Using the Horizon Propagator

Using the Horizon Propagator

In this example, we’ll extend some horizon interpretation using the automatic propagator. Using Insight’s picking modes, we can create picks to guide the automatic propagator. See How the Propagator works for background information on the propagator.

This horizon has already been through a round of propagation. It did a good job, but was unable to track the entire event.

If we look closely, we can see why. Towards the right of the image, the character of the waveform changes. A smaller event is introduced above our positive event, and the propagator stops tracking.

We’d like to continue the horizon along this positive event.

We start by stepping through regular IL and CL intervals and manually adding a sparse grid using the guided manual propagator.

Next, we’ll configure the automatic propagator. In the Horizons tab, select the horizon and look in the details panel for the Picking settings.

We’ve chosen the Propagator picking mode. This will use the waveform propagator to pick a horizon automatically, starting from seed points. We then choose our horizon + sparse grid as seed points.

In areas where the waveform changes, we can add extra seed points. Click the pencil icon to add seed points for additional guidance. The seed points can be placed in any view, including the 3D.

Let’s take a look at that in the 3D view. We can see:

  • The original picks
  • A coarsely picked grid
  • Additional seed points for control

Back to the settings!

We’ve specified a 60 ms window length for the waveform. Insight will use a window that is 30ms above and 30ms below the pick.

A Threshold (%) of 75 is set to limit the propagator to follow only the best picks (correlation above 75%). We’ll also set the shift limit to 5 ms at the Search (ms) box to prevent any large trace-to-trace jumps.

There are some faults intersecting the horizon. If had already interpreted the faults, we could tell the propagator to not extend across fault boundaries. Instead, we’ll be cautious with our settings, and confirm later that the propagator doesn’t skip to the wrong event. We can always adjust the parameters and run the propagator again if the propagation doesn’t extend far enough.

Now that we have configured the options, we can click Propagate. With a 3D view open, we can watch the propagation track its way through the volume. As it runs, we can see that newly picked areas appear opaque, while the previously picked horizon is transparent.

After this run of the propagator, the missing patches have been filled in and the horizon extends along the event. In section view, you’ll notice newly propagated areas appear as dashed lines. When you save the propagation, these dashed areas will appear solid with the rest of the horizon.

Before saving the propagated horizon, you can increase the Stop Threshold to a value greater than what was previously specified. As the value increases, only the highest confidence picks are kept, removing the lower correlation areas of the result. This fine control is helpful if there are jumps in the propagated horizon that don’t appear geological.

We’ll keep the threshold at 75% percent, save the propagation, and view the horizon in the 3D view.

The 3D view is great for QC. Let’s take a look at how the propagator performed across faults. In this area, there is a downthrown section which looks like it could be a small graben. To confirm that the horizon has been mapped correctly across the fault, let’s limit the horizon display to where it intersects the section. See how to do this on this page: Show Whole or Intersecting Horizon

Here’s that section with the propagated horizon across the fault. We’ve changed the colorbar to black and white to allow for a clearer view. The horizon appears to have propagated nicely across the event. Our interpretation that small downthrown portion of the horizon is a graben is correct.

Insight’s waveform propagator provides an efficient method to automatically add interpretation over large areas. The propagator’s customizable parameters give a high level of control and allow for accurate results in fewer iterations. It is great for mapping regional surfaces and structures at the prospect scale.