Machine Learning Method for Carbon Stock Classification with Drone and GEE Data

Accurately classifying carbon stock is essential for tackling climate change, as it helps improve forest management and carbon storage efforts. However, traditional measurement methods are often costly, time-consuming, and require extensive fieldwork. To address these challenges, remote sensing combined with machine learning offers a more efficient and scalable solution. This study explores the use of XGBoost and Random Forest classifiers to classify carbon stock levels using drone and Google Earth Engine (GEE) imagery, with VGG16 extracting features from the images. The dataset, collected from field plots at Telkom University in Bandung, Indonesia, consists of 2,114 drone images and 2,526 GEE images, labeled into three categories: low, medium, and high carbon stock. The results show that XGBoost applied to drone imagery achieves the highest accuracy of 90.79%, outperforming Random Forest and GEE-based models. This study underscores the potential of deep learning and ensemble methods in improving carbon stock estimation, supporting better environmental conservation.