Yu Li-Li, Sun Li-Shuang, Zhang Dan-Hua, Liu Miao, Xie Zhi-Wei, Ping Xiao-Ying
School of Transportation Engineering, Shenyang Jianzhu University, Shenyang, 110117, China.
CAS Key Laboratory of Forest Ecology and Management, Institute of Applied Eco-logy, Chinese Academy of Sciences, Shenyang 110016, China.
Ying Yong Sheng Tai Xue Bao. 2020 Dec;31(12):4091-4098. doi: 10.13287/j.1001-9332.202012.014.
The land cover of Bohai Rim region has changed greatly due to urbanization and economic development. Monitoring the land cover with high accuracy and real time is the most important basis for relevant researches. Traditional single-machine processing mode is difficult to realize rapid monitoring for large-scale and long-time series. The emergence of remote sensing big data makes it possible to combine computing platform and massive data. The land cover maps of study area were interpreted based on Google Earth Engine (GEE) platform with decision tree (CART) method from 2000 to 2019. The land cover change was analyzed, and the interpretation results using different data sources were compared. The results showed that the GEE platform could realize the rapid land cover interpretation in a large area, which interpreted coastal wetlands and other cover types with high accuracy over 80% comparing the surveyed points. Compared with Landsat images, the Sentinel-2A images interpretation results had a great improvement in accuracy, which increased from 85% to 95%, and thus more detailed surface information could be reflected. In 2000, the area of wetland, build-up area, farmland, forest, and water in the study area were 1612.5, 5734.9, 32074.8, 11853 and 3504.3 km, accounting for 2.9%, 10.5%, 58.6%, 21.6% and 6.4% respectively. By 2019, wetlands had been reduced by 775.1 km, with a decline of 40.1%; built-up area increased by 5310.5 km with an increasing rate of 92.6%. The area of farmland, forestland and water area decreased 1841.6, 1823.5 and 870.3 km, with a decreasing rate of 5.7%, 24.8% and 48.1%, respectively. The coastal urbanization process caused the occupation of built-up area to other land use types, which was the main driving force of land cover change in the study area.
由于城市化和经济发展，环渤海地区的土地覆盖发生了很大变化。高精度实时监测土地覆盖是相关研究的最重要依据。传统的单机处理模式难以实现对大规模、长时间序列的快速监测。遥感大数据的出现使得计算平台与海量数据的结合成为可能。基于谷歌地球引擎（GEE）平台，采用决策树（CART）方法对2000年至2019年研究区域的土地覆盖图进行了解译。分析了土地覆盖变化，并比较了使用不同数据源的解译结果。结果表明，GEE平台能够实现大面积土地覆盖的快速解译，与实测点相比，其对沿海湿地等覆盖类型的解译精度超过80%。与陆地卫星图像相比，哨兵-2A图像的解译结果在精度上有了很大提高，从85%提高到95%，从而能够反映更详细的地表信息。2000年，研究区域内湿地、建成区、农田、森林和水域面积分别为1612.5、5734.9、32074.8、11853和3504.3平方千米，分别占2.9%、10.5%、58.6%、21.6%和6.4%。到2019年，湿地减少了775.1平方千米，降幅为40.1%；建成区增加了5310.5平方千米，增长率为92.6%。农田、林地和水域面积分别减少了1841.6、1823.5和870.3平方千米，减少率分别为5.7%、24.8%和48.1%。沿海城市化进程导致建成区占用其他土地利用类型，这是研究区域土地覆盖变化的主要驱动力。
