Aerial Laser Scanning for Forest Change Assessment: An Evaluation Framework for Forest Change Map Products

Accurate mapping of forest cover changes is essential for monitoring the health of the vegetation, evaluating the forests' capacity to store the carbon dioxide absorbed from the atmosphere, and comply with international sustainable forest management goals. In the literature, the Global Forest Change product proposed by Hansen and estimated at a 30 meters spatial resolution from Landsat optical data is a widely used dataset for detecting forest loss. Another widely used approach involves studying how a Land Use/Land Cover (LULC) product changes over time. In this regard, the partnership between Google and the World Resources Institute (WRI) has led to the development of the Dynamic World dataset which provides LULC information at a high spatial resolution of 10 meters and a daily temporal frequency, estimated from Sentinel 2 optical data. These change maps are essential for researchers worldwide monitoring dynamic shifts in forest landscapes.

 

Among various Earth observation (EO) technologies, Light Detection and Ranging (LiDAR) scanning stands out for its potential in obtaining detailed information on forest structures over large geographical areas with high spatial resolution and accuracy. Airborne LiDAR scanning data can be used to measure the height of trees above the ground topography producing Canopy Height Models (CHMs). This work proposes a robust procedure for CHM computation at the desired spatial resolution using the LiDAR point cloud detected by Aerial Laser Scanning (ALS) in the area of interest. The differences between CHMs at different observation times define tree cover change maps with high accuracy in the specified area.

 

The study employs publicly available ALS data covering Estonia to calculate CHMs with a spatial resolution of 5 meters. Utilizing these data, we compute high-definition tree cover change maps in the temporal window between 2018 and 2021 providing an accurate quantification of forest loss. The resulting tree cover change maps play a crucial role in evaluating widely used forest change maps such as Global Forest Change and Dynamic World derivatives. Through the establishment of a robust evaluation framework, including the comparison of common metrics (e.g. commission and omission error) with existing forest change maps, the study contributes significantly to the reliability analysis of forest change map products. The research addresses challenges in quantifying regional forest changes and offers valuable insights for researchers and policymakers engaged in sustainable forest management.