Model Validation

As seen before, the semivariogram analysis includes the experimental semivariogram survey and later the theoretical modeling family adjust. In all this sequence, there is always an uncertainty level about the parameters adjusted to the models. This uncertainty is the estimate error, which can be obtained through the procedure called model validation. Summarizing, the validation process involves the re-estimate of the known values through the adjusting parameters of the semivariogram model.

Before executing the kriging, it is recommended to verify the validation results. Obvious problems can be identified with the input parameters (for instance, the semivariogram specification) or with data (for instance, outliers).

The validation module developed in the Spring uses the "kt3d" subroutine of the GSLIB (Deutsch and Journel, 1992) and gives the following outputs:

  • Error Spatial Diagram;
  • Error Histogram;
  • Error Statistics;
  • Observed Estimated Diagram values;
  • Numerical Results.


Executing the Model Adjust Validation:

  • In the Spring main menu press on Analysis -> Geostatistics -> Model Validation...;
  • the "Model Validation" window is presented;
  • observe that the active IL is presented in the superior part of the window. It has to be an IL using the adequate numerical model. If the IL is not of the numerical model and has no sample representation, a warning message will be displayed;
  • before executing the validation it is desirable to check the defined model. For this press the Model Verification... button;
  • The Minimum and Maximum Interpolation Parameters are related to the Number of Points in the Searching Ellipsoid. They are filled with the default values (4 and 16 respectively). Next, the Searching Ellipsoid radius and orientation are defined. The R.min, R.max, and angle fields are initialized, for an isotropic case with the following default values: R.min and R.max in meters corresponds to the diagonal of the Project involving rectangle and the angle has any value, for instance zero. Evidently if anisotropy is present, these parameters must be adjusted and selected according to the (Deutsch and Journel, 1992);
  • Press on the Apply button. The available results for visualization and analysis are in the Results list options (See below).


 Results

For example, press the Error Spatial Diagram option. The graphical window presented in the Figure is opened:


Configure your Chart

  • The cross type symbols in the figure above indicates the geographical location of the samples and the error magnitude (the smaller the symbol the smaller the error). More specific Information are obtained as follows: point, using the mouse, the location desired and next press on the mouse left button. The result is presented in the windows footnote area, as presented in the figure above.
  • Similarly, press on the Error Statistics and Error Histograms option and the windows presented in the Figures below will be presented.

 
Configure your Chart

  • To visualize the Observed x Estimated diagram values, press on the Observed x Estimated Diagram option. This action opens the window presented in the Figure below.


Configure your Chart

  • Besides the graphical results, the validation module gives a numerical results window, which is visualized selecting the Numerical option. This option opens the window presented in the figure below.

See also:
SPRING - Spatial Analysis
Thematic Models for experimental variogram adjusts
How to Execute Semivariogram Modeling ?
How to Execute Kriging ?