the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 01 Apr 2025
Passive seismic imaging of the Lower Palaeozoic in the Sudret area of Gotland, Sweden
Abstract. Passive seismic data were acquired together with active seismic data along a 2.8 km long profile in the Sudret area of Gotland, Sweden, as part of a feasibility study for storage of CO2 below the Baltic Sea. Seismic interferometry using cross-correlation and cross-coherence was employed on the passive seismic data. Correlation was used to retrieve virtual shot gathers containing mainly surface waves, while cross-coherence was used to retrieve mainly seismic reflections. Inversion for shear wave velocity and CDP processing of the passive data result in velocity profiles and images that correlate well with borehole data, synthetic seismograms and the active seismic data. Both the passive surface wave and body wave results provide geological information which complement the active data, the surface waves providing S-wave velocity information and the body waves providing a lower frequency image. The passive data are consistent with the active data and there is no indication of any large-scale faults in the area. Furthermore, analysis of the frequency and direction of the ambient noise using power spectral density and beam forming shows that ocean waves and human activity around the island of Gotland makes the Sudret area an ideal location for passive imaging. Our results illustrate that passive seismic imaging can be an important complement to active seismic data for evaluating the subsurface with respect to CO2 storage and monitoring in the Gotland area, Sweden, and perhaps, elsewhere.
- Preprint
(2969 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2025-1325', Anonymous Referee #1, 23 Apr 2025
-
AC1: 'Reply on RC1', Zhihui Wang, 15 Jun 2025
Thanks for your comments and suggestions. We would like to give responses point by point as follows.
Comment 1: Figure 1. Add labels for Figure 1(b) (e.g., x (m) and y(m)). If possible, a lighter background color will make the symbols more straightforward to distinguish from the background.
Response 1: We have reproduced Figure 1 following your suggestion.
Comment 2: Line 90. Since the operating period is 14 hours, it is better to state the exact start time and end time of the data.
Response 2: Thanks. We have state the exact time form 17:00 on 12 November, 2023 to 07:00 on 13 November, 2023 (UTC+1) instead of 14 hours.
Comment 3: Line 122. Describe the strengths and weaknesses of two methodologies and why two approaches were conducted for the surface and body wave extraction, with appropriate references.
Response 3: A lot of comparisons have been conducted by many researchers. We will cite some references to describe the strengths and weaknesses of the two methodologies.
Comment 4: Line 208. The figure citation here may be Figure 7, not Figure 5. I wonder why the numbers 150, 280, and 360 ms here do not match 170, 280, and 360 ms in line 185 for the synthetic seismogram section.
Response 4: Thanks for pointing out the mistakes. It is Figure 7 in Line 208. And we got a mistake to state the correlation of the first reflections in Figure 5 and Figure 7. Both amplitude peaks are at about 170 ms.
Comment 5: Figures 7 and 8: To contrast the background seismic images, it would be good to use a color other than green for the fonts.
Response 5: Thanks again. We will change the color of the fonts to make them much clearer.
Comment 6: Line 236: It would be good to have references for the relevance of 1, 5, 12, and 16 Hz originating from ocean waves and human activities. Can we assign the four frequencies to either ocean waves or human activities?
Response 6: Usually, the noise sources with the frequency of < 1 Hz come from atmospheric activities, earthquake and ocean waves etc and the ambient noise with the frequency of > 1Hz is generated by human and environmental activities. The research area is located on an island and surrounded by the Baltic sea, so the noise source with the frequency of 0.7-1 Hz is probably ocean waves.
Comment 7: Line 278: Why does the geometric location of sensors at the southern tip of an island allow noise from nearly all directions?.
Response 7: This is because the ambient noise sources originated from different countries around the island of Gotland, such as in mainland Sweden, Finland, Estonia, Poland, Germany, Denmark, Norway, etc.
Citation: https://doi.org/10.5194/egusphere-2025-1325-AC1
-
AC1: 'Reply on RC1', Zhihui Wang, 15 Jun 2025
-
RC2: 'Comment on egusphere-2025-1325', Anonymous Referee #2, 01 May 2025
Review of the manuscript “Passive seismic imaging of the Lower Palaeozoic in the Sudret area of Gotland, Sweden” submitted to the Solid Earth journal
In this research work, the authors consider ambient seismic noise interferometry as a possible tool for the structural studies and monitoring of CO2 storage. They justify their choice by a high level of ambient seismic noise in the studied area, as well as that passive seismic methods are cost-effective and environmentally friendly. The authors recorded passive and controlled source seismic data by a linear seismic array and processed them using the conventional ambient noise seismic interferometry technique. They compared results obtained by controlled source and passive methods and made a conclusion about structural features of the studied environment, and that further development of the method is needed. They consider synthetic tests as a further development of the method.
General comments
The topic of the research is important because the development of cost-effective and environmentally friendly methods for structural studies of the near surface is an actual task. Despite actuality and significant work done by the authors, I think that the manuscript cannot be accepted for publication in its current shape and should be resubmitted after additional work for the following reasons:
- The authors don’t justify why they separated data processing into the evaluation of surface wave and body wave parts of empirical Green’s functions (EGF). What is the reason to use different formulas to evaluate these parts? They should justify this by the results of previous research or their observations. Why is cross-correlation better for extracting surface waves but cross-coherence for body waves? Is it possible to evaluate the surface wave part of the empirical Green's function by cross-coherence and the body wave part by cross-correlation? Why is it impossible to have the same procedure for both types of waves?
- They don’t prove that they extracted reflected waves. I am concerned that the ambient noise, analysed by them, consists of only surface waves, because the dominant source of seismic energy is the ocean in their case. Despite mentioning anthropogenic sources, they don’t provide any polarisation analysis of ambient noise and EGF. In my opinion is more reasonable to split the whole record into “body wave dominated” windows of ambient noise, and “surface wave dominated” windows. This is possible to do with polarisation analysis of ambient noise seismic records in the selected frequency band. After this is possible to cross-correlate them separately and stack them accordingly to separate the body wave and surface wave parts of EGF. In my opinion, separating them by applying cross-correlation and cross-coherence is not possible.
- Analysis of directivity and frequency content of ambient noise should be done prior to cross-correlation. Providing this analysis in the discussion part makes reading confusing. Difficult to understand why they selected particular frequency ranges and the EGF evaluation method. I recommend moving this text from the discussion part to a separate chapter (after the description of the experiment).
- The authors calculated cross-correlation functions in the frequency domain, which is wrong in their case. This is possible only for the case of symmetric EGF, i.e. in the case when sources of ambient seismic noise are distributed almost homogeneously on azimuths. This is justified in experiments where ambient noise records have the length of months or years, but not for short-term experiments. From the azimuthal distribution presented by the authors, it’s obvious that the sources are mainly outside of the Fresnel zone. In that case, only cross-correlation in the time domain with further velocity correction can be applied.
- Authors don’t show signal (EGF). Therefore is difficult to estimate the correctness of dispersion curves. Moreover, they don’t provide enough information about dispersion curve inversion, which makes it difficult to estimate the quality of the results.
- In the conclusion part authors write “A next step would be to set up synthetic models to test how much CO2 is required in the bedrock in order for it to be detectable by the passive methods presented in this paper”. By this conclusion, they state that they did not obtain the expected results. What is the reason to publish this manuscript in that case? I recommend conducting a significant study with synthetic data first. For example, they could calculate synthetic seismograms of waves propagating through the synthetic model of the considered environment (fortunately, they have drillhole data) with different content of CO2. These results will help in understanding limitations, and authors should include these results in the manuscript. In that case would be possible to make clearer conclusions about the structure of the studied environment.
Specific comments
P1 L14-16. The sentence should be rephrased. Seismic waves cannot provide geological information. Their interpretation can provide structural information about the geological medium. What does “geological information complement active data” mean? What is low low-frequency image? Maybe “image, obtained by analysing the signal of low frequency”? “The passive data are consistent with active data” Better to rephrase: “Results obtained by analysing controlled source seismic data and passive seismic data are in a good correlation”.
P1 L 19. “Passive seismic imaging ….” should be better to add “…through ambient noise passive seismic interferometry”, because this is also possible through the analysis of earthquake records.
P1 L20 “evaluating” sounds strange. “for structural studies of near surface ” sounds better.
P2 L55 “limited source energy” This is a poorly justified statement. It is possible to use an artificial source of high enough energy to study the crust down to dozens of kilometres.
P3 L69 “Ordovician generates are clear reflection”. Structure itself cannot generate reflected waves, but it can reflect waves. Better to write: “The top of the Ordovician has strong reflectivity according to results of controlled source seismic data analysis”.
P3 L71-72 The sentence is unclear. Please rephrase.
P5 Table 1. What do the shot interval and the number of source points in passive seismic data mean?
P 7 “4.1. Calculating and picking of dispersion curves” better to write “Dispersion curve extraction”
P 8 “4.2. Inversion for the velocity model” better to write “Dispersion curve inversion”.
P 10 Figure 4. Distances instead of channels are more useful. Arrivals of all phases and apparent velocities should be added to the plot.
P10 Figure 4 caption. “noise reduction” sounds better than “noise attenuation”
P 11 “7.1. Shear wave velocity model” sounds better.
P 11 What initial model (parameters of the medium) did you use for dispersion curve inversion? Better to provide this as a table. More details are needed about the inversion process.
P13 There is poor correlation between sections obtained by analysing passive and controlled source seismic data. Only one boundary is visible on both sections. I guess this is because of the absence of reflected waves on the ambient noise. The authors did not prove that they extracted body waves. I would recommend conducting polarisation analysis of the extracted signals.
P14 L236-237 “Particularly, body waves and surface waves can be separated from each other at a frequency of 7 Hz.” How do you separate these waves?
P15 L246. “maximum values of the beam forming analysis in…”. Do you mean “maximum beam power”?
P 15 Figure 10. The sources' azimuths are mainly near-perpendicular to the profile in subplot b. Why are such differences in velocities observed? Have you corrected the phase velocities, taking into account this distribution? This also shows that formula (1) cannot be used to evaluate empirical Green’s functions.
P 15 L260-261. “No clear top of basement reflection is observed in the active data, but this may be due to the higher frequency nature of these data.” Why? How does reflectivity correlate with the frequency of the signal?
Citation: https://doi.org/10.5194/egusphere-2025-1325-RC2 -
AC2: 'Reply on RC2', Zhihui Wang, 15 Jun 2025
Thanks for your comments and suggestions. We would like to give responses point by point as follows.
General comments
Comment 1: The authors don't justify why they separated data processing into the evaluation of surface wave and body wave parts of empirical Green's functions (EGF). What is the reason to use different formulas to evaluate these parts? They should justify this by the results of previous research or their observations. Why is cross-correlation better for extracting surface waves but cross-coherence for body waves? Is it possible to evaluate the surface wave part of the empirical Green's function by cross-coherence and the body wave part by cross-correlation? Why is it impossible to have the same procedure for both types of waves?
Response 1: Thanks for your questions. We are sorry for missing this part in the manuscript. A lot of comparisons have been conducted by many researchers, such as Snieder et al.(2009), Nakata et al.(2011), Zhang et al.(2019). In this research, they describe the strengths and weaknesses of cross-correlation and cross-coherence and state that cross-correlation is stable, clearer and less noisy and cross-correlation can enhance temporal resolution and be able to retrieve the body wave data. So we followed their research results. We will supplement why the two approaches were conducted for the surface and body wave extraction with appropriate references. It is possible to evaluate the surface wave part of the empirical Green's function by cross-coherence and the body wave part by cross-correlation, but we have not mentioned that it is impossible to have the same procedure for both types of waves in the manuscript. In fact, there are some papers that justify the same procedure for extracting body waves and surface waves.
Comment 2: They don't prove that they extracted reflected waves. I am concerned that the ambient noise, analysed by them, consists of only surface waves, because the dominant source of seismic energy is the ocean in their case. Despite mentioning anthropogenic sources, they don't provide any polarisation analysis of ambient noise and EGF. In my opinion is more reasonable to split the whole record into "body wave dominated" windows of ambient noise, and "surface wave dominated" windows. This is possible to do with polarisation analysis of ambient noise seismic records in the selected frequency band. After this is possible to cross-correlate them separately and stack them accordingly to separate the body wave and surface wave parts of EGF. In my opinion, separating them by applying cross-correlation and cross-coherence is not possible.
Response 2: Thanks again. It is much more difficult to retrieve body waves than surface waves from ambient data, but extracting body waves from ambient data is possible in some cases. Our research area is located on an island and surrounded by mainland Sweden, Finland, Estonia, Poland, Germany, Denmark, Norway, etc, so it is possible to record human activities noise. In our view, polarisation analysis of ambient noise would be conducted with 3 component data, however, our SmartSolo nodes only recorded the vertical component. Of course, we can provide EGF in the revised manuscript. Before cross-correlation and cross-coherence, we did a lot of frequency and directional analysis of ambient data and separated surface waves and body waves with different frequency windows as shown in Table 2. Surface waves dominate at the frequencies of 0.7 - 6 Hz in Figure 2b, so we filtered out ambient data with the frequency of less than 6 Hz for body waves data processing. One retrieved shot gather shown in Figure 4 gave us confidence to continue data processing for body waves, because a direct, or diving, P-wave with a velocity of c. 3750 m/s, a direct S-wave with a velocity of c. 17003 m/s, air waves with a velocity of c. 350 m/s and reflections from the top of Ordovician with a NMO velocity of c. 3200 m/s correlate with active shot gathers. We will post the comparisons in the revised manuscript. A synthetic seismogram was produced from convolving the reflection coefficient log with a 10 Hz Ricker wavelet and it correlated well with our stacked section. Both the virtual shot gather and stacked section support that we extracted body waves from the ambient noise data.
Comment 3: Analysis of directivity and frequency content of ambient noise should be done prior to cross-correlation. Providing this analysis in the discussion part makes reading confusing. Difficult to understand why they selected particular frequency ranges and the EGF evaluation method. I recommend moving this text from the discussion part to a separate chapter (after the description of the experiment).
Response 3: We totally agree with your comments 3 that the directional and frequency content of ambient noise should be moved prior to cross-correlation and cross-coherence. Also, much more analysis of ambient noise data will be provided.
Comment 4: The authors calculated cross-correlation functions in the frequency domain, which is wrong in their case. This is possible only for the case of symmetric EGF, i.e. in the case when sources of ambient seismic noise are distributed almost homogeneously on azimuths. This is justified in experiments where ambient noise records have the length of months or years, but not for short-term experiments. From the azimuthal distribution presented by the authors, it's obvious that the sources are mainly outside of the Fresnel zone. In that case, only cross-correlation in the time domain with further velocity correction can be applied.
Response 4: Thanks for your comments, but we respectfully disagree with some of your points. As shown in Figure 10a , sources of ambient noise contributing to extracting surfaces with a velocity less than 4000 m/s came from the NNW and the azimuth of the ambient noise sources are consistent with the seismic survey line and located in the stationary phase zone, so we can calculate cross-correlation functions in the frequency domain and do not need to correct for velocity.
Comment 5: Authors don't show signal (EGF). Therefore is difficult to estimate the correctness of dispersion curves. Moreover, they don't provide enough information about dispersion curve inversion, which makes it difficult to estimate the quality of the results.
Response 5: Thanks for your question. We will provide some information about the data processing and inversion.
Comment 6: In the conclusion part authors write "A next step would be to set up synthetic models to test how much CO2 is required in the bedrock in order for it to be detectable by the passive methods presented in this paper". By this conclusion, they state that they did not obtain the expected results. What is the reason to publish this manuscript in that case? I recommend conducting a significant study with synthetic data first. For example, they could calculate synthetic seismograms of waves propagating through the synthetic model of the considered environment (fortunately, they have drillhole data) with different content of CO2. These results will help in understanding limitations, and authors should include these results in the manuscript. In that case would be possible to make clearer conclusions about the structure of the studied environment.
Response 6: Thanks. This paper focuses on presenting passive seismic imaging of the Lower Palaeozoic in the Sudret area, and provides a friendly to the environment and cost effective complement for active seismic. In the discussion part, we discuss the future work to monitor CO2 injection. However, synthetic modeling is beyond the scope of this paper currently. And surface waves are not very sensitive to the presence of CO2 since replacement of saline water by CO2 does not change the S-wave velocity very much.
Specific comments
Comment 1: P1 L14-16. The sentence should be rephrased. Seismic waves cannot provide geological information. Their interpretation can provide structural information about the geological medium. What does "geological information complement active data" mean? What is low low-frequency image? Maybe "image, obtained by analysing the signal of low frequency"? "The passive data are consistent with active data" Better to rephrase: "Results obtained by analysing controlled source seismic data and passive seismic data are in a good correlation".
Response 1: Thanks for your comments and good suggestion. We will rephrase these sentences.
Comment 2: P1 L20 "evaluating" sounds strange. "for structural studies of near surface " sounds better.
Response 2: Thanks. We have corrected it following your suggestion.
Comment 3: P2 L55 "limited source energy" This is a poorly justified statement. It is possible to use an artificial source of high enough energy to study the crust down to dozens of kilometres.
Response 3: Thanks. We meant that our source, a 500 kg weight drop hammer, has limited source energy and penetration depth.
Comment 4: P3 L69 "Ordovician generates are clear reflection". Structure itself cannot generate reflected waves, but it can reflect waves. Better to write: "The top of the Ordovician has strong reflectivity according to results of controlled source seismic data analysis".
Response 4: Thanks. We have changed the wording to better describe the physics.
Comment 5: P3 L71-72 The sentence is unclear. Please rephrase.
Response 5: We have rephrased the sentence with “The top of the Ordovician is highly reflectivite according to results from the active seismic data analysis”.
Comment 6: P5 Table 1. What do the shot interval and the number of source points in passive seismic data mean?
Response 6: It means the spacing interval and number of retrieved shots.
Comment 7: P 7 "4.1. Calculating and picking of dispersion curves" better to write "Dispersion curve extraction"
Response 7: Thanks. "Dispersion curve extraction" is much better.
Comment 8: P 8 "4.2. Inversion for the velocity model" better to write "Dispersion curve inversion".
Response 8: Thanks. We rewrote it with "Dispersion curve inversion".
Comment 9: P 10 Figure 4. Distances instead of channels are more useful. Arrivals of all phases and apparent velocities should be added to the plot.
Response 9: Figure 4 will be reproduced and arrivals of all phases and apparent velocities will also be added to the plot.
Comment 10: P10 Figure 4 caption. "noise reduction" sounds better than "noise attenuation"
Response 10: Yes, "noise reduction" sounds better.
Comment 11: P 11 "7.1. Shear wave velocity model" sounds better.
Response 11: Thanks. We have corrected it.
Comment 12: P 11 What initial model (parameters of the medium) did you use for dispersion curve inversion? Better to provide this as a table. More details are needed about the inversion process.
Response 12: The parameters and methods of data processing and inversion will be added to the manuscript.
Comment 13: P13 There is poor correlation between sections obtained by analysing passive and controlled source seismic data. Only one boundary is visible on both sections. I guess this is because of the absence of reflected waves on the ambient noise. The authors did not prove that they extracted body waves. I would recommend conducting polarisation analysis of the extracted signals.
Response 13: We cannot expect that the passive seismic reflection section provides the same result as active data. They have good correlations at R1, R2, R3 and R4 in Figure 7 and Figure 8. Currently, we cannot provide polarisation analysis of the extracted signals because only 1C sensors were used to collect data in our study.
Comment 14: P14 L236-237 "Particularly, body waves and surface waves can be separated from each other at a frequency of 7 Hz." How do you separate these waves?
Response 14: Thanks. There is a mistakes in the sentence. It is 6 Hz, not 7Hz. Surface waves dominate below 6 Hz (Figure 2b, 2c and Figure 9), so we separated them through different frequency bands (Table 2).
Comment 15: P15 L246. "maximum values of the beam forming analysis in…". Do you mean "maximum beam power"?
Response 15: Yes, it means maximum beam power.
Comment 16: P 15 Figure 10. The sources' azimuths are mainly near-perpendicular to the profile in subplot b. Why are such differences in velocities observed? Have you corrected the phase velocities, taking into account this distribution? This also shows that formula (1) cannot be used to evaluate empirical Green's functions.
Response 16: For surface wave frequency, the sources' azimuth (Figure 10a) is mainly along NNW, consistent with the seismic survey line and located in stationary phase zone. So we did not correct the phase velocities.
Comment 17: P 15 L260-261. "No clear top of basement reflection is observed in the active data, but this may be due to the higher frequency nature of these data." Why? How does reflectivity correlate with the frequency of the signal?
Response 17: Reflectivity does not correlated with the frequency of the signal. We intended to state that “No clear top of basement reflection is observed in the active data, but this may be due to the lack of lower frequencies of these data and the power limitation of weight drop hammer”. If the transition to the basement is gradual then lower frequency signals may reflect whereas higher frequency ones will not.
Citation: https://doi.org/10.5194/egusphere-2025-1325-AC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 454 | 53 | 23 | 530 | 20 | 40 |
- HTML: 454
- PDF: 53
- XML: 23
- Total: 530
- BibTeX: 20
- EndNote: 40
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The manuscript used seismic interferometry to analyze 14 hours of passive seismic data in the Sudret area of Gotland, Sweden, to obtain a shear-wave velocity profile helpful in validating the storage of CO2. The purpose, procedures, results, discussion, and conclusions of this study are generally well written and can be published after checking a few minor points.
Minor comments:
Figure 1. Add labels for Figure 1(b) (e.g., x (m) and y(m)). If possible, a lighter background color will make the symbols more straightforward to distinguish from the background.
Line 90. Since the operating period is 14 hours, it is better to state the exact start time and end time of the data.
Line 122. Describe the strengths and weaknesses of two methodologies and why two approaches were conducted for the surface and body wave extraction, with appropriate references.
Line 208. The figure citation here may be Figure 7, not Figure 5. I wonder why the numbers 150, 280, and 360 ms here do not match 170, 280, and 360 ms in line 185 for the synthetic seismogram section.
Figures 7 and 8: To contrast the background seismic images, it would be good to use a color other than green for the fonts.
Line 236: It would be good to have references for the relevance of 1, 5, 12, and 16 Hz originating from ocean waves and human activities. Can we assign the four frequencies to either ocean waves or human activities?
Line 278: Why does the geometric location of sensors at the southern tip of an island allow noise from nearly all directions?