Diane
A semi-variogram assumes that 'relationship' between two values is a function of distance between their locations and possibly direction. This is the intrinsic hypothesis.
A trend or drift is a change in the expected value from location to location within your study area. It is not important unless it interferes with the semi-variogram. You see it as a parabolic upturn added to the 'real' semi-variogram. You need to remove it before you can see the distance relationship, whether you are just using it for interpretation or proceeding to estimation or simulation.
If you are getting a clear shape for your semi-variogram with a sill, you not only have data which satisfies the intrinsic hypothesis but probably the stronger general stationarity too.
You cannot MAKE your data stationary - you can possibly subdivide a non-stationary area into more
homogeneous units.
Hope this helps
Isobel
