 This study compares the performance of four machine learning regression models for rice dry biomass estimation using Landsat 8, Sentinel-2A, HJ-1A and B, and GF-1 data in a major rice cultivation area in southeast China during the 2016 and 2017 growing seasons. The results indicate that the gradient boosting decision tree, GBDT, model is the most accurate for before rice-heading scenario while K-nearest neighbor, KNN, performs best after heading. Studies recommending extending the evaluation of these models to other parameters and microwave imagery are suggested. This article was offered by Laminar Mansaree, Adam Shikakon, Lingbo Yang, and others.