High-Resolution Global Soil Organic Carbon Stocks and Uncertainty Maps (30 cm & 100 cm, 100 m Resolution)

     Accurate quantification and monitoring of soil organic carbon (SOC) stocks and their changes are essential for closing the global carbon budget, assessing the land carbon sink, and supporting carbon accounting in nature-based climate solutions. However, uncertainty in SOC estimates remains a major challenge. To address this, we present globally consistent, spatially explicit SOC maps at 100 m spatial resolution, developed to align with the scale of land-use dynamics and improve the precision of carbon accounting across biomes and land-use categories.

     To develop the maps, we harmonized over 100,000 depth-specific SOC observations (0-30 cm and 0-100 cm) and integrated them with high-resolution remote sensing and in situ spatial covariates. We applied 14 biome- and ecosystem-specific random forest (RF) models to capture the distinct biophysical and climatic controls on SOC within each biome and ecosystem. Peatland- and mangrove-specific models, trained on recent field data, were applied to better represent the unique biophysical characteristics and large carbon stocks of these ecosystems. SOC predictions are reported in tons of carbon per hectare (t C/ha) at each depth. To quantify model uncertainty, pixel-level standard deviation maps were produced using a bootstrapped ensemble of 20 Random Forest models. From these, the 95% confidence interval (95CI) half-widths were calculated by multiplying the standard deviations by 1.96. All files are provided as GeoTIFFs in EPSG:4326, with 100 m resolution, and are encoded as 16-bit integers (UInt16).