基于星载ICESat-2/ATLAS数据的天然林郁闭度空间异质性研究
Spatial heterogeneity of natural forest canopy cover based on spaceborne ICESat-2/ATLAS data
[目的]星载激光雷达能够提供全球尺度的采样观测,借此机会,本研究基于高密度的ICESat-2/ATLAS采样足迹,探讨香格里拉市天然林郁闭度的空间异质性及其与地形因子的关系,为科学经营管理天然林提供参考。[方法]本研究以香格里拉天然林为研究对象,基于ICESat-2/ATLAS数据,结合 51 块实测样地,采用四种机器学习算法(k-NN、SVM、RF、GBRT)构建光斑尺度的郁闭度估测模型,选择性能最优的模型预测天然林地内的 1100 个光斑足迹郁闭度,通过采用空间自相关分析、变异函数分析、相关性分析、方差分析、地理加权回归等方法,对光斑足迹内的郁闭度空间异质性及其与地形因子的关系进行研究。[结果]四种机器学习(KNN、SVM、GBRT、RF)算法中,GBRT模型具有最好的预测精度(R2 = 0.94,RMSE = 0.041);研究区天然林郁闭度的空间变异最适合用指数模型来描述,郁闭度在 0～10200 m范围内存在强烈的空间自相关;基于Co-Kriging的光斑郁闭度空间插值具有良好的精度(R2 = 0.59,RMSE = 0.07);地形在一定程度上解释了天然林郁闭度的空间变异,海拔对空间变异的影响较大。[结论]天然林郁闭度在一定的空间尺度上显示出较强的异质性,且会由于地形因子的变化产生差异,在影响程度方面,海拔最大,坡度次之,坡向最小。
[Objective]Spaceborne LiDAR can provide global scale sampling and observation.Based on the high-density ICESat-2/ATLAS sampling footprint,this study explores the spatial heterogeneity of natural forest canopy cover(NFCC)in Shangri-La and its relationship with topographic factors,so as to provide references for scientific management of natural forests.[Method]This study takes Shangri-La natural forest as the research object.Based on ICESat-2/ATLAS data and combined with 51 measured plots,a machine learning model is built and 1100 light spot footprints in natural forest land are predicted.Spatial autocorrelation analysis,variance analysis,correlation analysis,geographical weighted regression and other methods are adopted.The spatial heterogeneity of canopy in light spot footprint and its relationship with topographic factors were studied.[Result]Among the four machine learning algorithms(KNN,SVM,GBRT,RF),GBRT model has the best prediction accuracy(R2 = 0.94,RMSE = 0.041).The spatial variation of natural forest canopy density in the study area is best described by the exponential model,and there is a strong spatial autocorrelation in the range of 0～10200 m.The spatial interpolation of spot density based on Co-Kriging has good accuracy(R2 = 0.59,RMSE = 0.07).The spatial variation of natural forest canopy density in the study area is best described by the index model. Topography explains the spatial variation of natural forest canopy to a certain extent,and elevation has a greater effect on the spatial variation.[Conclusion]The NFCC showed strong heterogeneity on a certain spatial scale,and the difference is caused by the change of topographic factors.In terms of the impact degree,the altitude is the largest,followed by the slope,and the slope direction is the least.
余金格;舒清态;胥丽;罗绍龙;国朝盛;王书伟;周文武;
西南林业大学 昆明 650224;西南林业大学 昆明 650224;西南林业大学 昆明 650224;西南林业大学 昆明 650224;西南林业大学 昆明 650224;西南林业大学 昆明 650224;西南林业大学 昆明 650224;
forest canopy cover natural forest spaceborne LiDAR topography factor spatial heterogeneity.
1994-1995 / 2
评论