A Novel Carbon Stocking Estimation Through Continuous Catalog Learning

TL;DR: A novel carbon stocking estimation method using continuous learning and high-resolution imagery data improves accuracy and provides valuable insights into diverse forest types.

Abstract: Abstract The role of forests as a significant mitigator of anthropogenic \coo emissions is integral to the global efforts to combat climate change. Precise monitoring of carbon sequestration is a high priority for governments and organizations striving to achieve a zero atmospheric carbon balance.The combination of remote sensing and machine learning has emerged as a powerful tool for estimating above-ground biomass (AGB) in diverse ecosystems.However, accurately measuring the carbon sequestrated by forests at various scales and forest types is still a technological and practical challenge carrying the risk of carbon overestimation, which may lead to wrong decision-making, flawed climate change mitigation actions, financial losses, and more.In this study, we propose a novel method to address this need using a catalog-based carbon stocking approach integrated within a continuous learning mechanism.Our method estimates forest carbon stocking based on high-resolution aerial LiDAR and multispectral imagery, offering valuable insights beyond the limitations of satellite-based imagery. Through the combination of unsupervised learning and a ground-based calibration procedure, we successfully delineated 10 distinct forest types within a vast area of mixed forest spanning 55,000 hectares.The calibrated carbon stocking estimation demonstrated superior accuracy compared to satellite-based analysis, as evidenced by rigorous cross-validation using an unprecedented dataset of 802 ground-surveyed plots.Employing a continual learning mechanism, the system can estimate carbon stocking on a 25x25m grid, enabling generalization across multiple forest types and scales of aggregation within a unified framework.This work serves as a starting point for further research to enhance the accuracy of carbon stocking monitoring and contribute to the momentum of carbon sequestration efforts.In addition to the scientific significance of this paper, we have made a notable contribution by providing access to a comprehensive dataset. This dataset encompasses high-density LiDAR point cloud data and multispectral imagery data, covering more than 13,000 acres and including samples from 1,725 individual trees.To our knowledge, this represents the most extensive combined aerial-ground dataset published in this field to date. Our objective in sharing this dataset is to facilitate ongoing research and set a benchmark for advancements in the domain of carbon stocking estimation.