Create a Prior Model

Priors are an important part of the Bayesian prediction process. By default, Insight will use "equal priors", where the prior probabilities of the lithofluids are equal. If good knowledge of geology is available, results may be more accurate by including a Prior Model in the prediction process. This allows a priori information to be included in the prediction (e.g. a prediction interval has a higher likelihood of containing brine sand than shale).

Prior models are applied as part of the Lithofluids process (see Lithofluid Probability) and can be visually inspected using the Predict window (see Using the Predict Window).

Available in: Lithofluids module

Add prior model

Add prior model
  1. In the Control Panel, open the QI tab.
  2. At the tab header, click on the Add icon and select Add Prior Model.
  3. Type a new name for the process and click OK.

Edit prior probabilities

Edit prior probabilities
  1. In the Details Panel, select a LithoFluid PDF (see Importing a LithoFluid PDF). 
  2. If the selected PDF resource has more than one LID model available, a model selection drop down appears below it. This is hidden if there is only one model to choose from.
  3. At PDF Attributes, choose which attribute space to use for the prior model (e.g. AI vs Vp/Vs, AI vs SI or muRho vs lambdaRho). This is only shown if the selected PDF supports more than one attribute space.
  4. At the Windowing Model, a windowing model can be used, instead of a sculpted volume, to define a series of bounded time/depth windows. Alternatively, select « Whole Trace » (the default) to define priors for the whole trace with no windowing.
  5. Under the Edit Prior Probabilities section, choose which window to edit at the Current Layer. This is based on the windowing model that you have chosen, and is hidden entirely when the windowing model has only one layer.
  6. Select the Layer Enabled check box to enable layers. When creating your priors, each layer in the windowing model can be selected independently, and different prior probabilities set.

Choose how priors are entered

For each enabled layer, choose how to define the lithofluid prior probabilities at Prior Entry:

  • Use Constants for Priors — each lithofluid has a single ratio, the same at every trace location.
  • Use Horizons for Priors — each lithofluid's ratio comes from a horizon map, and can vary laterally.

Using constant priors

  1. The lithofluids will be displayed in a table at the bottom of the Details Panel. Click on the green/amber circle to the left of the name to hide/show the items.
  2. You can change the colour of the lithofluids by clicking on the colourbar and picking a different colour.
  3. Type the ratio of the lithofluid in the box or use the slider to adjust the values. Select the check box next to it to turn on auto normalise: this keeps the lithofluid's own ratio fixed and adjusts the other enabled lithofluids to keep the total at 1.
  4. To enable normalisation of all lithofluids, click the Enable/Disable all check box at the bottom. Enabling this causes a normalisation.
  5. To save the values, click on the Copy as XML to clipboard icon at the bottom and paste it into a file.
  6. The status icon at the bottom right shows the running total of the enabled ratios: green with a tick ("All Good. Total is 1") when valid, or red with an exclamation mark ("Warning: Total is not 1") when the ratios don't add up to 1. Click the Give even value to all enabled fields icon to adjust the normalised ratios to even values.

Using horizon-based priors

  1. Each lithofluid in the table shows a horizon drop down list in place of the ratio box and slider. Select a horizon to supply that lithofluid's prior probability at each trace location.
  2. Only single-attribute 3D IL/CL horizons are supported — X/Y horizons cannot be used. All lithofluids in the same layer must reference horizons on the same survey.
  3. If a horizon does not meet these requirements, a warning is shown, for example: "no horizon selected", "horizon must have a single regular attribute", "only 3D il/cl horizons are supported at present", or "horizon is on a different survey from other lids".

Warning: Use a horizon map on a 1 IL x 1 CL grid. Coarser priors maps (e.g. 4 IL x 8 CL) are sampled onto the output grid using nearest-neighbour interpolation, which can imprint visible vertical striping in the output probability volumes, particularly where the map has a steep lateral gradient. Use Regrid (Horizon) to resample the map onto the 3D survey at 1 IL x 1 CL before building the prior model.

Project Above/Below

Project Above and Project Below only appear for a layer that is not the first or last layer of a multi-layer windowing model. They are not shown at all when using « Whole Trace » or a single-layer windowing model. In a multi-layer model, the top layer only shows Project Below, and the bottom layer only shows Project Above.

  1. At Project Above and Project Below, select Constant to enter a fixed value, or Loaded to use a horizon, to define the maximum/minimum extent above/below which data will not be displayed.
  2. When Loaded is selected, choose a horizon from the drop down list. An adjacent box lets you enter a constant offset to apply to the horizon.

Test your priors

  1. Click Test All Priors to validate every layer, everywhere.
  2. The icon and message beside the button report the result: "Priors are untested (hit the button to test all layers, everywhere)" before testing, "Priors checked out OK everywhere" if all priors are valid, or "Problems with priors found. Check warnings above" if not.
  3. For horizon-based priors, this also checks that the ratios sum to 1 at every IL/CL location, warning at any location that fails, for example: "At il/cl 1234/567 priors don't add up to 1. Total is 0.87."