Merging multi sensor time series remote sensing and machine learning empowers forest biomass estimation

Abstract

Effective monitoring of forest aboveground biomass (AGB) stands as a cornerstone for global carbon cycle research and strategies aimed at mitigating climate change. Accurate assessment of forest biomass provides crucial insights into carbon stocks, ecosystem health, and sustainable resource management. Remote sensing presents a practical solution for monitoring and assessing vast forest biomass across extensive areas, contrasting with the labor-intensive and time-consuming nature of conventional fieldwork. Despite decades of research focused on remote estimation of forest biomass, a significant gap persists in our understanding of how employing multi-platform approaches can address the saturation issue prevalent in high-density and high-biomass forests, particularly in tropical and subtropical regions. This study aims to bridge this gap by integrating Global Ecosystem Dynamics Investigation (GEDI), time-series remote sensing data from Advanced Land Observing Satellite-2 (ALOS-2) / Phased-Array L-band Synthetic Aperture Radar-2 (PALSAR-2) and Sentinel-2 datasets. By leveraging data from multiple sources and employing advanced machine learning techniques, our study seeks to provide a more depth understanding of how combining diverse remote sensing platforms can enhance our ability to accurately estimate forest biomass in areas previously susceptible to saturation issues. This holistic approach holds promise for refining our assessments and management strategies in vital forest ecosystems.