A Diagnostic to Assess the Fit of a Variogram Model to Spatial Data

Abstract

The fit of a variogram model to spatially-distributed data is often difficult to assess. A graphical diagnostic written in S-plus is introduced that allows the user to determine both the general quality of the fit of a variogram model, and to find specific pairs of locations that do not have measurements that are consonant with the fitted variogram. It can help identify nonstationarity, outliers, and poor variogram fit in general. Simulated data sets and a set of soil nitrogen concentration data are examined using this graphical diagnostic.