[ai-geostats] positiveness for variogram model
- Dear all:
We all use the commonly-known variogram models(such as gauss, exp and
spherical,etc.) to do variogram modelling and kriging since we know
these models will give you a positive krige variance.
I have a question if I extend the model, say, gauss model to
gamma(h)= c0+c1*(1-exp(-(x/r)^alpha)), then what requirements for
alpha in order for this model to be a valid variogram model? I have
tried a few alpha for one of my data, this model is not valid if
alpha is greater than 5.
To be more general, for any function f(x, a1,a2,a3,....), how do we
prove f(x) is a valid variogram function for parameters a1,a2,a3...?
Any suggestion will be appreciated!
P.S. I will do a summary for all replies