Dynamic stratification for vertical forest structure using aerial laser scanning over multiple spatial scales

The distinct vertical structure (i.e., "overstory-understory-grass") in natural forest stand causes significant dif-ferences in species composition, carbon-water cycles, and phenology variations in different vertical layers. Due to the complex forest structure and apparent topographic variations, it remains challenging to stratify a forest stand with highly overlapped tree crowns on a large spatial scale. This work successfully developed a lidar-based method to stratify vertical forest structures with different canopy covers and tree species by separating under -story using various stratification heights with the aerial laser scanning (ALS) data. Our results showed that: (1) The F-score values of forest overstory segmentation changed from 0.83 to 1 as forest density increased, and their values improved by 0.08 and 0.1 for coniferous and mixed forest stand by using our newly developed method, respectively. (2) The correlation coefficients (R2) between the visual-and ALS-based forest canopy cover increased from 0.72 to 0.84 by masking forest understory layers. The values of the F-score ranged from 0.5 to 1 as the plot density changed in the understory. (3) Varied stratification heights showed apparent effects on vertical forest structure separation. Our work provides a solid foundation for accurately estimating the forest structural parameters (i.e., leaf area index and biomass) and precision forestry.