Subtropical Forest Vertical Structure Classification And Above-ground Biomass Estimation In Large Area Using Airborne Lidar

Mapping the vertical structure of the forest canopy and accurately estimating above-ground biomass facilitates an in-depth understanding of the structure and function of forest ecosystems.Airborne Li DAR data and sample plot data were used to explore forest vertical structure classification and above-ground biomass estimation in a subtropical region of southern China with an area of 237.6?10~3km~2.The main research contents and methods are as follows:1)The method of forest vertical structure classification by airborne Li DAR.The vertical canopy profile(pseudo-wave)reflecting the vertical distribution of canopy was obtained by fitting the height-coverage frequency distribution of the point cloud using a 10th-order polynomial method.Canopy vertical structure parameters such as effective peak,canopy/canopy surface height,sub-layer height,and stand height/canopy surface height ratio were extracted by pseudo-wave and classification rules were established to classify the vertical structure of overstory(6types).Confusion matrix was used to assess the classification accuracy,and an area of 1369km~2 was selected for mapping to validate the generalizability of the classification rules.2)A method for estimating above-ground biomass of overstory based on vertical structure classification by airborne Li DAR.Based on the classification of the vertical structure of the forest canopy,the airborne Li DAR variables were divided into three groups from the three-dimensional structure of the forest canopy.The 44 multiplicative nonlinear regression model equations for above-ground biomass estimation in the overstory were constructed by a regular combination of exhaustive methods.The optimal above-ground biomass estimation models for each vertical structure type of four forest types(pine forest,fir forest,eucalyptus forest and broadleaf forest)were obtained by optimal selection,and the results of vertical structure classification of forests on the accuracy of biomass estimation were analyzed.3)Biomass estimation methods for shrub layer and herbaceous layer in forests with airborne Li DAR.Stepwise multiple regression,random forest,unary polynomial,unary quadratic polynomial,and multiplicative nonlinear regression methods were used to construct shrub and herbaceous layer biomass estimation models for the four forest types,and analyzed the effects of overstory layers cover on shrub layer and herbaceous layer biomass estimation.The main results of the study include:1)the overall classification accuracy of vertical structure types of overstory was 93.9%,and the Kappa coefficient was 0.913.The classification error rates of single-peaked,double-peaked and triple-peaked profiles were 6.2%,7.4%and9.1%,and the classification error rates of fir,pine,eucalyptus and broadleaf forests were 9%,6.4%,2.4%and 6.9%,respectively.The accuracy of each forest layer was higher than 96%,the omission ratio was less than 4%,and the commission error was less than 10%.The coverage rate of the classification rules of the mapped area reached 99.8%.2)Unclassified canopy vertical structure,the R-squared(R~2)of the above-ground biomass estimation models for the four forest types of fir,pine,eucalyptus,and broadleaf forests were0.65,0.65,0.74,and 0.50,and the relative root mean square errors(r RMSE)were 18.88%,20.73%,23.21%,and 32.46%.Following canopy vertical structure classification,the weighted average of r RMSE for the four forest types decreased by 7.20%,9.85%,1.91%,and 7.48%,respectively.When the vertical structure types of each forest type were clustered into two types,the weighted average of r RMSE decreased by 7.44%,7.17%,3.10%,and 1.78%,respectively.3)In each forest type,all shrub layer biomass estimation models had R~2 less than 0.45 and r RMSE greater than 100%,and all herbaceous layer biomass estimation models had R~2 less than 0.28 and r RMSE greater than 92%.After stratification according to overstory cover,biomass estimation of shrub and herb layers was slightly better than that without stratification when overstory cover was low,but the overall estimation was still poor.The estimation results of the overstory were worse than those of the non-stratified layer when the overstory coverage was higher.In this study,we developed a new method of airborne Li DAR forest vertical structure classification to achieve a clear spatial location and a rich ecological meaning of forest canopy vertical structure classification;through forest vertical structure classification,the accuracy of forest above-ground biomass estimation was improved effectively.It was difficult to estimate the biomass of shrub and herbaceous layers under the subtropical forest in a large region by low and medium density airborne Li DAR point clouds(mean point density 4.9 points/m~2)as influenced by the overstory.The study has an important reference value for the vertical structure classification and biomass estimation of airborne Li DAR subtropical forests in large regions.