Global Grassland Biomass Allocation Dataset – Created by Wu et al., 2025

Global Grassland Biomass Allocation Dataset (AGB, BGB, RS)This dataset provides global gridded estimates of aboveground biomass (AGB), belowground biomass (BGB), and root-to-shoot ratio (RS) in grassland ecosystems at a spatial resolution of 0.05°.
Biomass data were compiled from 2,406 field observations across 545 sites worldwide, collected from open-access repositories (e.g., ORNL DAAC) and peer-reviewed literature (1980–2024). Climatic, edaphic, topographic, and vegetation trait variables were extracted from widely used global datasets, including WorldClim, HWSD, MODIS/GLASS, and global leaf trait databases, yielding a total of 36 explanatory variables.
We applied an Extreme Gradient Boosting (XGBoost) machine learning framework to integrate these multi-source drivers and predict global grassland biomass allocation. Model training and validation employed ten-fold cross-validation and multiple accuracy metrics (MSE, RMSE, MAE, R²), ensuring robust predictive performance.
The final product includes:
Global gridded maps of AGB, BGB, and RS at 0.05° resolutionSpatially explicit patterns of biomass allocation across diverse grassland biomesA benchmark dataset for ecosystem modeling, carbon cycling studies, biodiversity conservation, and climate change assessmentsCoverage: Global grasslands
Period: Contemporary baseline (1980–2024 integrated)
Format: NetCDF (0.05° resolution grids)
Applications: Carbon cycle modeling, Earth system model evaluation, grassland management, and climate change mitigation research