Biomass is a crucial biophysical character of a forest ecosystem, which varies across global environments. The estimation of biomass within forest ecosystems has evolved from conventional destructive methods of tree-harvesting to the current state-of-the-art, non-destructive approaches enabled by remote sensing (RS) and machine learning (ML)-based technologies. The traditional methods primarily use volumetric equations to measure tree biomass for production forestry. But these destructive methods of biomass estimation needed to be transitioned despite providing nearly accurate values, as it involves harvesting of trees. The non-destructive methods are proved to provide efficient and effective biomass estimates across vast areas with significant accuracy while causing no adverse effects on the environment. The scope of these approaches currently ranges from simple linear regression to cutting edge algorithms like Artificial Neural Networks (ANNs) and integrated satellite imagery to model and predict biomass over time. Several recent researches have successfully used ML approaches based on multiple types such as, Tree-based methods, Neural networks, Ensemble learning, Bayesian and Hybrid methods in carbon quantification and valuation. The current chapter explores the existing approaches of tree-biomass estimation with a historical context – how forestry science advanced from destructive methods to the deployment of different ML-based methods for sustainable management of forest resource globally.