Impacts Of Airborne Li Dar Point Cloud Density On The Inversion Of Average Tree Height And Stock Volume,and Single Tree Segmentation In Plantation Forests

LiDAR(Light Detection and Ranging)is a mature and stable active remote sensing technology that has been widely applied in forest resource surveys,forest parameter estimation,forest biomass estimation,and individual tree identification.Currently,airborne LiDAR is mainly used to acquire three-dimensional forest data using small-footprint discrete-return systems in forestry.Point cloud density is one of the critical factors affecting forest parameter estimation and individual tree segmentation.The higher the point cloud density,the more detailed the forest structure,the clearer the tree crown shape,but also the higher the cost of point cloud data acquisition and processing,and the more information redundancy.In this study,we used artificially planted forests in Taocun Forest Farm,Song County,Henan Province,located in the warm temperate zone,as the research object,and merged airborne LiDAR point cloud data from 2020 and 2021 as the high-density point cloud data source.Using Normal Space Sampling(NSS)and Voxel Downsampling(VOX)methods,the high-density point cloud data were resampled into seven different point cloud density levels,namely 30,25,20,15,10,5,and 1 pts/m2.Stand average tree height,stand biomass,and individual tree segmentation results were obtained through modeling and tree segmentation at different point cloud densities.By comparing and analyzing,we explored the effects of point cloud density and forest type on LiDAR-derived stand average tree height,stand biomass,and individual tree segmentation results in warm temperate artificial forests.The main results of this study are as follows:1.Under different point cloud densities,the determination coefficient(R2)for LiDAR-derived stand average tree height ranges from 0.84 to 0.89,with the corresponding root mean square error(RMSE)ranging from 0.74 m to 0.84 m.Meanwhile,the R2 for LiDAR-derived stand biomass ranges from 0.78 to 0.96,with the corresponding RMSE ranging from 27.83 m3/hm2 to 12.27 m3/hm2.2.When the point cloud density reaches 20 pts/m2 and above,the accuracy(P),recall(R),and F1-Score of individual tree segmentation gradually increase to 70% and balance out.For the average tree height,when the point cloud density is greater than or equal to 5pts/m2,the RMSE of individual tree segmentation remains around 1 m.3.For different forest types,the RMSE of average tree height for coniferous forests under different point cloud densities ranges from 0.53 m to 0.7 m,while that of broadleaved forests ranges from 0.81 m to 0.95 m.The RMSE of coniferous forest biomass ranges from 13.46 m3/hm2 to 19.01 m3/hm2,while that of broad-leaved forests ranges from 10.17 m3/hm2 to 31.51 m3/hm2.In terms of individual tree segmentation,the average F1-score for coniferous forests under different point cloud densities ranges from 74% to75%,while that of broad-leaved forests ranges from 59% to 62%.When the point cloud density is greater than or equal to 5 pts/m2,the average tree height RMSE for coniferous and broad-leaved forests under point cloud segmentation methods are between 0.65 m-0.68 m and 1.077 m-1.083 m,respectively,while under CHM segmentation methods,the average tree height RMSE for coniferous and broad-leaved forests are between 0.83 m-0.95 m and 0.81 m-0.95 m,respectively.