Machine learning models for estimating above ground biomass of fast growing trees

Abstract

Biomass is a renewable and sustainable energy resource that can potentially be substituted for fossil fuels, which have a negative impact on the environment including the production of greenhouse gas (GHG) emissions. Forest carbon stocks are also of growing interest with regard to both GHG sequestration and renewable energy supply; fast-growing trees are of particular interest in this area. Producing a highly accurate estimation of the above-ground biomass (AGB) of any forest plantation is challenging. In this study, we apply machine learning (ML) techniques to model the AGB of fast-growing trees, namely E. camaldulensis, A. hybrid, and L. leucocephala. It is found that the random forest algorithm has the highest prediction accuracy (R2 of over 0.95, and normalized root mean square error of about 0.20), when compared to other ML algorithms and traditional allometric equations for estimating AGB. This work offers an alternative of estimating AGB for the tropical fast growing trees through the synergy of simple tree characteristics and modeling algorithms.