Technical specifications SoilGrids1km

Title: SoilGrids1km (www.soilgrids.org)
Period (temporal coverage; approximately): 1950━2005
Spatial resolution (covariates): 1/120 decimal degrees or about 1 km
Spatial resolution predictions and support size: 1 km (predictions at point support but using 1 km resolution covariates)
Total number of gridded maps: 288
Number of pixels with coverage per layer: 183M
Total size before compression: about 350GB
Total size after compression: about 19GB
Data license (IP policy): Attribution-NonCommercial International CC BY-NC
Source of point data: ca 110,000 soil profiles (World Soil Profiles)
Source of covariate data: ca. 80 layers (World Grids)
Target data standard: GlobalSoilMap specification

Software implementation: http://gsif.r-forge.r-project.org/
Frequently Asked Questions: http://isric.org/content/faq-soilgrids

 

Code: ORCDRC
Target variable (units): Soil organic carbon (fine earth fraction) in permilles (g/kg)
Spatial prediction method: 3D regression (GLM with a log-link function) with soil-depth relationship modelled using splines
% of variance explained (cross-validation): 22.9%
Most significant covariates (sorted): TNMMOD3a, TNSMOD3a, TX2MOD3a, TX4MOD3a, TX5MOD3a, EVSMOD3a, G09ESA3a, PX4WCL3a, GACHWS3a, GANHWS3a, GARHWS3a, GCRHWS3a, GHSHWS3a, GLXHWS3a, GLVHWS3a, GPLHWS3a, ns(altitude, df = 2)1, ns(altitude, df = 2)2

 

Code: PHIHOX
Target variable (units): pH in H2O × 10
Spatial prediction method: 3D regression with soil-depth relationship modelled using splines
% of variance explained (cross-validation): 50.5%
Most significant covariates: SLPSRT3a, TWISRE3a, TDSMOD3a, G02ESA3a, G13ESA3a, L12IGB3a, GLTUHA3xCarbonate sedimentary rocks, GLTUHA3xUnconsolidated sediments, GACHWS3a, GARHWS3a, GCLHWS3a, GCHHWS3a, GFRHWS3a, GPLHWS3a, GPZHWS3a, GSNHWS3a, GVRHWS3a, ns(altitude, df = 4)4

 

Code: SNDPPT
Target variable (units): Sand content mass fraction in %
Spatial prediction method: 3D regression (logit-transformed values) with soil-depth relationship modelled using splines
% of variance explained (cross-validation): 23.5%
Most significant covariates: TWISRE3a, TX2MOD3a, EVSMOD3a, G04ESA3a, G09ESA3a, G11ESA3a, G14ESA3a, G15ESA3a, G19ESA3a, L12IGB3a, GLTUHA3xBasic volcanic rocks, GLTUHA3xCarbonate sedimentary rocks, GARHWS3a, GKSHWS3a, GPHHWS3a, GPLHWS3a, GPZHWS3a, GVRHWS3a

 

Code: SLTPPT
Target variable (units): Silt content mass fraction in %
Spatial prediction method: 3D regression (logit-transformed values) with soil-depth relationship modelled using splines
% of variance explained (cross-validation): 34.9%
Most significant covariates: TDSMOD3a, TNSMOD3a, TX2MOD3a, TX5MOD3a, EVSMOD3a, G01ESA3a, L12IGB3a, GLTUHA3xBasic volcanic rocks, PX3WCL3a, GANHWS3a, GARHWS3a, GFRHWS3a, GHSHWS3a, GKSHWS3a, GLVHWS3a, GPHHWS3a, GPLHWS3a, GPZHWS3a

 

Code: CLYPPT
Target variable (units): Clay content mass fraction in %
Spatial prediction method: 3D regression (logit-transformed values) with soil-depth relationship modelled using splines
% of variance explained (cross-validation): 24.4%
Most significant covariates: TX2MOD3a, G02ESA3a, G03ESA3a, G04ESA3a, G05ESA3a, G11ESA3a, G15ESA3a, G19ESA3a, GLTUHA3xBasic volcanic rocks, GLTUHA3xCarbonate sedimentary rocks, GLTUHA3xIntermediate volcanic rocks, GLTUHA3xMixed sedimentary rocks, GLTUHA3xSiliciclastic sedimentary rocks, GANHWS3a, GARHWS3a, GLXHWS3a, GPZHWS3a, GVRHWS3a

 

Code: CFRVOL
Target variable (units): Coarse fragments (> 2 mm fraction) volumetric in %
Spatial prediction method: 3D regression (zero-inflated model) with soil-depth relationship modelled using splines
% of variance explained (cross-validation): -
Most significant covariates: L3POBI3bhigh mountains, L3POBI3blow mountains, L3POBI3brough low hills, L3POBI3bsmooth low hills, TNSMOD3a, TX6MOD3a, EVSMOD3a, G04ESA3a, G14ESA3a, L12IGB3a, GLTUHA3xSiliciclastic sedimentary rocks, GLTUHA3xUnconsolidated sediments, PX2WCL3a, GANHWS3a, GARHWS3a, GPHHWS3a, GVRHWS3a, ns(altitude, df = 4)4

 

Code: CEC
Target variable (units): Cation exchange capacity in cmol/kg
Spatial prediction method: 3D regression (GLM with a log-link function) with soil-depth relationship modelled using splines
% of variance explained (cross-validation): 29.4%
Most significant covariates: L3POBI3bsmooth low hills, TDSMOD3a, TX2MOD3a, TX4MOD3a, LASMOD3a, G11ESA3a, G20ESA3a, GLTUHA3xBasic volcanic rocks, GLTUHA3xCarbonate sedimentary rocks, GLTUHA3xIntermediate volcanic rocks, GARHWS3a, GCHHWS3a, GFRHWS3a, GFLHWS3a, GLXHWS3a, GPHHWS3a, GVRHWS3a, ns(altitude, df = 2)2

 

Code: BLD
Target variable (units): Bulk density of the fine earth fraction in kg / m3
Spatial prediction method: 3D regression with soil-depth relationship modelled using splines
% of variance explained (cross-validation): 31.8%
Most significant covariates: TDSMOD3a, TNMMOD3a, TNSMOD3a, TX2MOD3a, TX4MOD3a, TX5MOD3a, G05ESA3a, PX2WCL3a, PX3WCL3a, PX4WCL3a, GACHWS3a, GANHWS3a, GCHHWS3a, GLVHWS3a, GPLHWS3a, GUMHWS3a, GVRHWS3a, ns(altitude, df = 4)4

 

Code: BDRICM
Target variable (units): Depth to bedrock (R horizon) up to maximum 240 cm
Spatial prediction method: 2D regression using zero-inflated modelling
% of variance explained (cross-validation): -
Most significant covariates: -

 

Code: TAXGWRB
Target variable (units): World Reference Base (2006) groups
Spatial prediction method: Multinomial logistic regression (nnet package)
Kappa statistics (cross-validation): 28.1% (map purity: 32.7%)

 

Code: TAXOUSDA
Target variable (units): Soil Taxonomy (USDA) sub-orders
Spatial prediction method: Multinomial logistic regression (nnet package)
Kappa statistics (cross-validation): 40.3% (map purity: 44.2%)