极端干旱区苏干湖湿地植被分类与变化分析  

Classification and changes of vegetation in Sugan Lake wetland in the extreme arid region

作　　者：张腾 苗运法 邹亚国 张孜越 冯国平 Teng Zhang;Yunfa Miao;Yaguo Zou;Ziyue Zhang;Guoping Feng(Key Laboratory of Ecological Safety and Sustainable Development in Arid Lands,Northwest Institute of Eco-Environment and Resources,Chinese Academy of Sciences,Lanzhou 730000,China;University of Chinese Academy of Sciences,Beijing 100049,China;Management Station of Dasugan Lake Migratory Bird Provincial Nature Reserve in Akse Kazakh Autonomous County,Akse Kazakh Autonomous County 736400,Gansu,China)

机构地区：[1]中国科学院西北生态环境资源研究院干旱区生态安全与可持续发展重点实验室,甘肃兰州730000 [2]中国科学院大学,北京100049 [3]阿克塞哈萨克族自治县大苏干湖候鸟省级自然保护区管理站,甘肃阿克塞哈萨克族自治县736400

出　　处：《中国沙漠》2024年第4期81-90,共10页Journal of Desert Research

基　　金：国家重点研发计划项目(2020YFA0608400);中国科学院“西部之光”-西部交叉团队重点实验室专项(xbzg-zdsys-202204)。

摘　　要：湿地生物多样性极为丰富,湿地植被作为该生态系统中的核心部分,在维持系统稳定性和多功能性方面具有重要作用,掌握湿地植被类型与分布特征对生物多样性保护极其重要。然而,受湿地植被群落信息缺乏和遥感分辨率等因素的影响,对干旱区湿地植被分类的研究存在明显不足。选择中国西北极端干旱区柴达木盆地北部的苏干湖湿地作为研究对象,以116个野外植被调查点以及626个无人机影像植被样本点数据为基础,选择Sentinel-1合成孔径雷达数据(Synthetic Aperture Radar,SAR)和Sentinel-2光学遥感数据(MultiSpectral Instrument,MSI)构建新的遥感特征数据库,运用随机森林算法进行苏干湖湿地植被的分类与制图。结果表明:(1)SAR数据和MSI数据的结合使用能够提高湿地植被分类的精度,2019、2020、2021、2022、2023年湿地植被分类结果总体精度均超过0.81,Kappa系数分别为0.82、0.84、0.86、0.82、0.82。(2)2019—2023年苏干湖湿地面积总体呈稳定状态,植被分布面积为783.90km^(2)。芦苇(Phragmites australis)群落面积增加了28.49 km^(2),赖草(Leymus secalinus)群落面积增加了27.21 km^(2),而水麦冬(Triglochin palustre)、沼泽荸荠(Eleocharis palustris)群落面积减少了64.49 km^(2)。初步认为径流增加和禁牧政策是湿地植被分布变化的重要原因。Wetland is the most biodiverse ecosystem on Earth,and wetland vegetation plays a crucial role in maintaining the stability and functionality of these ecosystems,so mastering wetland vegetation types and distribution characteristics is extremely important for biodiversity conservation.Due to factors such as lacking or unsystematic vegetation community information and remote sensing resolution,the research on vegetation distribution in arid wetlands is limited.Taking the Sugan Lake wetland in the northern part of the Qaidam Basin of the extremely arid region in northwest China as the study area,based on the field vegetation survey data of 116 points and 626 unmanned aerial vehicle image sample points data,Sentinel-1 Synthetic Aperture Radar(Synthetic Aperture Radar,SAR)data and Sentinel-2 Multispectral Imager imagery(MultiSpectral Instrument,MSI)data were used to construct a new remote sensing feature database.The vegetation in Sugan Lake wetland was classified and mapped using the Random Forest algorithm.The results show that:(1)The combination of SAR and MSI data can improve the accuracy of wetland vegetation classification,with overall accuracy of wetland vegetation classification exceeding 0.81 for the years 2019-2023,and Kappa coefficients of 0.82,0.84,0.86,0.82,and 0.82 respectively.(2)From 2019 to 2023,the area of Sugan Lake wetland remained stable,with a vegetation distribution area of 783.90 km~2.The distribution area ofreed(Phragmites australis)communities increased by 28.49 km~2,and the area of leymus(Leymus secalinus)communities increased by 27.21 km~2.In contrast,the coverage of triglochin palustre(Triglochin palustre)and eleocharis palustris(Eleocharis palustris)communities decreased by 64.49 km~2.It is preliminarily considered that increased runoff and grazing prohibition policies are important reasons for the changes in wetland vegetation distribution.This study provides an effective method for surveying vegetation in arid area wetlands.High-quality dynamic monitoring of wetland vegetation offers th

关 键 词：遥感监测 干旱区 苏干湖湿地 随机森林 植被分类 

分 类 号：Q948[生物学—植物学]