GEE支持下联合植被物候特征与机器学习的入侵植物互花米草提取  

Extraction of invasive plant Spartina alterniflora by combining vegetation phenological characteristics and machine learning supported by GEE

作　　者：刘明月 郑浩 陈星彤 杨晓芜 宋敬茹 张永彬 满卫东 LIU Mingyue;ZHENG Hao;CHEN Xingtong;YANG Xiaowu;SONG Jingru;ZHANG Yongbin;MAN Weidong(College of Mining Engineering,North China University of Science and Technology,Tangshan 063210,China;Tangshan Key Laboratory of Resources and Environmental Remote Sensing,Tangshan 063210,China;Hebei Industrial Technology Institute of Mine Ecological Remediation,Tangshan 063210,China;Collaborative Innovation Center of Green Development and Ecological Restoration of Mineral Resources,Tangshan 063210,China)

机构地区：[1]华北理工大学矿业工程学院,河北唐山063210 [2]唐山市资源与环境遥感重点实验室,河北唐山063210 [3]河北省矿区生态修复产业技术研究院,河北唐山063210 [4]矿产资源绿色开发与生态修复协同创新中心,河北唐山063210

出　　处：《测绘通报》2023年第6期36-43,共8页Bulletin of Surveying and Mapping

基　　金：国家自然科学基金青年基金(41901375,42101393);河北省自然科学基金青年基金(D2019209322,D2022209005);河北省高等学校科学技术研究项目青年拔尖人才项目(BJ2020058);河北省引进留学人员资助项目(C20200103);唐山市科技计划重点研发项目(22150221J)。

摘　　要：外来物种入侵威胁生物多样性,破坏生态系统的结构与功能。互花米草作为我国首批16种外来入侵物种名单中唯一的海岸盐沼植物,对其实现快速精准的识别对滨海湿地可持续发展与管理具有重要意义。本文基于GEE密集时序Sentinel-2影像,利用HANTS算法拟合NDVI时序曲线,采用J-M距离优选互花米草识别关键物候期,综合关键物候期Sentinel-1雷达、Sentinel-2光学影像与DEM数据,构建光谱、雷达、地形、纹理特征集,分别采用支持向量机(SVM)、分类回归树(CART)、随机森林(RF)和梯度提升树(GTB)4种方法实现盐城滨海湿地互花米草信息提取。研究表明:①基于HANTS算法与J-M距离优选的互花米草识别关键物候期为成熟期和衰老初期(10—11月),关键物候期内互花米草与土著植物的可分离性显著提高;②利用关键物候期内多源特征集训练SVM、CART、RF与GTB分类器,互花米草F1值分别为0.95、0.93、0.97、0.95,RF分类效果最优;③盐城湿地珍禽国家级自然保护区核心区现有互花米草面积为3741.86 hm^(2),占比16.56%,互花米草斑块表现为点源和边界源扩散,侵占土著植物群落生态位,对盐城湿地珍禽国家级自然保护区生态平衡造成巨大威胁。本文依托于GEE云平台,联合植被物候特征和机器学习算法,综合利用多源遥感数据,能够准确、快速、高效地提取滨海湿地互花米草信息。The invasion of alien species threatens biodiversity and destroys ecosystem structure and function.As the only coastal salt marsh plant in the list of the first 16 invasive alien species in China,the rapid and accurate identification of Spartina alterniflora(S.alterniflora)is of great significance for the sustainable development and management of coastal wetlands.Based on time-series Sentinel-2 images on the GEE,HANTS algorithm is used to fit the NDVI time-series curves,then J-M distance is used to preferentially select the key phenological periods of the S.alterniflora.Sentinel-1,Sentinel-2,and DEM datasets during the key phenological periods are integrated to construct spectral,radar,topographic,and texture features.On this basis,four machine learning methods:SVM,CART,RF and GTB are applied to extract S.alterniflora in Yancheng coastal wetlands.The results show that:①the key phenology extracted by HANTS algorithm and J-M distance optimization for S.alterniflora identification are maturity and early senescence(October—November),and the discrimination of S.alterniflora from native plants in the key phenology is significantly improved;②4 classifiers are trained based on multi-source features of the key phenology,the F1-score of S.alterniflora classification resulted from SVM,CART,RF and GTB classifiers are 0.95,0.93,0.97 and 0.95,respectively,and RF had the best classification result;③The existing area of S.alterniflora in the core area of Yancheng Wetland Rare Bird National Nature Reserve(YNNR)is 3741.86 hm^(2),accounting for 16.56%.The patches of S.alterniflora show point-source and community boundary-source diffusion,which occupy the ecological niche of native plant communities and threaten the ecological balance of YNNR.Based on the GEE cloud platform,combined with vegetation phenology characteristics and machine learning algorithms,this study utilized multi-source remote sensing data to extract S.alterniflora information in coastal wetlands accurately,rapidly,and efficiently.

关 键 词：GEE 互花米草 植被物候特征 J-M距离 机器学习 

分 类 号：P237[天文地球—摄影测量与遥感]