For my research I want to calculate the total area of each WRB soil class on each continent in km2 using the 250m WRB soilgrids raster. However, the total area I calculated is almost 1.3 times as large as the total land surface area on earth.

I downloaded the TAXNWRB_250m_1l (2017 data) raster from here https://files.isric.org/soilgrids/former/2017-03-10/data/ and loaded it in arcgis pro. I used "extract by mask" to create a raster for each continent, using the continents as masks.

I assumed that because the resulution is 250m, I could calculate the total area in km2 by multiplying the number of cells for each soil class with (0.25*0.25). I used the "Count" column in the attribute table for the number of cells per soil class. However, there is a total of 3,275,963,752 cells for all continents together, which gives a total area of 204 million km2.

This is remarkable since the total land surface area is about 148 million km2. The original raster I downloaded has even more cells (3,280,092,749), so that would even more overestimate the surface area.

Area = number of cells* (cell_size*cell_size) is common practice I believe, so this makes me think it has to do with the cell size I use.

Could it be that this has something to do with the projection, so that the 250m resolution only applies at the equator and becomes smaller at higher lattitudes? The raster only has a GCS (WGS84) and I tried projecting it to both Mollweide and Homolosine, which give 217.7 and 224.3 as cell sizes, respectively, when I look into the properties. Using these as cell size the total area still exceeds the theoretical land surface area of 148 million km2.

Any advice on how to correctly calculate these areas would be most welcome!

  • Hi @Emma welcome to GIS.StackExchange. Your question is clear and detailed, but still take some time to read the Tour for all the details on how the web site functions. – Luís de Sousa Jun 9 at 15:05

The 2017 edition of SoilGrids was published with an irregular grid defined on the WGS84 datum. I.e. raster cells are of variable size and shape. When you load such a grid into a programme like QGis or ArcGIS the Marinus of Tyre projection is applied implicitly. Any area (or distance) calculations on this projection are meaningless.

You indeed need to project the map with an equal-area projection, making sure you obtain an approachable cell shape. With gdalwarp it looks something like this for the Homolosine:

gdalwarp -t_srs '+proj=igh +lat_0=0 +lon_0=0 +datum=WGS84 +units=m +no_defs' --config CHECK_WITH_INVERT_PROJ TRUE  -tr 250 250 -of GTiff TAXNWRB_250m.tif  TAXNWRB_250m_152160.tif

Or for the Mollweide:

gdalwarp -t_srs '+proj=moll +lat_0=0 +lon_0=0 +datum=WGS84 +units=m +no_defs' -tr 250 250 -of GTiff TAXNWRB_250m.tif  TAXNWRB_250m_54009.tif

After this it is all about map algebra. GRASS, for instance, provides modules that do it all automatically: r.stats and r.report. I do not know ArcGIS enough, but looking at this ESRI community thread it seems to have similar modules. The key really is to get the map projection right.

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.