Technology Highlight – Using Horizon Based Geometric Attributes to Accelerate Seismic Interpretation

In this blog post, I demonstrate how newly added horizon-based geometric attributes can accelerate fault identification, and discuss how they can serve as quality control checks for picked horizons.

Geometric attributes, as the name implies, characterize some property of the local surface geometry. There are two parts to the problem when we calculate volumetric attributes: 1) determine geometry and 2) infer some characteristic along this geometry, e.g. similarity. When we calculate geometric attributes on picked horizons, we already know the surface geometry.

Coherence, gradient and curvature are some of the many geometric attributes. Coherence (variance in SeisVision geometric attributes) calculates the similarity (or lack of it, in this case) in picked times in a local grid. The gradient (dip in SeisVision geometric attributes) and curvature attributes are finite difference approximations to first and second derivatives in space. If the horizon is noisy, there might be far too many spurious edges in the resulting attribute; so smoothing as a pre-conditioning step is often a good idea. Users have been able to calculate curvature on horizons historically, but coherence and gradient are recent additions to the attributes family in SeisVision.

Anyway, enough with the talk; here are a few attributes I calculated on a horizon in the F3 dataset:

Fig 1: (L-R, T-B) Time horizon; crossline dip; inline dip; azimuth; dip magnitude; variance

As we can see, different attributes (and vibrant SeisVision color palettes) really bring out the features of interest (the faults in this case).  All of these attributes are a recent addition to SeisVision and, besides identifying faults, can also double as quality or confidence maps for picked horizons.

As another example, let’s take a look at the Stratton dataset. I picked a horizon with a fault intersection to illustrate my point. The section view shows the picked horizon and the fault, while the horizon slice shows the same horizon in map view.

Fig 2 (L-R): Seismic section with horizon marked; Seismic horizon map view

I calculated the variance and curvature attribute for this horizon and I will show how both are powerful tools that can be used in real world interpretation settings.

The variance attribute does an excellent job of removing spurious noise and really brings out the spatial extents of the fault. This is extremely helpful in subsequent workflows such as identifying faults and constraining the gridding operation, etc.

Figure 3: Horizon Variance map

Next up is the maximum curvature attribute, which is not as clean as variance, but contains additional information. Extreme values in the curvature correspond to significant concavity in the data (faults, fractures, etc.), while the polarity indicates the upthrown and downthrown directions. The spacing between the upthrown and downthrown sides corresponds to the fault hade – another important characteristic.

Figure 3: Horizon Maximum Curvature map

As we have seen, faults are much easier to identify on geometric attributes than it is on traditional horizon maps. This has the potential to speed-up many subsequent workflows in seismic interpretation.

I encourage SeisVision users to explore this feature in more detail.

Acknowledgement

The F3 dataset was made available by dGB Earth Sciences B.V. (through OpendTect).