Journal
FOREST SCIENCE
Volume 60, Issue 1, Pages 128-136Publisher
OXFORD UNIV PRESS INC
DOI: 10.5849/forsci.12-088
Keywords
k-NN imputation; LiDAR remote sensing; random forest; boreal forest; forest structure
Categories
Funding
- Fonds quebecois de la recherche sur la nature et les technologies
- Ministere des Ressources Naturelles et de la Faune du Quebec
Ask authors/readers for more resources
Within the managed boreal forest of Quebec (Canada), there is a need to ensure the maintenance of age diversity across the forest resource, and, as a result, forest age has become an important variable in forest management plans. Forest age is a major driver of stand height, with site factors such as climate, water, and light availability, modifying height based on local conditions. The link between stand age and height is used to evaluate the site index, a commonly used measure of productivity, which is the basis of many forest management decisions. Stand age is extremely costly and laborious to measure in situ and, as a result, new technologies are required to provide added predictive power, such as light detection and ranging (LiDAR) which can provide insights into forest structure and site characteristics. In this article, we examined the utility of airborne LiDAR data to estimate stand age across 158 sample plots within a naturally regenerated boreal forest in Quebec, Eastern Canada. A suite of forest structure (height, first return penetration rate, and corresponding Weibull distribution parameters) and site attributes (elevation, slope, aspect, solar radiation, wetness index, catchment area, and flow path length) was derived from the LiDAR data. The k-nearest neighbor approach allowed imputation of stand age across the forest area, as well as providing key information on the predictive power of LiDAR metrics for estimating age in this forest region. Results indicate that LiDAR-derived predictors can estimate age well (R-2 = 0.83, root mean square error [RMSE] = 19%) with a predicted error of less than 10 years. Estimations using only forest structure variables were lower (R-2 = 0.74, RMSE = 22%). The most important forest structural predictors were stand height and Weibull crown shape parameters. Site predictors, in contrast, which had less predictive power, included elevation, slope, and catchment area with low to moderate importance.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available