Abstract

Geostatistics is an efficient and effective method to continuously assess the content, the spatio-temporal distribution and the correlation of a discretely sampled deposit. It begins with an exploratory analysis that evaluates the consistency and distribution of data through histograms and QQ plots, and then a structural analysis that evaluates data correlation and dependency through variogram and finally a predictive analysis using kriging. This predicting method is used in various geographical investigations: meteorology, demography, hydrology, orography, economy, and pollution, etc. Even when using related software, it is generally of the duty of the user to manually select the suitable variogram model. The main objectives of this paper were to highlight how the choice of a variogram model can affect the results of an interpolating predictive analysis and to show how a best-fitted model can be selected. The results, illustrated with an example, show that the choice of the variogram model inevitably influences the results of a kriging at both endpoints and amplitude of the range of the estimated values. However, the direction of variation of the interpolated values is independent of the variogram model: different variogram models (with the same characteristics) produce different thematic maps but, the areas of minimum and maximum values remain unchanged. Fortunately, the computation of some cross validation tests such as mean error (ME), mean square error (MSE), root mean square error (RMSE), average standard error (ASE) and root mean square standardized error (RMSSE) can help to ascertain the performance of the developed models.

Keywords: Kriging; Predictive analysis; Spatial analysis; Structural analysis; Variogram