Modeling post-logging height growth of black spruce-dominated boreal forests by combining airborne LiDAR and time since harvest maps

Abstract

Increase in forest disturbance due to land use as well as climate change has led to an expansion of young forests worldwide, which drives global carbon dynamics and timber allocation. This study presents a method that combines a single airborne LiDAR acquisition and time since harvest maps to model height growth of post-logged black spruce-dominated forests in a 1700 km2 eastern Canadian boreal landscape. We developed a random forest model in which forest height at a 20 m × 20 m pixel resolution is a function of stand age, combined with environmental variables (e.g., slope, site moisture, surface deposit). Our results highlight the model's strong predictive power: least-square regression between predicted and observed height of our validation dataset was very close to the 1:1 relation and strongly supported by validation metrics (R2 = 0.74; relative RMSE = 19%). Environmental variables thus allowed to accurately predict forest productivity with a high spatial resolution (20 m × 20 m pixels) and predicted forest height growth in the first 50 years after logging ranged between 16 and 27 cm·year−1 across the whole study area, with a mean of 20.5 cm·year−1. The spatial patterns of potential height growth were strongly linked to the effect of topographical variables, with better growth rates on mesic slopes compared to poorly drained soils. Such models could have key implications in forest management, for example to maintain forest ecosystem services by adjusting the harvesting rates depending on forest productivity across the landscapes.