[R-sig-Geo] Global gridded soil data (https://SoilGrids.org)

[Tomislav Hengl]

For those of you that need soil data for your modeling / analytics,

This is to inform you that we have recently done a major update of our 
SoilGrids system (https://soilgrids.org). The new predictions are now 
available at 250 m globally for 7 standard depths (0, 5, 15, 30, 60, 
100, 200 cm). We distribute predictions of standard chemical (soil pH, 
organic carbon, CEC) and physical soil properties (texture fractions, 
bulk density, coarse fragments, depth to bedrock) but also predictions 
of soil classes (USDA and WRB classification systems). These data are 
distributed under the Open Data Base License (i.e. the same license used 
by OpenStreetMap).

You can download the data directly via FTP 
(ftp://ftp.soilgrids.org/data/recent/) or by using the Web Coverage 
Service (http://webservices.isric.org/geoserver/wcs). I wrote a short 
tutorial that explains how to grab blocks of data using GDAL WCS driver 
(http://gsif.isric.org/doku.php?id=wiki:tutorial_soilgrids#wcs_data_access). 
Let me know if you are aware of any 'easier' way to subset and resample 
SoilGrids via WCS.

SoilGrids are also available via REST API (http://rest.soilgrids.org) 
hence at point locations you can fetch majority of values by using GSIF 
package (http://gsif.r-forge.r-project.org/REST.SoilGrids.html). Please 
try not to use this function to fetch values for large number of points 
as this can become very time consuming (the average response time per 
point is about 0.6 sec).

I would also like to mention that this project was fully implemented in 
R / OSGeo software (which on the end worked out very smoothly even 
though we had to crunch terrabytes of remote sensing data). We are 
really grateful to all creators of packages we have used, especially to 
the authors of the ranger, xgboost, snowfall, caret, raster and rgdal 
packages and SAGA GIS and GDAL, which are the backbone of the spatial 
prediction system. I could spend a lifetime thanking the package authors 
for sharing their talent and creations with us.

PS: We have a separate mailing list for SoilGrids 
(https://groups.google.com/forum/#!forum/global-soil-information) mainly 
used by soil scientists / soil data experts, but if it is a generic 
spatial analysis problem, then I will do my best to answer it via 
R-sig-geo.

cheers,

T. Hengl