1

## How can I set my parameters for Spectral Decomposition?

I have 3D seismic data with quite wide frequency content. I want to look at the structural features in different frequency bands in more detail. How should I choose my input parameters for Spectral Decomposition to use the data in optimum frequency bands?

Thanks

First of all, I would like to recommend you to generate a 3D seismic sub-volume just for your zone of interest, because you would be calculating additional several spectrally decomposed 3D sub-volumes later in the process, and they can take up so much hard-drive space. You can refer to the below thread for generating sub-volumes:

http://ihskingdom.com/?p=1108

When you run Rock Solid Attributes (RSA), you will have all the input parameters already filled in the entry boxes. The default input parameters are set so that you can have fairly good result as your first attempt, however, they may not of course work at best for all geological settings, for your area in particular. Then, you would need to do some tests on your own to get the optimum parameters for your data set. Even though results may not be so much different from each other, it is still worth spending some time for the best result. Having a sub-volume is also good for such trials.

Here, I can give you some background information and tips in general workflow.

– Go to Tools> Trace Calculations> Survey Spectrum

– Select the survey and the data type and run it (you can even define a subset here on the fly)

– Save the survey spectrum to a file

– Open it in MS excel worksheet and graph it since this would give you much better visual flexibility to examine the spectrum

– Take a note of the pick frequencies, the low and high end slopes

– Run Rock Solid Attributes (RSA)

– Select the 3D sub-volume and the data type

– Select “Spectral Decomp Attributes” from the pull-down menu

– There would be two options for spectral decomposition in IHS Kingdom: SD Envelope Sub-band and SD Trace Sub-band

SD Envelope Sub-band: It gives the amplitude envelope of the frequency bands. It is good in searching for amplitude anomalies at different frequencies.

SD Trace Sub-band: It gives traces at different frequency bands.

– You need to put a descriptive prefix or suffix to identify the calculated sub-volumes

– Click Next for the input parameter dialog box

– You will notice that all the above input areas are grayed-out since you would only be running the spectral decomposition process

– It would come with the default parameters, but you can change them depending on your data, subsurface setting, etc.

– Enter the low pick frequency at amplitude 0.5 (recommended)

– Enter the high pick frequency at amplitude 0.5 (recommended)

– 10 is a good number of bands for a good coverage at the beginning

– Choose either octave scale or linear scale.

– Red dots refer to the center frequencies for the each spectrally decomposed sub-volumes. Make sure those center frequencies coincide with the pick frequencies on your survey spectrum for better representation

Octave Scale: In this selection, you will have fewer side lobe effects and much cleaner data. As it is seen on the graph, there would be more data at the lower frequencies. At the higher frequencies, there will be ringing effects and the features will be less obvious due to much larger band range at fewer central frequencies.

Linear Scale: The distribution will be even, therefore, frequency bands are regular. There will be lesser center frequencies at lower end, therefore their spectrum range will be larger. Sub-volumes at the higher frequencies will have more ringing effect.

Therefore, you may want to use the both scales and compare to see which one gives the best result.

– You may need to adjust your input parameters for optimum result

– Click Next and choose 16-bit output for this round since you may not need 32-bit data for your first attempt, but for the final use, 32-bit output would be better to use

– Click Next to have the list of sub-volumes which will be calculated as sub-volumes

– If you are satisfied with the frequency coverage, then go ahead and run the calculation. Otherwise, go back and change your input parameters

– After the calculations are completed, flip through the spectrally decomposed seismic data at time-slices, in-lines or cross-lines over your prospective area

You would expect to see your structural features much better identified in certain bands. You can mark them to use in your horizon and fault interpretations. You can also calculate other seismic attributes on these spectrally decomposed sub-volumes.

Was this answer helpful?

Like