I have a soil
parameter that exhibits clear spatial autocorrelation with a nice semivariogram.
I have a secondary
parameter which exhibits no spatial autocorrelation and consequently I can
not model any semivariogram parameters. The reason is most probably too few
exists an accepatble correlation between these two parameters (corr. coeff.=0.7,
understand it, I have to define the semivariogram parameters (sill,
nugget and range) for both parameters in order to use co-kriging for
interpolation of the second parameter. This is not possible in my case. Are
there any good methods to deal with this situation?
Is it teoretically
1. Estimate my
second parameter (using linear regression) at each point where the first
parameter has been measured but not the second
2. Use kriging on
estimated and measured values to interpolate the secondary
I understand that
this approach is not optimal (Gerard and Pebesma 1999, Geoderma 89, 47-65) but
is it wrong?