Values from the global soil-erodibility (K) datasets on the NatCap Data Hub seem low

Hello everyone,

I recently encountered the global soil erodibility layers produced by NatCap and was really pleased to see these since it saves the effort of having to calculate K-factor oneself.

However, when inspecting the Williams K-Factor data from here Global Soil Erodibility - Williams K-Factor - Dataset - NatCap Data Hub, I am finding that the values seem very low, almost as if they are off by a factor of 10.

For example, I downloaded the William’s K-Factor data for a large swathe of sub-Saharan Africa and find that the mean value across this region is around 0.0035. When the conversion factor to US customary units is applied, this comes to 0.026. This is lower than all the values for different soil texture classes in the OMAFRA fact sheet table shown in the user guide, except for ‘Sand’ which has a very low K-factor value. However, sand is not the dominant soil texture across this region.

Furthering my suspicions is the fact that the maximum K factor value across this whole continental region is 0.013 or around 0.1 in US customary units, which is still lower than most of the values for different soil texture classes from the OMAFRA fact sheet. Considering the high prevalence of loamy and clay soils in some areas, I would expect values to be well above 0.1 (in US customary units) across much of the region (based on the OMAFRA factsheet), so am surprised to see that the absolute maximum value across the region is so low.

Based on this, it seems that either the OMAFRA K factor estimates for different soil textural classes are too high (even after using the 0.1317 conversion factor as I have described above) or the K-factor values from the NatCap hosted layer are coming out lower than expected. Could someone with some more knowledge of expected K-factor values perhaps take a look to see whether the values from the Williams layer do indeed seem too low?

Many thanks for any help,
Luke

Hi @lukezw, thanks for bringing this up! I am not a K-factor expert, but I wanted to let you know that I’m asking around our science staff so we can get to the bottom of this. Hope to have more soon.

James

1 Like

Thanks James, look forward to hearing further!

Hi @lukezw, just wanted to post back here with an update. In short, after reviewing the peer-reviewed equations for both the Renard and Williams K Factor layers, there are several obvious problems with how these layers were calculated relative to the source material that, when corrected, produce much more reasonable numerical values in both K Factor layers that we have been hosting.

It has also turned out that there are some mathematical discrepancies in how these equations are presented in papers. One coefficient in particular in the Williams equation remains unclear, and so I am chasing down a final couple of references to confirm what the equation should be. We expect to have corrected layers up on the data hub soon, likely in the next week or two if all goes well. The source material for both of these equations have mostly been old technical reports and model documentation, so chasing everything down has not been particularly easy.

Thanks again for bringing this to our attention, and I hope to have another update soon!
James