I have been working with some seismic data and recently submitted a paper to a journal.

My seismic data has 493 traces: each trace are 8300 seconds long with sampling frequency of 2000 Hz.

One of the reviewers is asking me to generate a frequency spectrogram at the time= 2000s for all signals. My knowledge about the spectrogram is that you can only do it for one signal and there should be a time window to run the frequency analysis on. How can I do it?

Here are more detail about my problem: These 493 traces recorded vibration around a cable in a oil producing horizontal well. The traces are 16ft away from each other along the well. The idea of the publication was to assess the spectrogram for traces on a specific location that was the target for hydraulic fracturing. The figure below is the spectrogram of one of the traces (signals) close to the target zone. Note that there are 492 other traces and all of them have the same number of samples, frequency, and length. All these traces recorded vibration for the entire 8300secon. The question is how I could generate spectrogram of all 493 traces for example at 2000 second. Here is the wording of the reviewer: "I welcome the addition of a spectrogram showing many channels at one particular time interval." The paper is accepted, But I think it's better to address the reviewer's comment. I will appreciate your name in my acknowledgment.

enter image description here


closed as too broad by A_A, Marcus Müller, Matt L., lennon310, Stanley Pawlukiewicz Nov 6 '18 at 18:17

Please edit the question to limit it to a specific problem with enough detail to identify an adequate answer. Avoid asking multiple distinct questions at once. See the How to Ask page for help clarifying this question. If this question can be reworded to fit the rules in the help center, please edit the question.

  • 1
    $\begingroup$ The way this question is posed makes it very broad. Can you share a bit more information? Maybe the reviewer is asking you to provide the spectrogram to check what is happening around that sample's instance in the dataset? $\endgroup$ – A_A Nov 2 '18 at 8:11
  • $\begingroup$ Yeah, in a DSP-adjacent field (like geology, for example), terminology isn't often used "strictly correct". We can't really guess from your rendition of what they wrote what they actually meant. Maybe the actual wording of your reviewer might help, as well as the figure you already have in there (make sure you don't violate any non-prior-publication conditions by posting that). $\endgroup$ – Marcus Müller Nov 2 '18 at 9:54
  • $\begingroup$ @A_A @ Marcus Müller, I edited my question to have more details about my problem. I appreciate your help. $\endgroup$ – Pk2018 Nov 2 '18 at 12:20
  • 1
    $\begingroup$ The way I read it, the reviewer would like to see a bunch of spectrum plots. Each from a particular signal. All for a particular moment in time. Line these up like a waterfall diagram, arranged so that the individual traces are positioned like the recording sensors (same order.) The idea being to show the distribution of sounds at varying distances from the well. The "loudest" spots could give information about the source of the sounds. $\endgroup$ – JRE Nov 2 '18 at 12:29
  • $\begingroup$ @JRE, this is exactly what the reviewer is asking. He wants to see if the low frequency phenomenon at the target zone can also be seen in all traces spectrogram. You are suggesting to generate for example 20 spectrograms in a figure as subplots? That would be easy to do. My understanding was that he wants only one spectrogram. Thank you for your help. $\endgroup$ – Pk2018 Nov 2 '18 at 13:33

"I welcome the addition of a spectrogram showing many channels at one particular time interval."

If you don't want to include it, or don't have enough space, it is always worthwhile to put them in a supplementary file, attached to the journal, or separately on the Internet. I suggest the reviewer is looking for an onset phenomenon, and it would be easy to generate a few crops in time interval $[1500-2500]$ side-by-side, potentially for only a tenth of your data (about 50 signals).

  • $\begingroup$ the problem is how to make " a spectrogram showing many channels at one particular time interval." $\endgroup$ – Pk2018 Nov 2 '18 at 12:25
  • $\begingroup$ My answer was delayed: putting them side-by-side in a mosaic could be enough. $\endgroup$ – Laurent Duval Nov 2 '18 at 12:27
  • $\begingroup$ My understanding is that the reviewer should like to whether this example is significant or not across the traces. You can put spectrograms in a line, symmetrically around the well location $\endgroup$ – Laurent Duval Nov 2 '18 at 12:37
  • $\begingroup$ Laurent, this is exactly the reason that reviewer is suggesting. So, I just generate 50 spectrogram and put them in a line? You mean as subplots in horizontal way using MATLAb or Python? $\endgroup$ – Pk2018 Nov 2 '18 at 13:37
  • $\begingroup$ For the eye, if the important phenomenon happen in time (around 2000) it is better to align stuff vertically, or rotate the time-frequency plot with time in the y direction, so that feature align. Conversely if the frequency patterns are more important. And you can also crop in frequency. With 49 signals, you can made a nice square $\endgroup$ – Laurent Duval Nov 2 '18 at 13:45

Not the answer you're looking for? Browse other questions tagged or ask your own question.