Ground Roll Attenuation of Multicomponent Seismic Data with the Noise-Assisted Multivariate Empirical Mode Decomposition (NA-MEMD) Method
Round 1
Reviewer 1 Report
The manuscript presents an improved method for ground roll attenuation in three-component seismic data. The method looks fairly straightforward and should be relatively easy to implement. Given the growing number of active-source experiments collecting three-component seismic data the methods is of potential interest for the scientific community. The paper is brief, but the main sections (description of methods, synthetic and field data and results) appropriately describe the method. I recommend the manuscript for publication after moderate revisions, which should address the following comments.
Comment 1: The manuscript was hard to read and I suggest English language to be reviewed/edited before publication. Not being native speaker myself I don't feel qualified to suggest particular changes, but I was having hard time understanding some sections of the paper and had to re-read them a few times. A few examples:
- The very first sentence of the abstract (line 8): "Multicomponent seismic exploration provides more wavefield information to image complex subsurface model.": Is model the right word? Or does it provide information enabling us to image complex subsurface structures/features?
- The second sentence (line 9): "Ground roll is strongly coherent noise in land multicomponent seismic data and presents similar feature in each component." What similar feature does it present? Does it relate to noise coherency? It either needs more information or the second part of the sentence can be left out.
Many sentences (including those in the abstract) use passive voice. Replacing with active voice would improve readability and reader's engagement.
Comment 2: In abstract, you state that "Ground roll attenuation is an important step in seismic data processing." You should say why, it may not be clear to all the readers.
Comment 3: Figure 8 and the related discussion. It looks like the NA-MEMD algorithms performs almost identically as FK filtering for frequencies > 20Hz. You state that NA-MEMD performs better at low frequencies as it keeps more useful low-frequency signals in the records.
- How do you know that the low-frequency signal is useful and not just an unfiltered remnant of the ground roll?
- Spectrogram Fig 8 d shows that more of the low-frequency signal/noise was retained by the FK filter although you claim the opposite. Can you comment on that?
- At line 255 you say that "low frequency range (<10Hz), .. is important to the later high resolution imaging". Why is that? Usually low frequencies enable greater depth range, but you need high frequencies to improve resolution. Perhaps it should be explained better.
Author Response
Dear Reviewers:
Thank you for your comments concerning our manuscript entitled “Ground roll attenuation of multicomponent seismic data with the noise-assisted multivariate empirical mode decomposition (NA-MEMD) method”. (ID: applsci-1576336). Those comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our researches. We have studied comments carefully and have made correction which we hope meet with approval. Revised portion are marked in the paper. The main corrections in the paper and the responds to your comments are as follows:
Responds to the reviewer’s comments:
Comment 1. The manuscript was hard to read and I suggest English language to be reviewed/edited before publication. Not being native speaker myself I don't feel qualified to suggest particular changes, but I was having hard time understanding some sections of the paper and had to re-read them a few times.
Response:
(1) We edit English language for this manuscript by professional Language Editing service (Editage). Some revisions are marked in paper.
(2) Revised: Ground roll is strongly coherent noise in land multicomponent seismic data and exhibits similar features, which are strong energy, low frequency, low velocity and dispersion, in each component.
(3) I revise the manuscript and express as much as possible with active voice.
Comment 2: In abstract, you state that "Ground roll attenuation is an important step in seismic data processing." You should say why, it may not be clear to all the readers.
Response: Ground roll attenuation is an important step in land seismic data processing. In introduction section (line 6), I say the reason.
Comment 3: Figure 8 and the related discussion. It looks like the NA-MEMD algorithm performs almost identically as FK filtering for frequencies > 20Hz. You state that NA-MEMD performs better at low frequencies as it keeps more useful low-frequency signals in the records.
Response: Ground roll and reflection wave exhibit difference in main frequency and frequency band. Reflection wave has high main frequency and broad frequency range from a few Hz to hundreds of Hz, but ground roll dominates at low frequencies. For NA-MEMD method, IMFs of low-frequency scale is subtracted for ground roll attenuation. We can suppress random noise by subtracting IMFs of high-frequency scale. F-K filtering method performs filtering operation in the frequency-domain domain, which cuts low-frequency and low velocity contents based on filter parameters.
- How do you know that the low-frequency signal is useful and not just an unfiltered remnant of the ground roll?
Response: For reflection wave, low-frequency information is useful. In the absence of low-frequency content, side lobe of seismic wavelet is powerful, which is not good for high resolution imaging. Seismic data are superposition of ground roll and reflection. For ground roll attenuation, it is tradeoff between maximum removal of ground rolls and minimum damage to effective reflection waves.
- Spectrogram Fig 8 d shows that more of the low-frequency signal/noise was retained by the FK filter although you claim the opposite.
Response: Ground roll dominates at low frequencies. For F-K filtering method, High cut-off frequency can help improve ground roll attenuation, but low-frequency reflection wave will be cut. To less damage to reflection wave, filter cut-off parameters of 10Hz frequency and max velocity of ground roll are selected, which can make ground roll retained in 10-20Hz frequency. For NA-MEMD method, frequency contents (<20Hz) is well attenuated, but some low-frequency contents are still preserved for the corresponding reflection wave.
- "low frequency range (<10Hz), .. is important to the later high resolution imaging". Why is that?
Response：We can expand high-frequency signal/noise by deconvolution, which improves resolution for reflection wave. Low frequencies are useful for late data processing. In the absence of low-frequency content, side lobe of seismic wavelet is powerful, which is not good for high resolution imaging and inversion, such as deconvolution and full waveform inversion.
Reviewer 2 Report
Applsci-1576336 Review
[1] This paper investigates the NA-MEMD method for suppressing ground roll in seismic reflection data and shows that it is slightly better than the FK method. The paper could be accepted after consideration of the following comments and questions, which are based on the PDF file of the paper.
[2] There are small and moderate errors in English grammar and style in the paper, but most of the paper can still be read and understood.
[3] Line 33: “Halliday” should be “Halliday et al.”. When there are three or more authors, the paper is usually cited as “First Author et al.”. The other citations in the paper should also be checked for this error.
[4] Lines 35-38: Some of the citations in these lines (Le Muir, Melo, Boustani) are missing in the References list.
[5] Some of the other citations elsewhere in the paper are also missing in the References list (see, e.g., lines 51 and 64). The authors should check to make sure that all their citations are included in the References list.
[6] Line 87: When step 1 is repeated, is r_n used as the new input signal?
[7] Line 126: Should this be “IMF1, IMF2 and IMF4, respectively”?
[8] Figure 4c and 4f: It is not clear what these figures show. What do you mean by “separated ground roll?
[9] Figure 6a-6c: Which is the vertical and which are the horizontal components? Is it x, y, z, left to right?
[10] Figure 6d-6f: The reflections do not show up very well.
[11] Lines 246-252: I see only Figure 7a and 7b. Figure 7c and 7d are missing in the PDF file of the paper.
[12] Lines 248-249: But according to the caption, Figure 7a and 7b show the FK filter, not the NA-MEMD method. (?)
[13] Figure 6f and 7a: Presumably, these both are the attenuated vertical component. When I compare 6f and 7a, I don’t see a big difference. Presumably the NA-MEMD method is more involved and costly and time-consuming than the FK method. I am wondering if the extra effort in the NA-MEMD method is worth it. I realize that the NA-MEMD method shows more low frequency content (below 10 Hz), as seen in Figure 8c and 8d, but does this make a big difference?
[14] Figure 8c: Why does the FK spectrum (green line) have a rising part but no falling part that decreases to zero?
[15] Figure 8c and 8d: The FK spectrum (green line) seems to be confined to the range 0-20 Hz, whereas the NA-MEMD spectrum (red line) has a lot of high-frequency content, going up to 120 Hz. That does not seem correct. Why is that? Should the two spectra not be more or less the same?
Author Response
Dear Reviewer:
Thank you for your reviewers’ comments concerning our manuscript entitled “Ground roll attenuation of multicomponent seismic data with the noise-assisted multivariate empirical mode decomposition (NA-MEMD) method”. (ID: applsci-1576336). Those comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our researches. We have studied comments carefully and have made correction which we hope meet with approval. Revised portion are marked in the paper. The main corrections in the paper and the responds to your comments are as follows:
Comment 1. This paper investigates the NA-MEMD method for suppressing ground roll in seismic reflection data and shows that it is slightly better than the FK method. The paper could be accepted after consideration of the following comments and questions, which are based on the PDF file of the paper.
Response: Thank you for valuable comment.
Comment 2. There are small and moderate errors in English grammar and style in the paper, but most of the paper can still be read and understood.
Response: We edit English language for this manuscript by professional Language Editing service (Editage). Some revisions are marked in paper.
Comment 3: “Halliday” should be “Halliday et al.”. When there are three or more authors, the paper is usually cited as “First Author et al.”. The other citations in the paper should also be checked for this error.
Response: I edit the citation format by reference numbers placed in square brackets. I also check citations for three or more authors.
Comment 4: Some of the citations in these lines (Le Muir, Melo, Boustani) are missing in the References list.
Response: These references are short conference proceedings. We cite references in reference list or revise text.
Comment 5: Some of the other citations elsewhere in the paper are also missing in the References list (see, e.g., lines 51 and 64). The authors should check to make sure that all their citations are included in the References list.
Response: I check all citations and reference list.
Comment 6: Line 87: When step 1 is repeated, is r_n used as the new input signal?
Response: Yes, EMD method is an iterative process to decompose a signal to different IMFs.
Comment 7: Line 126: Should this be “IMF1, IMF2 and IMF4, respectively”?
Response: I check and revise the text: “IMF1, IMF3 and IMF4, respectively”
Comment 8: Figure 4c and 4f: It is not clear what these figures show. What do you mean by “separated ground roll?
Response: Figure 4c and 4f show ground roll separated.
Comment 9: Figure 6a-6c: Which is the vertical and which are the horizontal components? Is it x, y, z, left to right?
Response: In Figure 6a-6c, they are horizontal components(x,y) and vertical component(z). I revise this description” Figure 6(a), 6(b) and 6(c) show the raw seismic data that are recorded X , Y and Z components from left to right, respectively. ”.
Comment 10: Figure 6d-6f: The reflections do not show up very well.
Response: The reflections have low signal/noise ratio, especially for Y component. Our method suppress only low-frequency ground roll.
Comment 11-12: Lines 246-252: I see only Figure 7a and 7b. Figure 7c and 7d are missing in the PDF file of the paper.
Response: Figure 7 (a) and (b) is included. For comparing, the corresponding NA_MEMD results is displayed in Figure6 (f) and 6(i).
Revise text: Figure 7(a) and 7(b) display the attenuation results and differences of F-K attenuation method for the vertical component, respectively. In Figure 6(f) and 6(i), we can see that the ground roll is well attenuated by the NA-MEMD method. However, the F-K method suppresses the ground roll by cutting operation in the frequency-wavenumber domain, resulting in deletion of the low-frequency content of the reflection wave. Therefore, some weak reflection event energy remains in the difference section, as shown in Figure 7(b).
Comment 13: The extra effort in the NA-MEMD method is worth it? The NA-MEMD method shows more low frequency content (below 10 Hz), as seen in Figure 8c and 8d, but does this make a big difference?
Response: NA-MEMD method is time-consuming than F-K method, but there are two advantages. The MEMD method is performed in time domain trace by trace. Therefore, space sampling requirement is not limited. Low frequencies are useful for late data processing. In the absence of low-frequency content, side lobe of seismic wavelet is powerful, which is not good for high resolution imaging and inversion, such as deconvolution and full waveform inversion.
Comment 14: Figure 8c: Why does the FK spectrum (green line) have a rising part but no falling part that decreases to zero?
Response: Ground roll dominates at low frequencies and has low velocity. For F-K filtering method, High cut-off frequency can help improve ground roll attenuation, but low-frequency reflection wave will also be cut. To less damage to reflection wave, filtering parameters of 10Hz cut-off frequency and max velocity of ground roll are selected, which can cause a rising part in 10-20Hz frequency.
Comment 15: Figure 8c and 8d: The FK spectrum (green line) seems to be confined to the range 0-20 Hz, whereas the NA-MEMD spectrum (red line) has a lot of high-frequency content, going up to 120 Hz. That does not seem correct. Why is that? Should the two spectra not be more or less the same?
Response: In our work, we only suppress ground roll with low frequency and low velocity. The results of two methods are different in low-frequency band. Random noise attenuation with high-frequency content is not considered. Therefore, the high-frequency contents are completely preserved using both methods. We can suppress random noise by subtracting IMFs of high-frequency scale using the MEMD method, or by high cut filter using the F-K method.
