 This study aims to improve the accuracy of estimating above-ground biomass, AGB, in forests using a stacking framework with diverse algorithms. Five base learners were selected and combined using RIDGE or RF as meta-learners, resulting in improved AGB estimates with reduced bias. The stacking models outperformed the optimal base learner in bias reduction, but not in R2 and RMSC. The study also generated global forest AGB maps using the optimal stacking model. This article was authored by Yu Zhenzong, Zhen Ma, Shan Linliang, and others.