the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 26 Mar 2024
3D Geo-Modeling Framework for Multisource Heterogeneous Data Fusion Based on Multimodal Deep Learning and Multipoint Statistics: A case study in South China Sea
Abstract. Relying on geological data to construct 3D models can provide a more intuitive and easily comprehensible spatial perspective. This process aids in exploring underground spatial structures and geological evolutionary processes, providing essential data and assistance for the exploration of geological resources, energy development, engineering decision-making, and various other applications. As one of the methods for 3D geological modeling, multipoint statistics can effectively describe and reconstruct the intricate geometric shapes of nonlinear geological bodies. However, existing multipoint statistics algorithms still face challenges in efficiently extracting and reconstructing the global spatial distribution characteristics of geological objects. Moreover, they lack a data-driven modeling framework that integrates diverse sources of heterogeneous data. This research introduces a novel approach that combines multipoint statistics with multimodal deep artificial neural networks and constructs the 3D crustal P-wave velocity structure model of the South China Sea by using 44 OBS forward profiles, gravity anomalies, magnetic anomalies and topographic relief data. The experimental results demonstrate that the new approach surpasses multipoint statistics and Kriging interpolation methods, and can generate a more accurate 3D geological model through the integration of multiple geophysical data. Furthermore, the reliability of the 3D crustal P-wave velocity structure model, established using the novel method, was corroborated through visual and statistical analyses. This model intuitively delineates the spatial distribution characteristics of the crustal velocity structure in the South China Sea, thereby offering a foundational data basis for researchers to gain a more comprehensive understanding of the geological evolution process within this region.
This preprint has been withdrawn.
-
Withdrawal notice
This preprint has been withdrawn.
-
Preprint
(4836 KB)
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2024-684', Anonymous Referee #1, 30 Apr 2024
Review comments on “3D Geo-Modeling Framework for Multisource Heterogeneous Data Fusion Based on Multimodal Deep Learning and Multipoint Statistics: A case study in South China Sea” by Dr Liu et al.
This paper proposes a new flowchart to construct a three-dimensional seismic wave velocity model by employing multipoint statistics and multimodal deep learning. The target area is the South China Sea and the authors incorporate existing 2D seismic velocity data from active-source experiments, satellite-based seafloor topography and gravity data and seafloor magnetic data in the region. Overall, this is an interesting paper and will be a good contribution to the related community. However, I have many major concerns and suggestions for improvements. Additionally, I have pointed out several minor text edits below. Therefore, I recommend returning the manuscript to the authors with major revisions.
1. Introduction
The Introduction appears verbose and could benefit from concise writing by avoiding redundant expressions. If needed, consider utilizing professional English proofreading services for assistance.
2. Input data
The authors adequately describe their technical processing framework of deep learning, but there seems to be less emphasis on the input data from geophysical observations. I think it is essential to consider the varying accuracy and uncertainty of seismic velocity models due to differences in data quality, such as OBS spacing during acquisition. Do you treat all seismic data equally in deep learning, or do you assign different weights? Additionally, while magnetic anomaly data reflects tectonic plate age, its relevance to seismic velocity below the seafloor is unclear. So, I was not sure why the magnetic data are included and how they contribute. Moreover, without information on the derivation of the Moho Surface Model (Fig. 2d), it is challenging to assess its reliability. Please address these issues for clarification.
3. Model reliability
Associated with the comment above, I am concerned about the reliability of the obtained 3D velocity model. To strengthen its credibility, I suggest incorporating more cross checks. For instance, have you considered comparing your results with independent data not used in your scheme, such as MCS reflection studies that constrain the thickness of shallow sedimentary layers? This comparison could provide valuable validation and enhance the robustness of your findings.
4. Terminology
The authors uses several seismological and geological terms that may be unclear or unnatural. For instance, terms like “OBS forward profiles” (Line 23) and “forward-simulated velocity structure profiles” (Line 152) are not commonly used in seismic refraction studies and may confuse readers. Similarly, the term “geological attributes” (Line 147 and elsewhere) lacks clarity in its meaning. Additionally, phrases like “artificial artifacts” (Line 291 and elsewhere) seems awkward and could be improved for better readability. To address these concerns and other problematic phrases (as noted in the minor comments below), I recommend having the manuscript carefully reviewed by English-native colleagues with expertise in seismology or using external proofreading services for clarity and accuracy.
5. Figures
The figures in the manuscript appear too small and need to be displayed more effectively. Many letters and labels in Figures 2, 6, 8, 9, 10, 11, 12, 13, 14 and 15 are difficult to read. The 3D view of the map figures (Figures 6, 10 and 15) is not effective, and I suggest presenting them in a 2D map view and enlarge them to improve readability. Additionally, some figure captions lack sufficient information. For instance, in Figure 11, clarification is needed on how the “Residual ratio” is defined. Furthermore, it would be beneficial to add iso-velocity contours to panels (a), (c), (e) and (g) of Figure 11.
Minor comments:
Table 1
Hou et al., 2019 is missing in the reference list. For OBS2019ZX1 and OBS2020-1, if you refer to the unpublished data, you need to provide more detailed descriptions on the velocity information.
Line 163-164: Sandwell et al. 2019 is missing in the reference list.
Line 183: Arpat and Caers, 2007 is missing in the reference list.
Line 187 Strebelle, 2002 is missing in the reference list.
Line 194: de Vris et al., 2009 is missing in the reference list.
Line 196: Honarkhah and Caers, 2012 is missing in the reference list.
Line 196: Straubhaar et al., 2021 > Straubhaar and Renard, 2021
Line 203: Minar and Naher, 2018 is missing in the reference list.
Line 205: “uesd” > “used”
Line 205: “Hinton (2006)” > “Hinton and Salakhutdinov (2006)”
Line 227: Zhang et al., 2021 is missing in the reference list.
Line 324: “constructa” > “construct”
Line 355: What is the unit of “70 x 400 x 400”?
Line 358: “-35km” should be replaced by “35km”.
Line 367: Where is “the SCS basin”? Please indicate its location and extent in map figures.
Line 368: Why are they so extremely thin? Are there any evidence for this from seismic refraction data? I would like to expect more discussion on these structural features from the view point of tectonics because Solid Earth is a geoscience journal rather than computer science.
Line 375: What is the definition of “HVL”?
Line 381 “similiar” > “similar”
Line 383: “17.94m” > “17.94km”
Line 468: “geologicial” > “geological”
Line 491: “hetereogeneous” > “heterogeneous”
Line 500: Chiles et al., 2012 and Pyrcz et al., 2014 are missing in the reference list.
Line 557: “imporve” > “improve”
Line 622: Please complete the reference information on Arpat (2005) although this is not cited in the text.
Line 673-674: Honarkhah 2011 is not cited in the text.
Line 681-691: There are two “Hou et al. (2023)” listed here. I was not sure which is cited in the text at each location.
Line 786-788: Sandwell et al. 2021 is not cited in the text.
Line 845-852: Xia et al. (2018) is repeated twice.
Citation: https://doi.org/10.5194/egusphere-2024-684-RC1 -
AC1: 'Comment on egusphere-2024-684', Henggaung Liu, 31 Jul 2024
Dear editor team, reviewers and readers, firstly, on behalf of the entire team, I would like to express heartfelt thanks for your hard work and dedication during the processing of the manuscript. The meticulous and professional guidance of the editor team, the valuable feedback and selfless dedication of the reviewers, as well as the attention and support from every reader, are indispensable forces for the improvement and perfection of this article. Here we pay high tribute to you all.
However, due to some uncontrollable factors, we regret to notify you that this manuscript needs to be withdrawn. After in-depth evaluation, it is found that the review cycle has been extended because a suitable second reviewer cannot be found for a long time, which has seriously affected our subsequent work arrangement. We are fully aware that this may cause inconvenience and trouble, especially for those reviewers who have invested time and efforts. We are deeply apologize for this.
Nonetheless, we are still grateful to every reviewer, reader and editor team. We will learn from this experience, strive to improve and optimize our work . We sincerely wish you all smooth work and pleasant life!
Citation: https://doi.org/10.5194/egusphere-2024-684-AC1
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2024-684', Anonymous Referee #1, 30 Apr 2024
Review comments on “3D Geo-Modeling Framework for Multisource Heterogeneous Data Fusion Based on Multimodal Deep Learning and Multipoint Statistics: A case study in South China Sea” by Dr Liu et al.
This paper proposes a new flowchart to construct a three-dimensional seismic wave velocity model by employing multipoint statistics and multimodal deep learning. The target area is the South China Sea and the authors incorporate existing 2D seismic velocity data from active-source experiments, satellite-based seafloor topography and gravity data and seafloor magnetic data in the region. Overall, this is an interesting paper and will be a good contribution to the related community. However, I have many major concerns and suggestions for improvements. Additionally, I have pointed out several minor text edits below. Therefore, I recommend returning the manuscript to the authors with major revisions.
1. Introduction
The Introduction appears verbose and could benefit from concise writing by avoiding redundant expressions. If needed, consider utilizing professional English proofreading services for assistance.
2. Input data
The authors adequately describe their technical processing framework of deep learning, but there seems to be less emphasis on the input data from geophysical observations. I think it is essential to consider the varying accuracy and uncertainty of seismic velocity models due to differences in data quality, such as OBS spacing during acquisition. Do you treat all seismic data equally in deep learning, or do you assign different weights? Additionally, while magnetic anomaly data reflects tectonic plate age, its relevance to seismic velocity below the seafloor is unclear. So, I was not sure why the magnetic data are included and how they contribute. Moreover, without information on the derivation of the Moho Surface Model (Fig. 2d), it is challenging to assess its reliability. Please address these issues for clarification.
3. Model reliability
Associated with the comment above, I am concerned about the reliability of the obtained 3D velocity model. To strengthen its credibility, I suggest incorporating more cross checks. For instance, have you considered comparing your results with independent data not used in your scheme, such as MCS reflection studies that constrain the thickness of shallow sedimentary layers? This comparison could provide valuable validation and enhance the robustness of your findings.
4. Terminology
The authors uses several seismological and geological terms that may be unclear or unnatural. For instance, terms like “OBS forward profiles” (Line 23) and “forward-simulated velocity structure profiles” (Line 152) are not commonly used in seismic refraction studies and may confuse readers. Similarly, the term “geological attributes” (Line 147 and elsewhere) lacks clarity in its meaning. Additionally, phrases like “artificial artifacts” (Line 291 and elsewhere) seems awkward and could be improved for better readability. To address these concerns and other problematic phrases (as noted in the minor comments below), I recommend having the manuscript carefully reviewed by English-native colleagues with expertise in seismology or using external proofreading services for clarity and accuracy.
5. Figures
The figures in the manuscript appear too small and need to be displayed more effectively. Many letters and labels in Figures 2, 6, 8, 9, 10, 11, 12, 13, 14 and 15 are difficult to read. The 3D view of the map figures (Figures 6, 10 and 15) is not effective, and I suggest presenting them in a 2D map view and enlarge them to improve readability. Additionally, some figure captions lack sufficient information. For instance, in Figure 11, clarification is needed on how the “Residual ratio” is defined. Furthermore, it would be beneficial to add iso-velocity contours to panels (a), (c), (e) and (g) of Figure 11.
Minor comments:
Table 1
Hou et al., 2019 is missing in the reference list. For OBS2019ZX1 and OBS2020-1, if you refer to the unpublished data, you need to provide more detailed descriptions on the velocity information.
Line 163-164: Sandwell et al. 2019 is missing in the reference list.
Line 183: Arpat and Caers, 2007 is missing in the reference list.
Line 187 Strebelle, 2002 is missing in the reference list.
Line 194: de Vris et al., 2009 is missing in the reference list.
Line 196: Honarkhah and Caers, 2012 is missing in the reference list.
Line 196: Straubhaar et al., 2021 > Straubhaar and Renard, 2021
Line 203: Minar and Naher, 2018 is missing in the reference list.
Line 205: “uesd” > “used”
Line 205: “Hinton (2006)” > “Hinton and Salakhutdinov (2006)”
Line 227: Zhang et al., 2021 is missing in the reference list.
Line 324: “constructa” > “construct”
Line 355: What is the unit of “70 x 400 x 400”?
Line 358: “-35km” should be replaced by “35km”.
Line 367: Where is “the SCS basin”? Please indicate its location and extent in map figures.
Line 368: Why are they so extremely thin? Are there any evidence for this from seismic refraction data? I would like to expect more discussion on these structural features from the view point of tectonics because Solid Earth is a geoscience journal rather than computer science.
Line 375: What is the definition of “HVL”?
Line 381 “similiar” > “similar”
Line 383: “17.94m” > “17.94km”
Line 468: “geologicial” > “geological”
Line 491: “hetereogeneous” > “heterogeneous”
Line 500: Chiles et al., 2012 and Pyrcz et al., 2014 are missing in the reference list.
Line 557: “imporve” > “improve”
Line 622: Please complete the reference information on Arpat (2005) although this is not cited in the text.
Line 673-674: Honarkhah 2011 is not cited in the text.
Line 681-691: There are two “Hou et al. (2023)” listed here. I was not sure which is cited in the text at each location.
Line 786-788: Sandwell et al. 2021 is not cited in the text.
Line 845-852: Xia et al. (2018) is repeated twice.
Citation: https://doi.org/10.5194/egusphere-2024-684-RC1 -
AC1: 'Comment on egusphere-2024-684', Henggaung Liu, 31 Jul 2024
Dear editor team, reviewers and readers, firstly, on behalf of the entire team, I would like to express heartfelt thanks for your hard work and dedication during the processing of the manuscript. The meticulous and professional guidance of the editor team, the valuable feedback and selfless dedication of the reviewers, as well as the attention and support from every reader, are indispensable forces for the improvement and perfection of this article. Here we pay high tribute to you all.
However, due to some uncontrollable factors, we regret to notify you that this manuscript needs to be withdrawn. After in-depth evaluation, it is found that the review cycle has been extended because a suitable second reviewer cannot be found for a long time, which has seriously affected our subsequent work arrangement. We are fully aware that this may cause inconvenience and trouble, especially for those reviewers who have invested time and efforts. We are deeply apologize for this.
Nonetheless, we are still grateful to every reviewer, reader and editor team. We will learn from this experience, strive to improve and optimize our work . We sincerely wish you all smooth work and pleasant life!
Citation: https://doi.org/10.5194/egusphere-2024-684-AC1
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 524 | 181 | 20 | 725 | 24 | 19 |
- HTML: 524
- PDF: 181
- XML: 20
- Total: 725
- BibTeX: 24
- EndNote: 19
Viewed (geographical distribution)
| Country | # | Views | % |
|---|---|---|---|
| United States of America | 1 | 277 | 38 |
| China | 2 | 196 | 27 |
| Germany | 3 | 44 | 6 |
| France | 4 | 18 | 2 |
| undefined | 5 | 17 | 2 |
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
- 277
