Speaker
Description
Accurately estimating the aboveground biomass in oil palm plantations is crucial for sustainable resource management, particularly as agricultural sectors globally work toward achieving net-zero carbon emissions and adapt to the realities of climate change. Precise biomass estimates support carbon stock assessments, contributing to climate change mitigation strategies and enabling plantations to monitor and enhance their carbon sequestration potential. However, traditional ground-based methods for aboveground biomass assessment, while precise at the plot level, are often labor-intensive, time-consuming, and costly, limiting their scalability for large areas. Non-destructive allometric approaches, commonly employed for plot-based estimations, face challenges in capturing the spatial variability of tree growth across heterogeneous tropical landscapes, which can reduce the accuracy of biomass modeling. To address these limitations, this study introduces a novel methodology that integrates drone-based LiDAR and geospatial features within a machine learning framework, further enhanced by Multiscale Geographically Weighted Regression (MGWR). The methodology involves the collection of drone-based LiDAR data for detailed individual tree detection and structural parameters, and the integration of these datasets with topographical information. Machine learning techniques are applied for initial modeling, while MGWR is utilized to account for localized environmental differences, providing a more refined and spatially adaptive analysis compared to conventional global models. The preliminary results highlight significant improvements in AGB estimation accuracy, demonstrating the potential of this integrated, data-driven approach to enhance plantation management. By offering a cost-effective and scalable alternative to traditional ground-based methods, this research supports informed decision-making, contributing to sustainable oil palm cultivation and aligning with climate change mitigation and net-zero carbon targets. This methodology underscores the utility of advanced remote sensing combined with machine learning to navigate the complexities of tropical plantation landscapes and optimize resource allocation.
Keywords: Drone-Based LiDAR, Aboveground Biomass Estimation (AGB), Machine Learning, Multiscale Geographically Weighted Regression (GWR), Oil Palm.
