Vertical Bad Traces in Marine Seismic !

Oil and natural gas exploration -- geology and geophysics
Post Reply
StanMarsh
Member
Member
Posts: 3
Joined: Thu Sep 15, 2016 8:51 am

Vertical Bad Traces in Marine Seismic !

Post by StanMarsh »

Hey guys,

I am trying to apply some steps in seismic processing in Matlab, but the data has some vertical bad traces as you can see in the picture:
Sem filtro.png
I have applied a spectrum analysis in these traces, but the spectrum varies a lot (i.e., if I take the spectrum at an interval from 3-4 seconds or 5-6 and so on). I´ve applied some filters and I got a good result (figure below):
Depois do filtro.png
If I try to add some gain afterwards (regardless if is AGC, IAC...) these traces comes "to life" again.

Any suggestions? The filtering approach is the wrong path in this case? I have simply to remove them and substitute for an average of the traces around?

Thanks for the attention :)

GuyM
VIP Member
VIP Member
Posts: 620
Joined: Sat Mar 24, 2012 11:35 pm

Re: Vertical Bad Traces in Marine Seismic !

Post by GuyM »

Hi -

The "bad traces' look to be caused by swell noise; broadly speaking that is noise caused by the pressure changes and action of waves as they roll down the cable.

Typically the "base" frequency of these is about 0.1Hz, however the energy tends to get highly spatially aliased to give energy that sits in the FK domain across a broad range of spatial frequencies and anywhere from 2.5Hz to 15Hz as a maximum.

There's a bunch of approaches you can use to identify them,

If you create a "peak" and "RMS" amplitude value in a deep time window (say 8000-12000ms) and then display this as a grid, using "SHOTID" as one axis and CHANNEL as another, this will clearly identify channels that have an issue.

You can then :

- select just these channels and kill them using tuned peak/rms values (then interpolate)
- select just these channels and use a low-cut filter
- use a mute in the FK domain shot-by-shot to remove low frequencies at wider spatial frequencies (+ and -)
- employ "projective filtering" type approaches
- split the data up by frequency bands and "gain down" amplitudes above a given threshold

While its not in Matlab, this is about detection of bad traces :
http://seismicreflections.globeclaritas ... tudes.html

This is about removal of swell noise:
http://seismicreflections.globeclaritas ... noise.html

We use MatLab for R+D work (aided by our adoption of HDF5 as a seismic data format, which MatLab reads pretty efficiently.
Main challenge is that its quite slow - once we have a Matlab prototype we then move onto an "engineered" production version.
Last edited by GuyM on Tue Sep 20, 2016 6:25 am, edited 1 time in total.

Tesseract
Silver Member
Silver Member
Posts: 19
Joined: Thu Jun 09, 2016 8:17 am
Location: cyberspace

Re: Vertical Bad Traces in Marine Seismic !

Post by Tesseract »

I would add that in typical marine processing flow, after the "swell noise attenaution" by time-frequency denoise or whatever module is in the package, next step would be decon in tau-p domain. That does a pretty good job of squashing the swell noise, as it is incoherent in the tau-p domain. I don't know if you can do that with Matlab though.

StanMarsh
Member
Member
Posts: 3
Joined: Thu Sep 15, 2016 8:51 am

Re: Vertical Bad Traces in Marine Seismic !

Post by StanMarsh »

Hey there GuyM,

Thank you so much for the complete answer on this one. I´ve read about this issue and it makes total sense. I´m gonna try to use the approaches you´ve mentioned trying to solve this!

regards
GuyM wrote:Hi -

The "bad traces' look to be caused by swell noise; broadly speaking that is noise caused by the pressure changes and action of waves as they roll down the cable.

Typically the "base" frequency of these is about 0.1Hz, however the energy tends to get highly spatially aliased to give energy that sits in the FK domain across a broad range of spatial frequencies and anywhere from 2.5Hz to 15Hz as a maximum.

There's a bunch of approaches you can use to identify them,

If you create a "peak" and "RMS" amplitude value in a deep time window (say 8000-12000ms) and then display this as a grid, using "SHOTID" as one axis and CHANNEL as another, this will clearly identify channels that have an issue.

You can then :

- select just these channels and kill them using tuned peak/rms values (then interpolate)
- select just these channels and use a low-cut filter
- use a mute in the FK domain shot-by-shot to remove low frequencies at wider spatial frequencies (+ and -)
- employ "projective filtering" type approaches
- split the data up by frequency bands and "gain down" amplitudes above a given threshold

While its not in Matlab, this is about detection of bad traces :
http://seismicreflections.globeclaritas ... tudes.html

This is about removal of swell noise:
http://seismicreflections.globeclaritas ... noise.html

We use MatLab for R+D work (aided by our adoption of HDF5 as a seismic data format, which MatLab reads pretty efficiently.
Main challenge is that its quite slow - once we have a Matlab prototype we then move onto an "engineered" production version.

StanMarsh
Member
Member
Posts: 3
Joined: Thu Sep 15, 2016 8:51 am

Re: Vertical Bad Traces in Marine Seismic !

Post by StanMarsh »

Thanks for the complement Tesseract. Since this noise is really strong compared with the signal, this can help solve this.

Regards,
Tesseract wrote:I would add that in typical marine processing flow, after the "swell noise attenaution" by time-frequency denoise or whatever module is in the package, next step would be decon in tau-p domain. That does a pretty good job of squashing the swell noise, as it is incoherent in the tau-p domain. I don't know if you can do that with Matlab though.

Post Reply
  • Similar Topics
    Replies
    Views
    Last post