1

I am looking for a function to do the following.

Given a boundary file, and soil sample points within the boundary I am wanting to use the Leave one out validation technique.

What I mean is create a Kriged map leaving out n point, and then compare the kriged value with the known value.

The purpose of this is to identify areas within the boundary that need resampling if the kriged value is significantly different from the actual value.

Does anyone know of a method within QGIS or custom python script to do this?

Improve this question
2
  • 1
    You may be doing more that identifying areas that need dual-sampling. As large deviation of a local value in a Jackknife validation could also indicate a nonstationarity problem which, would be a serious violation of, even relaxed, kriging assumptions. I do not know about this in Python but in in R this would be a simple random sampling and re-estimation for loop. Let me know if you would like an R example.
    – Jeffrey Evans
    Commented Feb 18, 2016 at 18:47
  • Actually, here is an R example using thin plate splines. The basic idea would be the same with a kriging estimate rather than a spline. gis.stackexchange.com/questions/82550/…
    – Jeffrey Evans
    Commented Feb 18, 2016 at 18:49

1 Answer 1

Reset to default
3

In R, here is an example:

library(sp)
library(gstat)
demo(meuse, ask = FALSE)
summary(krige.cv(log(zinc)~1, meuse, vgm(1, "Sph", 900)))

Read the docs in ?krige.cv

Improve this answer

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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