Overview
Probability of occurrence of each of 30 World Reference Base (WRB 2006) reference soil groups, and the most probable group per cell, for the globe on a 250 m grid. These are the soil classification maps of SoilGrids 2.0, but they are not new predictions: ISRIC summed the qualifier-level class probabilities predicted in the SoilGrids250m 2017 release (multinomial neural network and random forest ensemble on about 150,000 WoSIS profiles) into one map per reference soil group, and took the highest as the most probable group. Anthrosols, Technosols and Retisols are not included. Each map is a GDAL virtual raster (VRT) that mosaics a folder of GeoTIFF tiles in place, so it opens remotely with GDAL-based tools.
Important note
The underlying model is the 2017 one, not the quantile regression forests of the SoilGrids 2.0 property maps, and ISRIC says a new unified soil classification method is under development. The files are in EPSG:4326 although the ISRIC catalogue record states Homolosine. The VRTs reference their tiles by relative path, so open them over HTTPS with GDAL's /vsicurl rather than downloading the .vrt alone, and read windows. Probabilities of the 30 groups in a cell do not always sum to exactly 100 because of rounding.
Intended uses
- soil type inputs to crop, land evaluation, and ecosystem models at national to global scale
- stratifying or masking analyses by dominant soil group
Not recommended for
Why: values are 250 m predictions from the 2017 model with modest class accuracy, not survey observations.
Use instead: national soil survey maps or local profile descriptions.
Access & use
Files, endpoints, and worked examples for each distribution.
Data
4096x4096 GeoTIFF tiles, one folder per reference soil group at {rsg}/.
Format: image/tiff
File indexes
GDAL VRThttps://files.isric.org/soilgrids/latest/data/wrb/{rsg}.vrt30 files
One VRT per reference soil group beside its tile folder, with a .vrt.ovr overview file
List of files 30 filesDownload CSV
Individual file
https://files.isric.org/soilgrids/latest/data/wrb/Acrisols.vrt
Usage examples
import rioxarray
# The VRT opens as one raster; its tiles are read on demand
da = rioxarray.open_rasterio("https://files.isric.org/soilgrids/latest/data/wrb/Acrisols.vrt", masked=True)library(terra)
# The VRT opens as one raster; its tiles are read on demand
r <- rast("/vsicurl/https://files.isric.org/soilgrids/latest/data/wrb/Acrisols.vrt")# The VRT opens as one raster; its tiles are read on demand
gdalinfo /vsicurl/https://files.isric.org/soilgrids/latest/data/wrb/Acrisols.vrtOther access methods
4096x4096 GeoTIFF tiles of the most probable group.
Format: image/tiff
File indexes
GDAL VRThttps://files.isric.org/soilgrids/latest/data/wrb/MostProbable.vrt
VRT mosaicking the MostProbable tiles, with a .vrt.ovr overview file
Usage examples
import rioxarray
# The VRT opens as one raster; its tiles are read on demand
da = rioxarray.open_rasterio("https://files.isric.org/soilgrids/latest/data/wrb/MostProbable.vrt", masked=True)library(terra)
# The VRT opens as one raster; its tiles are read on demand
r <- rast("/vsicurl/https://files.isric.org/soilgrids/latest/data/wrb/MostProbable.vrt")# The VRT opens as one raster; its tiles are read on demand
gdalinfo /vsicurl/https://files.isric.org/soilgrids/latest/data/wrb/MostProbable.vrtOther access methods
Additional assets
Pixel value to reference soil group name for MostProbable.
Format: application/json
SLD colour style for the MostProbable map.
Format: application/vnd.ogc.sld+xml
Global rendering of the MostProbable map.
Format: image/png
ISRIC WMS serving the WRB maps as rendered images.
Access the data
https://maps.isric.org/mapserv?map=/map/wrb.map
Data dictionary
What each file holds. A layout describes how the data is stored in a file: its variables, the dimensions they vary over, and any keys.
File layouts
| Files | Format | Variables | Dimensions | No data |
|---|---|---|---|---|
| probability | image/tiff | probability | cell × rsg | 255 |
| most-probable | image/tiff | MostProbable | cell | 255 |
In every file
Dimensions
7.5 arc-second cells (~250 m at the equator).
Grid spacing: 0.0020833333 degree
Only in probability
Variables
| Variable | Type | Unit | Description |
|---|---|---|---|
| probability | uint8 | % | Predicted probability that the cell's soil belongs to the reference soil group, in percent. |
Dimensions
WRB reference soil group, the {rsg} token of the file names (WRB 2006 names, so Albeluvisols rather than Retisols).
Type: string
Coded against: https://www.fao.org/soils-portal/data-hub/soil-classification/world-reference-base/en/
AcrisolsAlbeluvisolsAlisolsAndosolsArenosolsCalcisolsCambisolsChernozemsCryosolsDurisolsFerralsolsFluvisolsGleysolsGypsisolsHistosolsKastanozemsLeptosolsLixisolsLuvisolsNitisolsPhaeozemsPlanosolsPlinthosolsPodzolsRegosolsSolonchaksSolonetzStagnosolsUmbrisolsVertisols
Show all 30 valuesShow fewer
Only in most-probable
Variables
| Variable | Type | Description |
|---|---|---|
| MostProbable | uint8 | Reference soil group with the highest predicted probability in the cell.
|
Technical metadata
The full spatial, temporal, and modeling detail behind the summaries elsewhere on this page.
Spatial
- CRS
- EPSG:4326
- Extent (W S E N)
- [-180, -56, 180, 84]
- Coverage
- World
- Grid spacing
- 7.5 arc-seconds (~230 m at the equator)
Temporal
- Reference date
- 2020
Provenance
Ordered processing steps, from source to publication.
- source2017-03-10
SoilGrids250m 2017: ensemble of multinomial log-linear neural network (nnet) and random forest (ranger) models fitted on WoSIS soil profiles with WRB classes and global covariates, predicting per-class probabilities (Hengl et al. 2017).
- aggregate2020-05-04
The 2017 qualifier-level probabilities summed into one map per reference soil group, and the group with the highest probability written as MostProbable. Published as part of SoilGrids 2.0.
- regenerate2022-03-01
Layers regenerated with ISRIC's WRB mapping scripts to fix a distortion at low latitudes, and tiled as VRT mosaics.
How to cite
Please cite the dataset and any associated peer-reviewed publications.
L. Poggio, L. M. de Sousa (2020-05-04) SoilGrids250m 2.0 - WRB classes and probabilities. Version 2.0. ISRIC - World Soil Information. https://doi.org/10.17027/isric-soilgrids.c4dc161c-d62d-11ea-a1a3-292680b15169. Accessed through the CGIAR Climate Action Data Hub, https://cgiar-climate-data-hub.github.io/catalog/soilgrids-wrb_2.0/.
BibTeX
@misc{soilgrids_wrb_2020,
title = {SoilGrids250m 2.0 - WRB classes and probabilities},
author = {Poggio, L. and de Sousa, L. M.},
year = {2020},
version = {2.0},
publisher = {ISRIC - World Soil Information},
doi = {10.17027/isric-soilgrids.c4dc161c-d62d-11ea-a1a3-292680b15169},
note = {Accessed through the CGIAR Climate Action Data Hub, https://cgiar-climate-data-hub.github.io/catalog/soilgrids-wrb_2.0/}
}Contacts
Who produced this dataset, and who to reach about it.