摘要
洪涝灾害一直是自然灾害研究的热点问题,具有突发性、频发性、影响范围广泛等特点。现有的洪水灾害遥感研究存在监测时相不足、时效性低等问题,难以对其过程进行有效监测。针对这一问题,基于光学与雷达影像,首先评估随机森林、支持向量机、最小距离算法对鄱阳湖洪涝灾害淹没范围提取的效果;然后采用提取精度最高的随机森林算法对鄱阳湖区2020年6月至7月的洪涝灾害过程进行监测;最后结合洪涝灾害发生前的土地利用类型图,分析洪涝灾害淹没的地物类型及面积。结果表明:1)随机森林算法对洪涝淹没范围提取效果最好,总体分类精度为95.50%,Kappa系数达到0.91;2)集成光学与雷达遥感数据能够有效地对洪峰过程进行监测,相比于传统光学遥感监测,雷达遥感传感器不受云雨天气影响,应受到更加多的重视;3)2020年6月26日至7月14日,鄱阳湖区域新增淹没面积达1367.18 km^(2),主体淹没范围呈持续扩大趋势,此后洪水逐渐退水消亡;整个过程淹没面积最大的地物类型为农田,淹没面积达到246.68 km^(2)。
Flood disasters have always been a hot topic in natural disaster research,which are characterized by suddenness,frequency,and wide-ranging impacts.The current remote sensing research on flood disasters faces challenges such as insufficient monitoring time and low timeliness,making it difficult to effectively monitor their processes.To address this issue,based on optical and radar images,the study initially evaluates the effectiveness of random forests,support vector machines,and minimum distance algorithms in extracting the inundation range of the Poyang lake flood disaster;Then,the highest accuracy random forest algorithm is used to monitor the flood disaster process in Poyang lake area from June to July 2020;Finally,based on the land use type map before the occurrence of flood disasters,the types and areas of land features submerged by flood disasters are analyzed.The results show that:1)The random forest algorithm performs best in extracting flood inundation range,with an overall classification accuracy of 95.50%and a Kappa coefficient of 0.91;2)Integrating optical and radar remote sensing data can effectively monitor flood peak processes.Compared to traditional optical remote sensing monitoring,radar remote sensing sensors are not affected by cloud and rain weather and should be given more attention;3)From June 26 to July 14,2020,the newly submerged area in the Poyang lake area reached 1367.18 km^(2),with the main submerged area continually expanding,after which the flood gradually receded.The type of land with the largest inundation area throughout the entire process is farmland,with an inundated area of 246.68 km^(2).
作者
田申
艾金泉
李黄晶
钟超
刘慧玲
王诗铭
杨玉成
TIAN Shen;AI Jinquan;LI Huangjing;ZHONG Chao;LIU Huiling;WANG Shiming;YANG Yuchen(Key Laboratory of Mine Environmental Monitoring and Improving around Poyang Lake of Ministry of Natural Resources,East China University of Technology,330013,Nanchang,PRC;Key Laboratory of Watershed Ecological Process and Information of Jiangxi Province,East China University of Technology,330013,Nanchang,PRC)
出处
《江西科学》
2024年第5期982-987,1095,共7页
Jiangxi Science
基金
国家自然科学基金项目(42361059)
江西省大学生创新创业训练计划项目(202310405017,S202310405024)
江西省高校人文社会科学研究项目(JC23217)
东华理工大学2023年度“创新创业教育实践基地”建设项目(2023ZX03)。
关键词
洪涝灾害
随机森林
遥感
鄱阳湖
flood disasters
random forest
remote sensing
Poyang lake
