Abstract: Aboveground biomass (AGB) is a key indicator of grassland ecosystem function and quality, and is crucial for effective management. We combined a long-term time series of remote-sensing imagery with ground-survey data to develop a random-forest inversion model for grassland AGB in the Xilin Gol League. The best-performing model was applied to estimate annual AGB from 2000 to 2023, and the Mann-Kendall test and Sen's slope were used to evaluate spatiotemporal trends over this period. The result show that: (1) The model achieved high accuracy, yielding a coefficient of determination (R²) of 0.75, indicating robust predictive skill for regional-scale AGB estimation. (2) Estimated AGB exhibited pronounced spatial heterogeneity which gradually increased from west to east, that is in consistent with the distribution of grassland types. (3) From 2000 to 2023, interannual variation in AGB was small, indicating a relatively stable grassland ecosystem at the regional scale. The multi-decadal record provides a quantitative baseline for evaluating restoration outcomes and grazing management effectiveness. These findings show that random-forest-based remote sensing supports accurate, large-area AGB monitoring and provides practical evidence for regional management.