摘要
针对黑河流域大尺度环境下水体提取难度大、演变规律尚不明晰等问题,基于谷歌地球引擎(Google Earth Engine,GEE)处理黑河流域1986—2024年Landsat影像,采集7.8×10^(4)个水体/非水体样本并构建逐年样本数据集,通过将多波段水体指数(Multi band water index,MBWI)、增强型水体指数(Enhanced water index,EWI)、改进归一化差异水体指数(Modified normalized difference water index,MNDWI)与光谱波段进行单独与统一组合,构建并筛选出最佳融合水体指数的随机森林(Random forest,RF)水体提取方法,提取了研究区39个时相的逐年地表水体影像,采用曼-肯德尔(Mann-Kendall,M-K)法揭示了黑河流域逐年地表水体面积变化特征,基于主成分与敏感性分析探究了影响地表水体演变的主要驱动因素。结果表明:融合3种水体指数(MBWI、EWI、MNDWI)的随机森林水体提取方法对黑河流域Landsat影像的水体提取效果最佳,平均总体精度(Overall accuracy,OA)为96.16%,平均Kappa系数为0.9128;经M-K法检验,黑河流域1986—2024年地表水体面积呈波动减少态势;年降水量、人口、年蒸散量为黑河流域地表水体演变的最主要驱动因素。研究结果可为全流域地表水体的快速准确提取提供理论支持。
Aiming to address the challenges of extracting water bodies in large-scale environments and clarifying their long-term evolution patterns,the Landsat images of Heihe River Basin(1986—2024)were processed by using Google Earth Engine(GEE),approximately 78000 water and non-water samples were collected,and an annual sample dataset was built.Based on this dataset,a random forest(RF)-based water extraction method was developed by integrating the multi band water index(MBWI),enhanced water index(EWI),and modified normalized difference water index(MNDWI)with spectral bands,both individually and in combination.Through systematic screening,the optimal fusion index was selected,enabling the accurate extraction of surface water bodies across 39 temporal phases.The Mann-Kendall(M-K)test was applied to detect interannual trends in surface water area,while principal component analysis(PCA)and sensitivity analysis were used to identify the dominant driving factors influencing water body evolution.The results demonstrated that the RF method incorporating all three water indices(MBWI,EWI,and MNDWI)achieved the best extraction performance for Landsat images of the Heihe River Basin,with average overall accuracy(OA)of 96.16%and average Kappa coefficient(KC)of 0.9128.The M-K test indicated a fluctuating downward trend in surface water area from 1986 to 2024.Annual precipitation,population,and annual evapotranspiration were identified as the main driving factors for the evolution of surface water bodies in the Heihe River Basin.The research result can provide a theoretical foundation for the rapid and accurate extraction of surface water bodies at the basin scale and support future hydrological and environmental applications.
作者
赵文举
谢振东
徐文
俞海英
战国隆
王之君
ZHAO Wenju;XIE Zhendong;XU Wen;YU Haiying;ZHAN Guolong;WANG Zhijun(College of Energy and Power Engineering,Lanzhou University of Technology,Lanzhou 730050,China;Key Laboratory of Smart Agriculture Irrigation Equipment,Ministry of Agriculture and Rural Affairs,Lanzhou 730050,China;Hydrology and Water Resources Center of Gansu Province,Lanzhou 730030,China;Dayu Water-saving(Tianjin)Co.,Ltd.,Tianjin 301712,China)
出处
《农业机械学报》
北大核心
2025年第8期152-162,共11页
Transactions of the Chinese Society for Agricultural Machinery
基金
国家自然科学基金面上项目(52379042)
甘肃省重点研发计划项目(23YFFA0019、24YFFA033)
甘肃省青年基金项目(25JRRG013)。
关键词
水体识别
Landsat卫星
水体指数
时空演变
黑河流域
water body identification
Landsat satellite
water index
spatio-temporal evolution
Heihe River Basin
