以青海省8个县为试点区,利用Sentinel-2A和Landsat-8中分辨率遥感数据,结合2018—2020年试点单位的森林督查成果及森林资源管理“一张图”年度更新数据,协同运用LSMA光谱混合分析模型、Li-Strahler几何光学模型,对试点县2020年森林资源变化情况进行动态监测和实证研究。结果显示:在15个月的时间跨度中,共监测疑似林地变化图斑693个,总面积1 286.382 7 hm²,主要发生在公益林地范围内,占疑似变化图斑总面积97.48%,总体准确率达到87.30%,满足森林资源动态监管要求;10m分辨率的中分卫星影像能检测到的最小图斑面积达到0.1 hm²,但主要集中在郁闭度较大的林区;在计算机自动检测的基础上,加入人工识别,能进一步提高变化检测的准确率。

Based on Sentinel-2A and Landsat-8 middle-resolution remote sensing data,combined with the three-phase forest inspection results and the annual update data of the "one map" of forest resources man-agement from 2018 to 2020,LSMA spectral hybrid analysis model and Li-Strahler geometric optics model were used to study the dynamic monitoring and empirical research on the changes of forest resources in 8 pilot counties of Qinghai Province.The results showed that,in the 15-month time span,a total of 693 change patterns with total area of 1286.3827 hm² of forest land were monitored,which mainly occurred within the public welfare woodland and accounted for 97.48% of the total area.And the overall accuracy rate reached 87.30%,which met the requirements for dynamic monitoring of forest resources;Also,the smallest patch area that could be detected by the 10-m resolution satellite image was 0.1 hm²,but it was mainly concentrated in forest areas with greater canopy closure;Furthermore,on the basis of computer automatic detection,adding manual recognition could further improve the accuracy of change detection.