Construction of aboveground biomass models for different age groups of Casuarina equisetifolia
We used Casuarina equisetifolia of different age groups on Pingtan Island,Fujian Province as the research objects to extract the high-precision structural parameters of individual C.equisetifolia trees,establish diversified aboveground biomass estimation models for C.equisetifolia,and quickly and efficiently investigate and monitor the growth trends of C.equisetifolia stands.We utilized point clouds data acquired from unmanned aerial vehicles(UAV)equipped with LiDAR(UAV-LiDAR)and fusion laser point clouds(Fusion-LiDAR)formed by combining data from both UAV and terrestrial LiDAR scanners(TLS)to rapidly and accurately obtain the key structural parameters of C.equisetifolia at the individual tree scale,including tree height,crown width,and diameter at breast height.Combining field survey data,we used algorithms such as partial least squares(PLS),random forest(RF),and back propagation neural network(BPNN)to construct an aboveground biomass estimation model for C.equisetifolia.The results indicated the following:the canopy height model constructed based on Fusion-LiDAR point clouds significantly outperformed that constructed solely based on UAV-LiDAR in terms of individual tree segmentation accuracy,particularly in young forests;the determination coefficients(R2)for tree height and crown width derived from Fusion-LiDAR point clouds were generally higher than that extracted from UAV-LiDAR point clouds.Specifically,the R2 values for tree height and crown width increased by 11.41％and 16.73％,respectively,in overmature forests compared with those obtained from UAV-LiDAR point clouds;among the three models,i.e.,BPNN,PLS,and RF,the BPNN model exhibited superior performance in predicting the biomass of C.equisetifolia across various age classes,with R2 values exceeding 0.75 and relative prediction deviation values surpassing 1.40 for all age groups;and as the age class increased,the accuracy of individual tree segmentation,precision of stand parameter extraction,and predictive performance of the models tended to decline gradually.The fusion of UAV-LiDAR and TLS significantly improved the accuracy of individual C.equisetifolia segmentation and the extraction precision of individual tree structural parameters.The BPNN model demonstrated superior performance in predicting the aboveground biomass of C.equisetifolia of different forest ages,further enhancing modeling efficiency and prediction accuracy.
Casuarina equisetifoliaunmanned aerial vehicleterrestrial LiDARpartial least squaresrandom forestback propagation neural networkindividual tree segmentationaboveground biomass
万方数据
维普
