Conditioning to secondary data

When working in the Rock Property Modeling workflow, the Conditioning form (model > Rock Properties > Conditioning) only opens automatically if you selected Kriging or SGS as a modeling method. Depending on the selected method, the form offers different functionality. With the Conditioning form you can enforce the modeling algorithm to use another property (called secondary property) to influence the values simulated during the modeling run, e.g. use porosity to model permeability. You should only use conditioning if there is a statistical relationship between the two properties. If you do not want to use secondary data, select the 'No secondary data' option at the top of the form. Otherwise, either select 'Vertical & lateral' or '3D'.

Using the context menu options Copy and Paste, you can copy-paste settings between the VOIs which have methods that use conditioning (i.e. Kriging, SGS). Once you have copied settings, you can select Paste Settings Options. This will open a separate form, which allows you to paste the settings to multiple VOIs at once. Note that checkbox selection can be done in bulk by holding down SHIFT while selecting a range of VOIs (they turn blue), then checking the box of one of them. All other VOIs in the selected range will get checked simultaneously.

No secondary data  Select this option if you do not want to use secondary data. Next, select which type of Kriging to use for your primary data.

Simple Kriging  The search ellipsoid is used to find the data points to be used in the calculation (you set the Search Factor on the Control Method form, next step of the workflow). The global mean is a known value, and is held constant over the area of interpolation. Simple Kriging is the best approach when there is no or little evidence of significant non-stationarity and few hard data. The industry standard Kriging library is used. See Global Mean below.

  • Global mean  (Only visible on the form when no transformation is selected on the Transform & Trends form.) The Simple Kriging approach requires a mean which is held constant across the modeled domain, i.e. the VOI. Note that you can also enter a value outside the range of the upscaled property. Enter the mean or have it calculated from the upscaled property (). If transformations are active in T&T, then the mean is automatically obtained from transformed data.
Uncertainty   This parameter can be propagated as a parametric uncertainty in a probabilistic volumetric calculation with the study strip. Click the uncertainty symbol behind the 'Global mean' entry field to open the Uncertainty Parameter dialog. For how to use the controls on the dialog, see The Uncertainty Parameter dialog. For information about parametric uncertainties and how to use them in JewelSuite Subsurface Modeling, see Incorporating uncertainty in static or dynamic modeling.

Ordinary Kriging  The search ellipsoid is used to find the data points to be used in the calculation (you set the Search Factor on the Control Method form, next step of the workflow). The mean is not a-priori known but based on the data points found in the search ellipsoid. Good approach when there are many hard data and there is some evidence of non-stationarity. The industry standard Kriging library is used.

Use global kriging  If you check the box, by definition the search ellipsoid spans the entire set of input data points (as opposed to standard kriging where only the data points within the search ellipsoid is taken into account). Because of this, the number of calculations is reduced and global kriging is usually faster. Global kriging can be used in combination with both Simple and Ordinary Kriging. (Note: this option is unrelated to the 'Global mean' setting above.)

Vertical & lateral  This selection enables the use of vertical and lateral secondary data.

Vertical  From the drop-down list select which property you want to use: From upscaled property or Target proportions for modeling. You can open a dedicated VTC form where you can assign target proportions for each k-layer in your VOI, see Editing a Vertical Trend Curve.

Lateral  From the drop-down list, select a 2D grid. This will populate the next drop-down list with all the properties that you can use as a secondary property. Select a property.

3D  This section enables the use of 3D secondary data.

Property  From the drop-down list select a property. This property is used as a secondary property, which is suitable only if the scale of this data is similar to that of the data being modeled, and if there is a clear statistical correlation between the modeled data and the secondary property data.

Integration method  Select the type of method you want to use.

Locally Varying Mean  Select this option when the scale of spatial variation of the secondary property is larger than the one of the upscaled property. For this option the secondary property supplies the expected mean and needs to have the same property template as the primary property.

CoKriging  Select this option when the scale of spatial variation of the secondary property is similar to upscaled property, and there is a clear statistical correlation between the two.

Collocated cokriging always assumes a linear relationship between transformed primary and transformed secondary properties at the same location. The scatter chart on the form shows the transformed properties. Only the correlation coefficient is used by the algorithm. The transformed properties are modeled as standard normal variables, so the intercept of the linear relationship is always zero, and the slope is determined by the correlation coefficient.

If one or both of the transforms are non-linear, the resulting relationship between the back-transformed properties will also be non-linear. The default Normal Score transform based on the upscaled data is non-linear, but there is no user control. If you wish to control the resulting relationship you must use one or more of the parametric transforms for both primary and secondary properties. For primary properties, go to the Transform and Trends form, for secondary use the 3D Property Preparation Transform and Trends button on this form. The parametric transforms are the User-defined normal distribution (an option in the Normal Score transform), the Shift Scale transform, and the Logarithmic transform.

  • To produce a linear relationship in Kriging use Shift Scale transforms. For SGS a similar effect can be achieved with User-defined normal transforms.
  • To produce exponential or logarithmic relationships add a Logarithmic transform to the primary or secondary data respectively.
  • To produce a power-law relationship add Logarithmic transforms to both data.

3D property preparation  If your primary property is transformed, then you must transform your secondary property as well. Click on the Transform and Trends button to open the dedicated 3D Property Preparation form. For more on data transformations, see transformations.

Correlation coefficient  Correlation coefficient between the property to be modeled and the secondary property, applied when you select a secondary property. This option is available when you have selected CoKriging. You can enter a value or use the calculator button () to provide one for you. The data used is displayed in the diagram below. You can control how to display the data in the diagram with the following options:

  • Changing format and units - A right mouse button click on the horizontal or vertical axis opens a context menu. You can change the units and the format by making a selection from the options provided to you.
  • Zooming - In the plot area, left mouse button click and keep pressed down to activate zooming. You can move your mouse in any direction. A preview window indicates the new display window. When you let go of the left mouse button, this is the part of the chart that is shown. To zoom out, double click in the chart area.
  • Adjusting values along the axis - Hover over the chart with your mouse until the tool tip changes into a bidirectional arrow. Left mouse button click and keep pressed down while moving the mouse will adjust the values of the axis.
  • Show/hide legend - A right mouse button click in the view opens a context menu with the Show/hide legend option. When the legend is shown, you can also use it as a display filter: when you click on a name, that data is hidden from the chart, and the legend is grayed out.
Uncertainty   This parameter can be propagated as a parametric uncertainty in a probabilistic volumetric calculation with the study strip. Click the uncertainty symbol behind the Correlation Coefficient entry field to open the Uncertainty Parameter dialog. For how to use the controls on the dialog, see The Uncertainty Parameter dialog. For information about parametric uncertainties and how to use them in JewelSuite Subsurface Modeling, see Incorporating uncertainty in static or dynamic modeling.

Click Apply to save and apply the settings you have specified and keep the form open, or click OK to save and apply the settings you have specified and move to the next step of the workflow, Control Method.

For rules of thumb to capture trends, see Capturing trends with JewelSuite Subsurface Modeling - rules of thumb.