Why We Should Perform PSDM and Calculate Seismic Attributes in the Depth Domain, Examples from the Woodford Formation

Wednesday, March 28, 2018 • 11:30 AM
Calgary Petroleum Club, Devonian Room (+15 Level), 319 – 5 Ave SW

We are commonly challenged to generate a drill plan in depth from time migrated seismic. In order to meet the needs of interpreters and drillers, PSTM volumes are frequently converted to pseudo depth volumes by stretching. This process affects every part of the drill planning procedure as these pseudo-depth data are input into earth modeling packages and used to design lateral well paths. Our study shows that even using a sophisticated velocity model, the drilling interval does not have the correct thickness and/or interlayer structuring on a depth-stretched dataset; making it difficult to achieve the primary geologic goal of staying in zone.

We examine the difference between time to depth stretch and anisotropic PSDM on a 3D dataset. The selected 60 mi2 3D seismic survey is located in the Anadarko Basin, Oklahoma, with the main target being the Woodford formation that varies in thickness from 100ft to 400ft but the actual drilling window is around 30ft. The seismic velocity within the Woodford is much slower than in the overburden (13,000ft/s and 16,000ft/s respectively). Because it violates the Dix equation, this large velocity contrast cannot be corrected through a time processing sequence and the apparent layer thickness is too large. On the PSDM data, a geologically meaningful velocity model incorporating anisotropy, produces a more accurate thickness estimation and better reflects the inter-formation geology. A comparison between the anisotropic PSDM and the depth-stretched data along a lateral well path demonstrates why drilling on depth-stretched data results in inconsistencies between the predicted and encountered geology.

The detailed anisotropic, amplitude preserved PSDM dataset was used to investigate the possibility of performing impedance inversion in the depth domain. The main challenge is the complexity of the depth wavelet but by focusing on the target interval, a balanced result is achieved.

The inversion test consists of 3 phases; first, the anisotropic PSTM stack is inverted using log data of 3 wells. This was done to produce a base result with a well understood technology. Next, the amplitude preserved PSDM stack is inverted. Lastly, the PSDM stack is stretched to time using the final PSDM velocities, inverted to impedance and the results are stretched back to depth, using the same velocity field. All 3 results are compared to each other and analyzed for accuracy and resolution. The products reveal that the depth inversion is not affected by any artifacts and actually has a higher resolution compared to the other two inversion as the PSDM input data has a wider frequency bandwidth.

Last but not least, let’s look at azimuthal velocity attributes that are being used to map fractured intervals within the shale reservoir units. Wide-azimuth 3D seismic data are suitable for anisotropic/ azimuthal processing in both, the time and depth domain. Adjusting for the vertical and horizontal transverse isotropy (VTI and HTI) during the pre-stack time/depth migration improves the final image. By-products of this process are azimuthal velocity attributes, which in theory could be used to predict inhomogeneous areas within the target interval. As this is becoming increasingly common, we wanted to investigate the repeatability and actual usefulness of these attributes. We designed a study that awarded the same project to 4 individual seismic processing companies and compared and analyzed the results in the time domain. Following this study, a similar analysis in the depth domain was performed.

Our findings show that the migrated seismic image clearly benefits from HTI corrections if a sinusoidal behavior is present in the COCA gathers. However, the resulting velocity attributes are of varying quality and only conform to the geological attributes when those are filtered back to represent larger scale features. This made the attributes less useful for mapping smaller scale discontinuities such as localized fracture corridors/ swarms within the Woodford interval. When interpreting the azimuthal velocities in depth, the velocity difference between Vfast and Vslow became insignificant which confirmed our previous learnings.

In conclusion, we believe that it is more accurate to perform PSDM processing instead of stretching the time migrated data to depth. Our post and pre-stack inversion tests indicate that it is doable to perform inversions in the depth domain, which again, eliminates stretching the data. When investigating the usefulness of azimuthal velocities we conclude that they can help sharpening the image but azimuthal velocity variations are most likely related to geological changes and not to fracturing. When this analysis is performed in depth, most of the azimuthal velocity differences become so small that they are considered to fall within the noise cone.