文献类型：期刊文献

中文题名：基于LandTrendr算法和森林资源清查数据的海南岛森林扰动监测与分析

英文题名：Forest disturbance monitoring and analysis based on LandTrendr algorithm and forest resource inventory data in Hainan Island

作者：王晨[1,2] 庞勇[1,2] 袁智文[1,2] 蒙诗栎[1,2] 余涛[1,2] 孙乡楠[1,2,3]

第一作者：王晨

机构：[1]中国林业科学研究院资源信息研究所,北京100091;[2]中国林业科学研究院资源信息研究所,国家林业和草原局林业遥感与信息技术重点实验室,北京100091;[3]国家林业和草原局林草调查规划院,北京100714

年份：2025

卷号：45

期号：16

起止页码：8054-8070

中文期刊名：生态学报

外文期刊名：Acta Ecologica Sinica

收录：;北大核心:【北大核心2023】；

基金：国家重点研发计划(2023YFE0105100)。

语种：中文

中文关键词：海南岛;LandTrendr;森林扰动;遥感;森林资源清查

外文关键词：Hainan Island;LandTrendr;forest disturbance;remote sensing;National forest inventory

分类号：S771.8

摘要：海南岛作为中国重要的热带林区,对其森林生态系统的持续监测对于生态保护和可持续发展至关重要,而国家森林资源清查(National forest inventory,NFI)为揭示森林动态变化提供了权威的数据支撑。为了厘清海南岛近30年来的森林变化情况,利用1990—2023年海南岛生长季Landsat 5/7/8时间序列影像在谷歌地球引擎(Google Earth Engine,GEE)云平台上开展长时间序列森林变化分析研究,采用LandTrendr算法对海南岛森林扰动进行监测并结合地面调查数据和NFI数据进行精度验证,评估5个不同光谱指数对不同扰动类型的敏感度。结果表明:(1)基于地面调查精度评估发现归一化燃烧指数(Normalized Burn Ratio,NBR)监测扰动年份与实际扰动年份的相关性最高,R^(2)为0.91。(2)基于NFI数据精度评估结果表明NBR指数监测的总体精度最高(77.38%),缨帽变换的绿度分量(Tasseled Cap Greenness,TCG)监测的总体精度最低(61.31%)。(3)灾害因素中,增强型植被指数(Enhanced Vegetation Index,EVI)对突发性扰动如火灾、风倒的敏感度较高;TCG指数对渐进性扰动如病虫害、其他灾害的敏感度较高。人为因素中,NBR指数对土地利用变化的敏感度较高;TCG指数对森林更新活动的敏感度较高;EVI指数对采伐的敏感度最高。总体而言LandTrendr算法能有效监测海南岛的森林扰动;不同植被指数扰动监测对不同扰动类型的敏感度存在一定差异,通过集成不同植被指数的扰动结果有望提供分类型森林动态监测的精度。
As a significant tropical forest region in China,the continuous monitoring of Hainan Island’s forest ecosystem is crucial for ecological conservation and sustainable development.The National Forest Inventory(NFI)provides authoritative data for understanding the dynamic changes in forest ecosystems.In order to clarify the forest changes in Hainan Island in the past three decades.This study used the Landsat 5/7/8 time series images during the growing season of Hainan Island from 1990 to 2023,and conducted a long?term forest change analysis on the Google Earth Engine(GEE)cloud platform,using the LandTrendr algorithm to explore the disturbance monitoring technology suitable for forests in Hainan Island and evaluate the sensitivity of five different spectral indices to different disturbance types.The results showed that:(1)The accuracy evaluation based on the ground survey shows that the correlation between the disturbance year monitored by Normalized Burn Ratio(NBR)and the actual disturbance year is the highest,with an R^(2)of 0.91.(2)When combined with NFI data,NBR also exhibited the highest overall accuracy(77.38%),while Tasseled Cap Greenness(TCG)had the lowest accuracy(61.31%).(3)Based on NFI data,the sensitivity of different spectral indices in LandTrendr algorithm to different disturbance types was further evaluated.The results showed that the Enhanced Vegetation Index(EVI)was more sensitive to high intensity disturbance such as fire,wind throw.TCG was more sensitive to low intensity disturbance such as plant diseases and insect pests and other disturbances.Among the human factors,NBR showed a higher sensitivity to land use changes,TCG was more sensitive to forest regeneration activities,and EVI had the highest sensitivity to logging.Overall,the LandTrendr algorithm can effectively monitor forest disturbances in Hainan Island.There are certain differences in the sensitivity of disturbance monitoring using different vegetation indices for various disturbance types.By integrating the disturbance results from different vegetation indices,it is expected to improve the accuracy of classified forest dynamic monitoring.