Kriging/Collocation Parameters

Select the Kriging/Collocation gridding method from the Gridding window and then click on <More Options> to open the Options for Kriging Gridding dialog box. View image

The upper left-hand corner of the window contains covariance model, half-correlation length, and minimum variance of measurement parameters, which are specific to the Kriging algorithm.

See: Gridding Methods - More Options to learn about the other options found in this window.

Empirical Covariance – A measure of validity for data projection. Describes how data values vary as a function of distance. One half of the computed covariance is used as a value for setting a margin for including data outside of your grid in grid node calculations.
Data Search Option – Search for data points based on proximity to the grid node or to use a directional bias. (Not an option for minimum curvature.)

Covariance Model

Click on the down arrow next to Variogram Mathematic Model in the upper left-hand corner of the More Options window to select one of seven models from the drop-down menu:

Second-order Markov Model
Gaussian Model
Exponential Model
Quadric Model
Linear Model
Power Model
Wave Model

To determine which covariance model to use, calculate a covariance model from your data by clicking on the <Compute Empirical Variogram> button in the More Options window. It will calculate and insert a recommended half-correlation length into the parameter box.

Choose the model that most closely matches the covariance computed from your data. Select each model from the drop-down menu and keep the recommended half-correlation length to compare it with your computed covariance model.

Half-Correlation Length

The further the half-correlation point lies to the right, the higher the data correlation to distance is and the better your grid interpolation results will be. Short half-correlation lengths produce steep weighting curves, providing maximum weight to the closest input data points. Too steep a curve, however, can produce unusable results.

For high recommended half-correlation lengths, try gridding the data using the recommended value. If you would like to try a stronger weighting setting, experiment with increasingly shorter half-correlation lengths to get the results you require. For very low recommended half-correlation lengths, try gridding the data using the recommended value. The recommended value may be too low to produce coherent results. You will need to increase half-correlation length, add a minimum variance of measurements factor, or use a combination of both approaches.

Minimum Variance of Measurements

Use this parameter to minimize measurement errors in the original input data. Enter a non-zero value to allow the grid values to deviate from input data values that lie nearby. Try a setting of 4 to compute interpolated grid values to replace anomalous input data values. Increase or decrease the setting as needed.