摘要
滇西南地区拥有丰富的丛生竹林景观和珍稀特有竹种资源,但竹资源分布储量不清、监测技术缺乏等问题很大程度限制了竹资源开发与利用。基于Sentinel-2A影像数据,采用反向传播神经网络、支持向量机、随机森林三种机器学习分类方法进行沧源县丛生竹林信息提取及精度评价,利用Google Earth影像和DEM数据对竹资源分布的空间和地形特征进行了分析。结果表明,随机森林分类精度优于支持向量机和反向传播神经网络,分类总体精度达90%,Kappa系数达0.87,竹林用户精度达81%。沧源县共有竹林138.07 km2,主要分布于城镇村庄、道路、水系和耕地周边,以四旁竹和防护竹林为主,采用Sentinel-2A10 m的分辨率很好地提取了空间上分布分散的丛生竹林。沧源县竹林主要分布在海拔900~2000 m,坡度范围大都位于缓坡和斜坡。研究结果可为沧源县竹资源开发利用提供数据支持,研究方法可作为大型丛生竹遥感监测的参考。
Southwest Yunnan is enriched in landscapes of cluster bamboo forests and rare but unique resources of bamboo species.The development and utilization of bamboo resources are greatly limited by the uninformed distribution and growing stock of bamboo forests and the lack of proper monitoring techniques.Based on Sentinel-2 A image data,we compared three machine learning methods:back-propagation neural network,support vector machine,and random forest,for classifying cluster bamboo forests and other land use types in Cangyuan County.Google Earth images and DEM data were used to analyze the spatial and topographic characteristics of the distribution of bamboo forests.The results showed that random forest achieved the best accuracy in classification,with overall accuracy of 90%,Kappa coefficient of 0.87,and user accuracy of81%.With a total bamboo forest of 138.07 km2,bamboo forests in Cangyuan are mainly located in towns and villages,along roads,rivers and cultivated lands.At a resolution of 10 m,Sentinel-2 A data is good at characterizing spatially dispersed cluster bamboo forests.Cangyuan’s bamboo forests are mainly located at gentle slopes or incline at altitudes from 900 to 2000 m.Our results provided basis for the development and utilization of bamboo resources in Cangyuan County.The methods used here provided reference for monitoring cluster bamboo forests.
作者
严欣荣
张美曼
郑亚雄
尹子旭
黄兰鹰
姜小雨
官凤英
YAN Xin-rong;ZHANG Mei-man;ZHENG Ya-xiong;YIN Zi-xu;HUANG Lan-ying;JIANG Xiao-yu;GUAN Feng-ying(Key Laboratory of Bamboo and Rattan Science and Technology,International Center of Bamboo and Rattan,Beijing 100020,China)
出处
《生态学杂志》
CAS
CSCD
北大核心
2020年第3期1056-1066,共11页
Chinese Journal of Ecology
基金
“十三五”国家重点研发计划专项(2018YFD0600103)
江苏宜兴竹林生态系统国家定位观测研究站运行补助(2019132146)资助。
关键词
丛生竹
反向传播神经网络
支持向量机
随机森林
分类方法
分布特征
沧源
cluster bamboo
back propagation neural network
support vector machine
random forest
classification method
distributing characteristic
Cangyuan
