基于Google Earth Engine的塔里木河下游土地利用变化

Land Use Changes in the Lower Reaches of the Tarim River Based on Google Earth Engine

在谷歌地球引擎(Google Earth Engine,GEE)遥感大数据云平台的支持下,基于随机森林算法,结合Landsat系列遥感影像、光谱指数和地形特征,解译塔里木河下游土地利用类型,分析2000-2020年塔里木河下游土地利用时空变化特征。结果表明GEE环境下,选用随机森林算法,能快速实现区域尺度土地利用分类,总体分类精度均大于85%,Kappa系数高于0.82,较难区分的林地、草地的生产者精度超过82%;2000-2020年塔里木河下游土地利用结构稳定,草地、未利用地和林地为主要土地利用类型,土地利用变化以草地和未利用地转出为主,变化区域呈分散分布,其次是水体连续扩增,增长区集中于大西海子水库以下的旧河道及台特玛湖,林地面积变化不稳定,先减少后增加,耕地、建设用地面积变化不明显。

With the support of the google earth engine(GEE)remote sensing big data cloud platform,based on the random forest algorithm,combined with the Landsat series of remote sensing images,spectral indices and topographic features,interpret the land use types in the lower reaches of the Tarim River,and analyze 2000-spatio-temporal changes of land use in the lower reaches of the Tarim River in 2020.Results indicate that in the GEE environment,the random forest algorithm can be used to quickly realize land use classification at a regional scale.The overall classification accuracy is greater than 85%,the Kappa coefficient is higher than 0.82,and the producer accuracy of difficult to distinguish forest land and grassland exceeds 82%.From 2000 to 2020,the land use structure in the lower reaches of the Tarim River will be stable,with grassland,unused land and woodland as the main land use types.The land use change is mainly due to the transfer of grassland and unused land,and the changed areas are scattered,followed by continuous expansion of water bodies.The growth area is concentrated in the old river channel and Taitma Lake below the Daxihaizi Reservoir.The woodland area changes steadily,first decreasing and then increasing,and the area of cultivated land and construction land does not change significantly.