This study aimed to investigate the accuracy of crown factor extraction from point cloud data by Terrestrial laser scanning (TLS), and the error of tridimensional green biomass calculated by the voxel simulation method, to verify the feasibility of using ground-based lidar data combined with the voxel simulation method to calculate tridimensional green biomass, and construct Larix gmelinii tridimensional green biomass models.

In this paper, 42 Larix gmelinii in the sample plots in Mengjiagang in Jiamusi City, Heilongjiang Province were studied, and the sample plot area was 0.06-0.12 hm², with 3 366 trees. Four Larix gmelinii sample plots (208 trees) were selected, and the single tree factor parameters could be obtained by TLS scanning and LiDAR360 software processing. The “Voxel-based simulation method” was used to realize the calculation of the tridimensional green biomass of individual trees in the four Larix gmelinii sample plots (208 trees) by Matlab programming, and then the precision of tridimensional green biomass calculated by the “voxel simulation method” and “crown formula method” was analyzed. The tridimensional green biomass model of individual Larix gmelinii was constructed based on the crown factors extracted from the TLS and the tridimensional green biomass calculated by Matlab. The tridimensional green biomass model of single wood was combined with the data of each wood survey in 42 sample plots to calculate the tridimensional green biomass of each wood. The tridimensional green biomass of each wood in the sample plots could be accumulated to obtain the tridimensional green biomass of the sample plots, which was converted into the corresponding tridimensional green biomass of the stand according to the area of the sample plots. The tridimensional green biomass model of stands was constructed by combining stand factor data of the 42 plots.

The crown factors obtained by TLS met the accuracy requirements, among which the extraction accuracy of diameter at breast height was the best (P=97.14%), followed by tree height (P=90.96%), crown width (P=89.25%) and crown length (P=77.92%). The “voxel simulation method” was the most stable to calculate the tridimensional green biomass value using Matlab with the step size k=0.1 m. The fitting and testing effect of the three-factor tridimensional green biomass model of individual trees was the best (R²=0.846, R_MSE= 0.250, M_AE=0.210, M_APE=0.099, P=97.314%). The result of the two-factor tridimensional green biomass model of stands was the best (R²=0.706, R_MSE= 0.138, M_AE=0.120, MAPE=0.013, P=99.142%).

The accuracy of crown factors obtained by TLS has high accuracy, and the “voxel simulation method” is more accurate to calculate the three-dimensional green value. The individual-tree model and stand model constructed by TLS data improve the accuracy of tridimensional green biomass estimation in large-scale areas, providing a new method for the tridimensional green biomass estimation in large-scale areas.