Estimation of forest aboveground carbon storage based on GEDI and GF-6 in Jiyuan City, China

文献类型： 外文期刊

第一作者： Yang, Liu

作者： Yang, Liu;Yuan, Yabo;Sun, Jinhua;Zhang, Lei;Wang, Ting;Liang, Yazhen;Yang, Liu;Zhao, Hui;Liu, Xiangfeng

作者机构：

关键词： aboveground forest carbon storage; global ecosystem dynamics survey; GF-6; remote sensing

期刊名称：JOURNAL OF APPLIED REMOTE SENSING （ 影响因子：1.4； 五年影响因子：1.5 ）

ISSN：

年卷期： 2025 年 19 卷 2 期

页码：

收录情况： SCI

摘要： The accurate estimation of forest aboveground carbon storage is important for forest ecological function evaluation, carbon sequestration and sequestration accounting, and carbon market trading. The Global Ecosystem Dynamics Survey (GEDI) satellite provides an accurate method for measuring various vertical structural parameters of forests by penetrating the forest canopy. However, the distribution of footprints along the orbital trajectory is uneven and discontinuous, making it impossible to obtain the formation of county-level spatial distribution of carbon storage. The GEDI spaceborne lidar data and GF-6 satellite data are combined to estimate the forest carbon storage in Jiyuan City. The surface information of GEDI L2B and L2A footprints was obtained by Kriging interpolation methods with previously preprocessed footprints; the vegetation indices extracted from GF-6 and vertical variables extracted from GEDI are used to construct the aboveground carbon storage estimate models. Then, the carbon storage estimation model was established using three methods: support vector machine (SVM), random forest (RF), and convolutional neural networks (CNN) algorithms. In the GEDI footprint parameter selection model based on semi-variance, the optimal model for the digital elevation model (DEM) and RH90 to RH100 is a stable model, whereas the optimal models for fhd_normal, pai, and cover are exponential, Gaussian, and spherical, respectively. Analyzing the extracted 11 independent variable factors related to carbon storage include DEM, fhd_normal, RDVI, GWDRVI, GRVI, GDVI, NDVI, OSAVI, ARVI, RH92, and DVI through recursive features elimination (RFE) method. The R2=0.755 and RMSE=6.179 t/hm2 of the carbon storage estimation model established by the RF method were obtained after variable selection, and its evaluation accuracy was better than that of SVM and CNN. It indicates that it is feasible to obtain county-level carbon storage data by combining GEDI LiDAR data with GF-6 optical data. The research results also provide a new perspective for combining GEDI L2B data with other remote sensing images to estimate other forest structural parameters.

分类号：

  • 相关文献
作者其他论文 更多>>

微信小程序