Machine Learning and Multi-source Remote Sensing in Forest Aboveground Biomass Estimation: A Review

TL;DR

This paper systematically reviews how machine learning and multi-source remote sensing are combined to estimate forest AGB, synthesizing 25 studies from 2014–2024. It finds Random Forest as the most common base method, while Extreme Gradient Boosting often yields the best results when compared against other methods, with Sentinel-1 as the predominant data source and strong gains from multi-sensor integrations including LiDAR. Key insights highlight the importance of feature selection, the value of LiDAR and DEM for 3D structure, and the benefits of phenology and time-series modeling for time-consistent biomass estimates. The work offers practical guidance on sensor combinations, data fusion strategies, and pipeline design, aiming to improve accuracy and applicability of forest carbon estimations across diverse ecosystems.

Abstract

Quantifying forest aboveground biomass (AGB) is crucial for informing decisions and policies that will protect the planet. Machine learning (ML) and remote sensing (RS) techniques have been used to do this task more effectively, yet there lacks a systematic review on the most recent working combinations of ML methods and multiple RS sources, especially with the consideration of the forests' ecological characteristics. This study systematically analyzed 25 papers that met strict inclusion criteria from over 80 related studies, identifying all ML methods and combinations of RS data used. Random Forest had the most frequent appearance (88\% of studies), while Extreme Gradient Boosting showed superior performance in 75\% of the studies in which it was compared with other methods. Sentinel-1 emerged as the most utilized remote sensing source, with multi-sensor approaches (e.g., Sentinel-1, Sentinel-2, and LiDAR) proving especially effective. Our findings provide grounds for recommending which sensing sources, variables, and methods to consider using when integrating ML and RS for forest AGB estimation.

Figures (7)

• Figure 1: Frequency of ML methods by Groups
• Figure 2: Heatmap: binary co-occurrence of two data sources
• Figure 3: Frequency of ML methods used
• Figure 4: Best ML methods found in studies that compared multiple methods
• Figure 5: Most common combinations of data sources