Modeling forest stand variability by means of remotely sensed vertical structure parameters

Vertical forest structural complexity plays a substantial role for biomass, natural regeneration, and habitat provision, yet in-situ mapping remains costly. Light Detection and Ranging (Lidar) has shown potential to derive relevant structural forest parameters. This study estimates the Standard Deviation of Tree Heights (SD HT ) and the Gini Coefficient of Basal Area per Hectare (GC BPH ) from Lidar metrics using Partial Least Squares (PLS) and Random Forest (RF) models. Together, SD HT and GC BPH capture height variability and tree size inequality, which influence cover and forage availability for mule deer winter habitat. The study was conducted in the Alex Fraser Research Forest, British Columbia. Models were applied to Lidar data from 2013 and 2019 to assess temporal trends. SD HT was well estimated (R 2 = 0.76, rRMSE = 13 %), while GC BPH showed moderate accuracy (R 2 = 0.42/0.46, rRMSE = 18 %). RF achieved R 2 of 0.72 and 0.50 for SD HT and GC BPH , respectively. Simultaneous modeling improved GC BPH estimates. L-Scale emerged as a key Lidar metric. Results support applications in habitat modeling, fuel modeling, biomass estimation, and disturbance assessment.