Geostatistical Modelling

Variogram Modelling

The first step in variogram modelling is to produce directional experimental variograms. These can be determined from almost any kind of data, for example, diamond core samples, blasthole chip samples, channel samples, etc. It is however advantageous if some thought is devoted to the objectives and purpose of variographic analysis. Having selected a suitable sample support, the experimental variograms are then generated in various directions.

The next step is to select a suitable model variogram that would best characterise the overall shape of our experimental variogram. Having done this, a visual fit is carried out and initial parameters are selected, a theoretical model variogram is calculated and overlaid on the experimental variogram plot. The parameters are then altered accordingly to achieve the best fit.

As the variograms are the final controlling factors in determining the sample weight configuration (Kriging), you might like to use a point kriging plan to determine 'the goodness of fit'. Each sample is removed in turn from the data set, and an estimated value at that point is determined using the model variogram and surrounding data. The estimated values are then compared with actual sample values to establish the suitability of the variographic parameters used in the selected model.