基于近期密集卫星数据的北江流域森林砍伐遥感方法  

Remote Sensing Method of Forest Logging in the Beijiang River Basin Based on Recent Intensive Satellite Data

作　　者：赵梓彤 陈水森 于国荣[1] 李丹 贾凯 赵晨尧 李健[1,2] 秦伯雄 Zhao Zitong;Chen Shuisen;Yu Guorong;Li Dan;Jia Kai;Zhao Chenyao;Li Jian;Qin Boxiong(Faculty of Electric Power Engineering,Kunming University of Science and Technology,Kunming 650500,China;Guangzhou Institute of Geography,Guangdong Academy of Sciences,Key Lab of Guangdong for Utilization of Remote Sensing and Geographical Information System,Guangdong Open Laboratory of Geospatial Information Technology and Application,Research Center of Guangdong Province for Engineering Technology Application of Remote Sensing Big Data,Guangzhou 510070,China;Joint Laboratory on Low-carbon Digital Monitoring,Guangdong Institute of Carbon Neutrality(Shaoguan)&Shaoguan ShenBay Low Carbon Digital Technology Co.,Ltd.,Shaoguan 512029,China)

机构地区：[1]昆明理工大学电力工程学院,昆明650500 [2]广东省科学院广州地理研究所,广东省遥感与地理信息系统应用实验室,广东省地理空间信息技术与应用公共实验室,广东省遥感大数据应用工程技术研究中心,广州510070 [3]广东碳中和研究院,韶关市深湾低碳数字科技有限公司,低碳数字监测联合实验室,广东韶关512029

出　　处：《热带地理》2024年第11期2091-2103,共13页Tropical Geography

基　　金：韶关市人才工程“南岭团队计划”项目“双碳空间大数据”;广东省科技计划项目(2023A1414020039);广东省科学院发展专项资金项目资助(2022GDASZH-2022010202-003)。

摘　　要：针对森林砍伐前后地物光谱特征变化特点,选用北江流域Sentinel-22017—2022年共930景光学遥感影像作为实验数据,基于Google Earth Engine云计算平台,优化归一化植被指数阈值分割方法提取多时相森林砍伐分布动态特征。结果表明:采用阈值分割优化方法识别森林砍伐验证精度达到72.05%,2017—2022年,北江流域森林砍伐量除2020和2021年外总体呈现逐年递增趋势,年均增长约9%;森林砍伐主要集中在坡度8°~25°的缓坡和较陡坡,共占所有森林砍伐面积的48%~57%;且在15°以下坡度中更易发生,砍伐比例为3.76%,比15°以上坡度高出1.14%;2018年森林砍伐56个特征点NDVI标准化均值在砍伐后逐年增加,2018—2022年平均每年增加0.08,砍伐后第三年基本恢复林地NDVI特征。Forests are important natural and strategic resources,and deforestation is a significant cause of soil erosion.Given the high uncertainty and limited temporal-spatial resolution of land features classified by remote sensing,especially the lack of regional studies on the dynamic distribution of forest deforestation,it is urgent to extract the multi-temporal dynamic distribution of forest deforestation using remote sensing techniques.Based on the spectral features of ground objects before and after deforestation,930 optical remote sensing images from Sentinel-2 in the Beijiang River Basin from 2017 to 2022 were selected as experimental data.The Google Earth Engine cloud platform was utilized for data collection and preprocessing to calculate the NDVI vegetation index from 2017 to 2022.Following the extraction of forest distribution using the threshold segmentation method,the dynamic change in deforestation distribution between 2017 and 2022 in the study area was analyzed.The results showed that:(1)229 sampling points were randomly selected in the deforestation area,and the accuracy of remote sensing mapping of deforestation in 2020-2021 was evaluated using historical high-resolution images,achieving a verification accuracy of 72.05%.(2)From 2017 to 2022,deforestation in the Beijiang River Basin exhibited an increasing trend year by year,except in 2020 and 2021,with an average annual increase of about 9%.In terms of distribution,the largest proportion of deforestation occurred in the Wujiang River Basin during 2017-2022,with an average annual deforestation rate of 3.27%of the total area of the basin.The lowest proportion of deforestation was observed in the Nanshui River Basin,with an average annual deforestation rate of 1.47%of the total area of that basin.(3)In the Beijiang River Basin,the distribution of deforestation across different slopes is more uniform.The deforestation area is primarily concentrated on slopes between 8°and 25°,which account for 53.3%of the total basin area and generate 48%to 57%of the def

关 键 词：Sentinel-2 森林砍伐 动态监测 Google Earth Engine 北江流域 

分 类 号：S771.8[农业科学—森林工程]