Stratifying Forest Vertical Layers Using Three Dimensional (3-D) Point Cloud Data

Natural forest has special overstory-shrub-grass layers structure,most of them can be separated into forest overstory layer and understory layer including shrub,grass,and moss.The forest background refers to all the materials below the forest canopy,including understory vegetation,soil,rocks,snow,and a mixture of them.Separating overstory layer and understory layer accurately and quantitatively has important scientific significance and practical value.Traditional passive optical remote sensing has limitation on direct acquisition of 3 dimensional information,which cannot effectively acquire forest stereo structure.Forest stratification is possible with the generation of lidar technique.This study use the aerial laser scanning system to acquire the point cloud data of Washington botanical garden and Panther Creek forest in USA.Segment the forest point cloud data from Washington botanical garden and Panther Creek forest using ALS-based segmentation algorithm,and acquire the height of single trees for forest stratification,which can be separated into the understory layer and overstory layer.Binary graph generated from overstory point cloud data is used for eliminating background information of HyMap data.Then,validate and analysis the result.Secondly,a new TLS based tree crown segmentation algorithm is proposed in this study.The algorithm performs tree segmentation process in order from bottom to top(trunk to canopy).Forest data in Evo,Finland is segmented into individual trees to obtain tree height information based on the algorithm proposed in this study.Forest is stratified three-dimensionally by trees height information to obtain understory layer and overstory layer.Two 3-D voxel space are created from understory layer extracted by algorithm and understory from manual reference data.Compare the voxel one to one,then,validate and analysis the result.The main conclusions of this study could be drawn as follows:(1)Removal of understory vegetation information can effectively improve the estimation accuracy of forest canopy leaf area index.Study result shows the correlation between Normalized Differential Vegetation Index(NDVI)and actually measured effective leaf area index(LAI)of canopy raised from 0.086 to 0.591.Besides,comparing with Simple Ratio(SR)(correlation with canopy raised from 0.209 to 0.560)and Reduced Simple Ratio(RSR)(correlation with canopy raised from 0.147 to 0.358),NDVI is the most sensitive to canopy leaf area index(correlation increased by 0.505).(2)Forest canopy closure has a significant impact on the penetration of ALS data.Study result shows understory layer can be effectively removed in medium density and low density forest.High-density forest are not penetrated by trees and leaves,so it is necessary to adjust a series of parameters to complete the separation of understory vegetation.(3)Understory layer can be acquired effectively by TLS.Study result shows low-density forest has less understory in canopy cover with a maximum coverage of 0.358 and a minimum coverage of 0.07.High-density forest has more understory in canopy cover with a maximum coverage of 0.628 and a minimum coverage of 0.3.