Confidence intervals on raster produced by kriging spatial interpolation?

Question

I have used spatial interpolation to generate a raster that estimates percentage of a variable of interest (e.g. expected percentage at unsampled locations given frequencies at nearby sites). As I am ultimately interested in identifying regions where this percentage is high, I have reclassified the raster so that values below a cutoff (e.g. 90%) are set to 0.

Initially I used inverse distance weighting spatial interpolation in ArcGIS but I'm wondering whether kriging would allow me to estimate some sort of confidence interval/estimates of error of the location of the boundaries where estimates of my variable drops to less than 90%? Does ArcGIS provide output that can be used for such purposes?

Answer 1

You will find papers in the literature that have used the kriging standard deviation to construct confidence intervals but this is a fundamental mis-understanding of how the kriging standard deviation is computed. In particular it does not depend on the data values at all.

IDW is simple to apply but it has no theoretical basis and hence one will have difficulty interpreting the results