SoilGrids
ISRIC - World Soil Information
Machine-learned maps of soil chemistry for the whole planet at 250 m, six depths down.
SoilGrids predicts soil properties worldwide on a roughly 250 m grid using machine learning trained on soil profile observations. It maps pH, organic carbon, bulk density, sand, silt, and clay, cation exchange capacity, nitrogen, and more at six standard depths, each with an uncertainty estimate. It is the default global soil layer for land, crop, and carbon modeling.
Access
- Cost
- FreeFree maps and downloads; the REST API is paused with no restoration date.
- License
- Open licenseCC BY 4.0
- Sign-up
- No sign-up
- How you get it
- Bulk downloadGeoTIFF, VRT
What's in it
- One row is
- One grid cell (about 250 m) per property per depth interval, with predicted value and uncertainty
- Covers
- Global land surface
- Years
- Static modeled snapshot built from soil profiles collected over decades
- Updated
- Irregular
- Size
- 14 properties at six depths on a 250 m global grid
Quality
Widely used and openly licensed with per-pixel uncertainty, but every value is a model prediction, not a measurement.
- Predictions are weak where few soil profiles exist
- REST API is paused
- Not a time series, so it cannot show soil change
Connects to
- Macrostrat Geologic Map and Column Database: Compare predicted soils with the bedrock they formed on
- GEBCO Global Bathymetry Grid: Its land elevation supplies terrain context for soil predictions
Questions it could answer
- Where are the world's most acidic cropland soils?
- How much organic carbon sits in the top 30 cm of soil in each country?