkgo trees in Changsha city,and to provide a reference for the conservation of ginkgo trees in Changsha city.【Method】160 ginkgo trees over 100 years old in Changsha city,Hunan province,China,were used as the research objects.Based on the provincial and municipal forest inventory data combined with the field survey data,the modeling parameters of 160 ginkgo trees were collected,including age,elevation,slope direction,slope,average annual precipitation,soil type,and average crown width,etc.,and the data were categorized according to the ratio of training:validation=4:1.After the parameters were processed by Pearson correlation screening and importance ranking,the growth model of chest diameter and tree height of Ginkgo biloba was established by using multiple linear regression,support vector machine regression and random forest regression methods.On this basis,the habitat suitability of ginkgo trees was evaluated using the maximum entropy weight model (MaxEnt), and mapping of suitable areas for ginkgo planting in Changsha city was carried out. 【Result】The study showed that the random forest regression model of ginkgo diameter at breast height and tree height had the best fitting effect, in which the coefficient of determination of the random forest model of diameter at breast height was the highest R^(2) of 0.86, and the root mean squared error (RMSE) was 2.67, which was higher than that of the support vector machine regression (R^(2) of 0.72) and the multivariate linear regression (R^(2) of 0.79), and that

关 键 词：银杏古树 生境适宜度 随机森林 MAXENT 长沙市 

分 类 号：S791.11[农业科学—林木遗传育种]