Variogram Model Selection

Abstract

A common problem in geostatistics is variogram estimation, in order to choose an acceptable model for kriging. Nevertheless, there is no standard method, first, to test if a particular model can be accepted as valid and, second, to choose among several competing variogram models. The problem is even more complex if, in addition, there are outliers in the data. In this paper we propose to use the distribution of some classical and robust variogram estimators to test, first, the validity of a particular model, accepting it if the p-value of the test, with this particular model as null hypothesis, is large enough and, second, to compare several competing models, choosing the model with the largest p-value among several acceptable models.