Scholarly record

ESTIMATION OF BIOMASS USING GEDI LIDAR DATA, HARMONIZED LANDSAT SENTINEL-2 (HLS) IMAGERY, AND GRADIENT BOOSTING MACHINE LEARNING

Zhiyong Hu

Abstract

Accurate estimation of vegetation biomass, particularly aboveground biomass (AGB), is critical for understanding carbon cycling, climate change mitigation, and ecosystem dynamics. Forests alone store a large proportion of terrestrial carbon, making biomass estimation central to initiatives such as REDD+ and global carbon accounting. Traditional field-based (destructive) methods, while accurate, are labor-intensive, costly, and impractical at large spatial scales. Remote sensing technologies have emerged as essential tools for large-scale biomass estimation, with spaceborne LiDAR and multispectral satellite imagery playing complementary roles. In recent years, their integration has become a dominant paradigm for improving both the accuracy and spatial coverage of biomass estimates. Existing studies have used random forest machine learning to model relationship between LiDAR observed point biomass and vegetation spectral indices derived from Sentinel and Landsat data for the estimation. This paper uses biomass data from spaceborne LiDAR system GEDI and vegetation indices derived from the Harmonized Landsat Sentinel-2 (HLS) imagery, and a machine learning method, gradient boosting to estimate ABG for northwest Floria, USA.

Publication details

Back to publication list