The Variogram Cloud - The Need For A New Method For Calculating Variograms

From a computational point of view calculating a variogram is a perfectly simple operation. For any particular study, we know which variable is to be studied. We also have a good idea of structural parameters which are likely to be important (the scale of the phenomenon, the probable direction of the anisotropy, etc.). Finally, we know the structure of the data file. So we can specify the parameters to be used to calculate the variogram (the distance classes, the angular classes, etc.). This allows us to decide which variogram program is most appropriate (1, 2 or 3 dimensional data, regularly or irregularly spaced, etc. ) . In general, the basic statistics (mean, variance, correlation? etc.) have been calculated beforehand, and so the distribution of the data is already known. Consequently, the choice of the parameter values fed into the variogram is by no means arbitrary. But it must nevertheless be made before the' variogram is calculated. If the resulting variogram is not satisfactory, the whole sequence of operations has to be repeated. Before going on to see how this can be avoided, it is worthwhile having a brief look at the common causes of "unsatisfactory" variograms.