摘要
为了构建有效的稀土矿区植被覆盖信息提取方法,准确掌握矿区土地治理的动态变化,对矿区周边的生态恢复效果进行评价,以江西省赣州市南部的稀土矿区为研究区,采用像元二分模型和随机森林算法,结合Sentinel-2 MSI和Sentinel-1 C波段雷达数据组成的多源遥感数据集,对稀土矿区的植被覆盖信息进行提取。结果表明:与像元二分模型相比,随机森林算法能更有效地提取矿区的植被覆盖分类结果,且总体精度和Kappa系数分别提高了8.49%和0.05。对2016—2020年矿区植被覆盖的时空变化进行分析得出,稀土矿区裸地面积不断减少,从3246 hm^(2)减少到1779 hm^(2);植被面积呈增加趋势,从2806 hm^(2)增加到4272 hm^(2)。整体来说,赣南各地的稀土矿区植被恢复效果均较好,其中赣县的植被恢复效果最明显。
In order to construct an effective method for extracting vegetation cover information from rare earth mining areas,to accurately grasp the dynamic changes of land management in mining areas,and to evaluate the ecological restoration effect around mining areas.In this study,a multi-source remote sensing dataset consisting of Sentinel-2 MSI and Sentinel-1 C-band radar data was used to extract vegetation cover information from a rare earth mining area in the southern part of Ganzhou City,Jiangxi Province,using the dimidiate pixel model and random forest algorithm.The results showed that compared with the dimidiate pixel model,the random forest algorithm could more effectively extract the classification results of vegetation coverage in mining areas,and the overall accuracy and Kappa coefficient were increased by 8.49%and 0.05,respectively.The analysis of the spatial and temporal changes of vegetation cover in mining areas from 2016 to 2020 showed that the area of bare land in rare earth mining areas had been decreasing,from 3246 hm^(2)to 1779 hm^(2);the area of vegetation cover had shown an increasing trend,from 2806 hm^(2)to 4272 hm^(2).Overall,the vegetation restoration effect of rare earth mining areas in southern Jiangxi is relatively good,and the vegetation restoration effect in Ganxian County is the most obvious.
作者
李媛
LI Yuan(College of Earth Sciences,East China University of Technology,Nanchang,Jiangxi 330013,China)
出处
《天津农业科学》
CAS
2023年第1期49-57,共9页
Tianjin Agricultural Sciences
关键词
稀土矿
植被覆盖度
像元二分模型
随机森林
时间序列分析
rare earth mining area
fractional vegetation cover
dimidiate pixel model
random forest
time series analysis
