Remote sensing inversion of grassland aboveground biomass based on high accuracy surface modeling

•Three machine learning algorithm were used to simulate grassland Biomass.•High-Accuracy-Surface-Modeling (HASM) achieved higher accuracy and deeply explicit the spatial heterogeneous of biomass.•Multi source data fusion can improve the accuracy of ecological monitoring.•Warm-wet climate promote grass growth, human intervention lead to biomass decrease. Accurate and effective accounting of grassland aboveground biomass (AGB) is essential for grassland carbon storage accounting and pastoral agriculture sustainability. In this study, we combined AGB field survey data and remote sensing data to build a suitable model to estimate the grassland AGB in the Three-River Source Region (TRSR) of China. Three machine learning models were used to simulate the grassland AGB from 2001 to 2019, including support vector machine (SVM), random forest (RF), and high accuracy surface modeling (HASM). The results show that (1) the HASM achieved better results than the RF and SVM models (R2 = 0.8459 > 0.72 > 0.5858; RMSE = 29