Well, based on my disucussion with prof Bogaert, there are several reasons why people tends to use variogram instead of covariogram. But before we discuss it, let us recall that variance (h) is equal to covar (0) – covar (h). So basically, the variogram is just the horizontal-twisted shape of covariogram. This equation is true ONLY in the case od second order stationary. While in non linear case this equation is not necessarily true. In spatial analysis though, the second order stationary is a priori assumed.
Using variogram instead of covario can be regarded as a matter of “taste”. In time series case, people tends to use covariogram and may not even familiar with variogram, WHILE in spatial case, people tends to use the variogram. The second order stationary in spatial analysis makes the variogram easy to use, derive calculation, and intituively easy to understand.
Well if that so, why we still need covariogram? we still need covariogram to check the consistency of our dataset. As shown from the equation, if we have stable , consistence dataset ( at least until second order moment), the covariogram and variogram will be horizontaly-opposed one to each other.
