Integrating area-based and individual tree detection approaches for estimating tree volume in plantation inventory using aerial image and airborne laser scanning data

Abstract

Remote sensing has been increasingly used to assist forest inventory. Airborne Laser Scanning (ALS) systems can accurately estimate tree height in forests, and are being combined with more traditional optical images that provide further details about the horizontal structure of forests. To predict forest attributes two main techniques are applied to process ALS data: the Area Based Approach (ABA), and the Individual Tree Detection (ITD). The first part of this study was focused on the effectiveness of integrating ALS data and aerial imagery to estimate the wood volume in Eucalyptus urograndis plantations using the ABA approach. To this aim, we analyzed three different approaches: (1) using only ALS points cloud metrics (RMSE = 6.84%); (2) using only the variables derived from aerial images (RMSE = 8.45%); and (3) the integration of both 1 and 2 (RMSE = 5.23%), which underestimated the true volume by 2.98%. To estimate individual tree volumes we first detected individual trees and corrected the density estimate for detecting mean difference, with an error of 0.37 trees per hectare and RMSE of 12.68%. Next, we downscaled the total volume prediction to single tree level. Our approach showed a better result of the overall volume in comparison with the traditional forest inventory. There is a remarkable advantage in using the Individual Tree Detection approach, as it allows for a spatial representation of the number of trees sampled, as well as their volume per unit area - an important metric in the management of forest resources.

Keywords

• Forest Inventory
• Airborne Laser Scanning
• Treetop Detection
• Eucalyptus Plantation
• Area-based Approach
• LiDAR