Machine learning approaches for estimating forest stand height using airborne LiDAR data in New Brunswick forests
Loading...
Date
2024-10
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
University of New Brunswick
Abstract
Accurate forest stand height estimation is crucial for effective forest management, ecological studies, and carbon stock assessment. Airborne LiDAR is an advanced remote sensing technology extensively used in forest inventory. This study investigates machine learning techniques for estimating forest stand height using airborne LiDAR data in New Brunswick, Canada, evaluating Linear Regression (LM), Random Forest (RF), Gradient Boosting Machine (GBM), and Support Vector Machines (SVM). Results show that Linear Regression and Gradient Boosting Machine models provide the highest accuracy, with R2s up to 0.76 and rMSEs as low as 1.06m. Conversely, the Random Forest model underperformed. This study demonstrates the value of combining high-resolution LiDAR data with machine learning models to improve forest stand height estimation accuracy, supporting sustainable forest management and conservation efforts.