“Old” is not precise enough: Airborne laser scanning reveals age-related structural diversity within old-growth forests

Abstract

Old-growth forests of different ages provide specific structures, habitats and ecosystem services. Methods to distinguish this internal diversity are still rare, especially in boreal forests. This research therefore aims to determine the ability of Airborne Laser Scanning (ALS) technology to identify age-related structural diversity in old-growth boreal forests. The study area was located in primary boreal forests in Quebec (Canada) dominated by black spruce (Picea mariana). This area contained 71.8 km2 of early old-growth forests (burned 110 years ago), 17.1 km2 of late old-growth forests (protected areas; unburned for at least 250 years) and 370 km2 of old-growth forests of unknown age (> 125-years-old). We divided the study area into 1 ha tiles, where we extracted seven ALS indices representing vertical and horizontal forest structure. We trained random forest models using an iterative approach to discriminate between early and late old-growth forests based on ALS indices. Model predictions were applied to the old-growth tiles of unknown age, and to 86 field plots (28 from provincial forest surveys and 58 from a dedicated survey of old-growth forests) to evaluate the predictive capacity of the models. The models very accurately distinguished early and late old-growth forests (error-rate = 4.9%). Old-growth survey plots confirmed model ability to discriminate early and late old-growth forests, but not provincial survey plots, possibly because of a lower reliability of these data when forest age exceeds 150 years. Model predictions for tiles of unknown age highlighted the presence of very large tracts of late old-growth forests within a matrix of old-growth forests of intermediate age (≈150–200 years). Overall, ALS-data can contribute to a finer structural age distinction and mapping of boreal old-growth forests. This enhanced knowledge of old-growth landscapes will greatly help to improve their protection, restoration and management. The scarcity of reliable field data for model evaluation is, however, a limitation to be addressed.