本文选取中国 32个省会及以上城市建成区和香港特别行政区为研究区,基于Google Earth Engine(GEE)遥感大数据云平台技术,通过长时序Landsat遥感影像,结合传统水体指数和城市阴影指数构建了一个基于地表温度和混合指数的水体自动提取方法,并在24894幅Landsat遥感影像上自动生产了中国主要城市1990-2020年30m空间分辨率的地表水体数据.结果表明:引入地表温度可以大幅提高城市地表水体提取的精度,城市地表水体的提取总体精度均在93%以上;研究时段内,地表水体面积整体呈增加趋势,从1990年的2758.3 km2 增长到2020年为3424.1 km2,2012年以后水体增长尤为明显.本研究提出的联立LST的城市水体提取方法可以大幅提升城市地表水体的遥感提取精度,同时本研究也可为中国城市水资源管理及规划提供科学数据支撑.
Urban surface water extraction and its spatial-temporal changes based on land surface temperature and remote sensing cloud platform in China
Urban surface water is increasingly important to urban ecology and development,but complex urb-an surface environment,shadow,and other noise interferences always restrict the extraction of urban surface water,and long time series urban surface water data sets are especially scarce.In order to reveal the real changes of surface water in major cities in China,this study selected 33 built-up areas of provincial capitals in China,inclding Hong Kong Special Administrative Region,as the study area.Based on Google Earth Engine(GEE)cloud computing platform for remote sensing big data,long time series Landsat remote sensing images were used.Combining traditional water index and urban shadow index,an automatic water extraction method with Land Surface Temperature(LST)was constructed.Based on 24894 Landsat remote sensing images,annual surface water data with 30 m spatial resolution in major cities in China during 1990-2020 were automatically produced.The results showed that the overall accuracy of urban surface water extracted in this study was above 93%.During this study period,sur-face water showed an overall increasing trend,increasing from 2758.3 km2 in 1990 to 3424.1 km2 in 2020,and the area of urban surface water increased by 665.7 km2.This study highlights LST can effectively distinguish urb-an water and other urban cover types,and also provide scientific data support for urban water resources man-agement in China.
urban surface waterland surface temperaturehighly reflective featureurban shadowsremote sensing big data
万方数据
维普
