Friday, January 28, 2011

Spectral decomposition at Dickman

In our ongoing effort to relate seismic attributes to geologic features at Dickman, we have generated a series of narrow-band attribute volumes from the migrated data. Unlike traditional time-frequency methods (Chakraborty and Okaya, 1994) that seek a trade-off of time and frequency resolution, we have chosen to use a pure frequency isolation algorithm. In fact, the method we use is traditional Fourier bandpass filtering with a very narrow response centered on the frequency of interest. In Figure 1 we show (a) the original 3D migrated data spectrum as extracted in a 5x5 bin area, (b) the same after narrow band filtering around 6 Hz and (c) after 43 Hz narrow band filtering.

Figure 1. Spectrum of Dickman seismic data and narrow band decomposition. (a) Fourier amplitude spectrum of 3D migrated data in a 5x5 bin area from 0-2 seconds. (b) Spectrum after narrow bandpass filtering centered on 6 Hz. (c) Spectrum after 43 Hz narrow bandpass filtering.

In the vertical view (Figure 2) the narrow band results are not very enlightening. It is tempting to conclude the new data has no time information content, but in fact there is time localization in the amplitude modulation of each trace although it seems to have little value in the vertical view.

Figure 2. Representative seismic line before and after filtering. (a) Original broadband data. (b) Narrow band 6 Hz data. (c) Narrow band 43 Hz data.

Figure 3 shows a time slice through the broadband data at 848 ms, roughly coincident with the Miss/Penn unconformity. The prominent incised channel is clearly shown. Note there are no clear trends in the data aligned with the yellow dash lines and the channel does not seem to approach the tip of the yellow arrow.

Figure 3. Broadband timeslice at 848 ms, approximately coincident with the Miss/Penn unconformity. Features observed in narrow band data (Figs 12 and 13) but not on the broadband data are indicated in yellow.

A coincident time slice through the 6 Hz data (Figure 4) shows a strong and remarkable diagonal alignment parallel to the yellow dash lines. One always suspects acquisition footprint when linear features show up in seismic data, so Figure 5 shows the shot and receiver orientation for the Dickman 3D data. While it is possible the NW-SE trend is related to source line orientation, the conjugate direction has no association with shooting geometry. We suspect these features indicate fracture orientation as described using curvature by Nissen et al. (2004, 2006). We plan to investigate this alignment further in the next quarter.

Figure 4. Narrow band (6 Hz) timeslice at 848 ms. Yellow dash lines indicate orientation of diagonal features (perhaps related to fractures) not seen in Fig 3.

Figure 5. Shooting geometry orientation for Dickman 3D, shots (L) and receivers (R).


At a higher frequency band (43 H, Figure 6) we observe a dark channel-like feature as indicated at the tip of the yellow arrow. We are currently working to confirm or deny the reality of this channel feature. The process will consist of studying well logs inside and outside the feature; particularly concentrating on basal Pennsylvanian sediments and indications of a subtle structural low at the top Mississippian.

Figure 6. Narrow band (43 Hz) timeslice at 848 ms. Yellow arrow indicates channel feature not seen in Figure 3.

Method

Spectral decomposition (SD) figures shown above were generated using SeismicUnix (SU) on segy data output from SMT's Kingdom software. Specifically the sufilter program was driven by a shell script that implemented the following steps:

for each frequency of interest, stepping by 1 Hz {
__read segy and convert to SU format
__set params to isolate one frequency
__apply sufilter
__extract time slice
__create pdf figure
}

This allowed quick narrow band data scanning for features of interest in a common time slice. Once a particular narrow band was selected for further analysis, sufilter was applied to the original data and the entire SD attribute volume was output in segy format. This was imported to SMT for further analysis.

While the multiple-frequency scanning cannot be currently done in SMT, a narrow band SD of the type described here can be calculated (Figures 8 and 9) using Tools>>Trace_Pak...>>Process_Multiple_Traces....

Figure 8. Parameter window to create narrow band data directly in SMT Kingdom using TracePak, generating a new data type named amp_6Hz_SMT.

Figure 8. Narrow band (6 Hz) time slice generated and displayed in SMT Kingdom. Compare Figure 4 showing 6 Hz data generated and displayed using SeismicUnix.

Similarity of the SU and SMT result is reassuring, suggesting the linear features are robust features of the data and not merely artifacts related to a particular implementation of bandpass filtering.

References

Nissen, S. E.,T. R. Carr, and K. J. Marfurt, 2006, Using New 3-D Seismic attributes to identify subtle fracture trends in Mid-Continent Mississippian carbonate reservoirs: Dickman Field, Kansas, Search and Discovery, Article #40189.

Nissen, S. E., K. J. Marfurt, and T. R. Carr, 2004, Identifying Subtle Fracture Trends in the Mississippian Saline Aquifer Unit Using New 3-D Seismic Attributes: Kansas Geological Survey, Open-file Report 2004-56.