Multi-scale forest typologies using novel vertical layer metrics from airborne LiDAR in temperate mixed forests

Precisely quantifying forest vertical structure is fundamental to understanding biodiversity, productivity, and ecosystem functions. Airborne Laser Scanning (ALS) provides an unparalleled capacity to capture three-dimensional canopy architecture, yet existing metrics often overlook attributes of individual layers. To address these limitations, this study developed novel vertical layer indices, applied clustering, and tested robustness across scales. We used high-density ALS data (over 200 points·m −2 ) covering 5.6 km 2 of temperate mixed forest in South Korea, encompassing diverse species compositions and successional stages. For scale-dependent analysis, circular analytical units with radii of 5- and 10-m were applied. A vertical layer segmentation procedure was optimized, achieving close agreement with field assessments (Mean Absolute Error ≈ 0.4 m), and provided the number of vertical layers and their ranges. From these results, ten indices were created to quantify layer height, gap size, and diversity at both species and unit scales. The indices were clustered into four to five interpretable types using Principal Component Analysis (PCA) and Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN), which explained over 70% of structural variance. For the 10-m units, the forest was classified as monolayer conifers, mixed mid-story, broadleaved multi-layer, stratified late-successional, and noise (structurally indeterminate units left unassigned), accounting for 31.8%, 39.8%, 11.3%, 5.9%, and 11.2%, respectively. Scale comparisons showed the Cumulative Layer Range (CLR) was stable, with differences within ±1.7% across six species out of eight. In contrast, Shannon's H′ was sensitive, increasing by 0.08 (25%) in 5-m units for Pinus densiflora . This index-based approach quantifies forest structural characteristics and captures scale-dependent vertical complexities.