Spreading of Localized Information across an Entire 3D Electrical Resistivity Volume via Constrained EMI Inversion Based on a Realistic Prior Distribution
Round 1
Reviewer 1 Report (Previous Reviewer 3)
The revised manuscript has been somewhat improved, but, unfortunately, the authors did not follow my main recommendation to get assistance of a geophysical (geoelectrical) professional, whose mother tongue is English. Since grammatically and even semantically the manuscript is generally ok, the assistance of non-geophysical language editor could be insufficient. For example, the authors are widely using words "map" and "mapping" not in the common geophysical sense, but rather as something like "converting" ("mapping the measurements into physical properties", "forward modeling maps the distribution of physical properties", etc). This could be ok with an English editor, but it is very confusing for a geophysical reader. There are many similar examples, which make reading the manuscript a challenge for geophysical professionals.
The situation even more severe with the long overloaded and hardly readable sentences. I pointed out on some of them as examples and the authors indeed changed them accordingly. But I specified that the problem exists throughout the whole manuscript and the authors are required to handle it accordingly everywhere.
And the last, but not least problem I have with the author's replies to my comments. It is perfectly ok, if the authors disagree with some comments and substantiate their disagreement in the report of the replies. But I do not understand the situation, when the authors agree with the comment, specify that the appropriate changes have been done, but the revised manuscript does not include these changes (e.g. the reply 53 to my comment recommending to include the forward modeling response in addition to the model and the inversion results). Again, it would be fine, if the authors have not accepted my recommendation and specified their reasons in the reply. Then, I could, probably, convince them that I recommend the most common presentation of the ERT results that gives some idea of how the measurements would look for the considered model and thus of how much the inversion improves the measurements.
I do estimate the originality and the significance of the content as very high. The rest of my estimates could be biased by the above described unclear presentation.
The quality of English Language in geophysical/geoelectrical terms is fairly low as it is hardly understandable and sometimes confusing. An extensive editing involving an appropriate geophysical (geoelectrical) professional is required.
Author Response
Please, see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report (New Reviewer)
The authors should make minor revisions to the manuscript.
Comments for author File: Comments.pdf
Author Response
Please, see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report (New Reviewer)
Excellent work. This is one of the best papers that I have reviewed.
Author Response
Please, see the attachment.
Author Response File: Author Response.pdf
This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.
Round 1
Reviewer 1 Report
Review on „From a single acquisition line to a 3D electrical resistivity distribution: How to constrain the inversion (based on realistic prior) of dense EMI measurements via pre-existing information (e.g., ERT results).”
The authors presented an interesting manuscript that aimed to demonstrate, how the EMI inversion could be enhanced by incorporating additional information, a reference model or an ERT section along a line across the survey area. Instead of using lateral constraints in a deterministic inversion, they applied realistic prior models and a minimum search approach to derive EMI inversion results. Finally, they tried to verify their results against ground penetrating radar (GPR) measurements and borehole data.
General comments:
Although the approach presented may be promising, the results, however, are not fully convincing. I am missing a comparison with standard EMI inversion results as well as a more detailed discussion of the pro and cons.
Title:
Shorter without brackets and “e.g.”? It should express the main theme of the paper:
a) Starting with a reliable ERT section (or model) and spreading this information via spatial, but improper EMI inversion models (focus on relative EMI variations),
b) Adjusting the EMI inversion models within the entire survey area using constraints derived from a single ERT section or model (focus on absolute EMI inversion models).
Style:
Writing is not always straightforward due to the use of too many nested sentences and filler words (e.g. lines 117-121).
Structure:
All information on field data should be moved to “Materials” (e.g. lines 228-231, 309-323,330-333) as well as the references for GPR and borehole data. The “Discussion” is more a “Conclusion” and several sentences comprising additional information should be moves from “Methods” to “Discussion” (e.g. lines 190-197, 203-225). Split "Discussion" to "Discussion" and "Conclusions".
Figures:
Most of the figures are too complex with hardly readable details and numbers. The resistivity colour bars should be the same in all figures.
Data:
Use of confidential EMI data – is that acceptable by Remote Sensing / MDPI?
References:
Some of the references should be checked, because they do not refer to ground EMI papers, but to airborne TEM (and others) papers (e.g. [6], [8], [10], [14], [16], [17], [29], [30]. That would be OK if there were no adequate EMI (or airborne FEM) papers available, but then that should be mentioned.
Specific comments:
P2, L95: Add information that GPR and borehole data will be used for validation.
P3, L103: Introduce here all the own field measurements and external data used.
P3. L112, Figure 1: blue/greens dots are hardly distinguishable.
P3, L122f: “making the data misfit smaller than the expected noise level ? in the observations” – is that meaningful?
P4. L140, Figure 2: black characters on dark blue background are hardly distinguishable. Add “F”, “s(m)” and “mappr”?
P4, L154: Explain “20 x 106”.
P4, L155: Which (kind of) model is used?
P4, L157: Why 30 m and not e.g. 15 m?
P4, L159: “from 3 to”?
P4, L161: Add depth of layer 17?
P5, L180, Figure 3: Is (a) used as mappr and (b) as prior P for EMI inversion? Add explanation in L185ff.
P5, L188: Define “F” (fomula or at least reference).
P6, L245: Add information on random noise (absolute or relative, value).
P6, L247: Unconstrained inversion = Figure 2b without Mc and F(P)? Explanation required.
P6, L251ff: Add explanation why the conductivity structure in Figures 4b/5c (conductor above resistor) is opposite to the model resistivity structure. How would standard deterministic inversion models look like?
P7, L254, Figure 4: Necessary?
P8, L264, Figure 5: Data and error bars a hardly distinguishable in (b), (d), (f) – far too small. Why are the unconstrained EMI inversion models (11 + 16) next to the pipe in (c) affected by the pipe? Explanation required.
P11, L368ff: The lithology of the borehole in Fig. 10 ranges from fine sand to gravels/stones, which normally cannot explain the different resistivities found in the inversion models. Is anything known on groundwater mineralization? Without that information, the lithology logs are not really helpful. It is further irritating that the inversion results close to the borehole differ enormously. So, the comparison with GPR and borehole results are not very helpful. (Same is valid for Fig. 11).
P14, L403ff: The demonstration of the effect of different noise levels with the help of field data is not very helpful in this case – Fig. 12 should be skipped. Instead, the effect of the EMI noise level (ranging from noise-free up to twice of a realistic maximal noise level of the EMI field data) should be studied based on synthetic data or small portion of field data.
P14, L407ff: The Discussion should include also the pro and cons of the approach presented in comparison with other approaches (standard EMI inversion, spatially constrained EMI inversion, …) as well as the dependency of the final inversion results on the prior P used (what happens, if the model used for creating prior P differs from the true geology and/or groundwater mineralization?).
P15, L438ff, Figures 10-12: Are all the details necessary? For the comparison with boreholes, only a few nearby inversion models are required, which could be displayed in enlarged form (and also the data/error bars). The comparison with the GPR section is questionable, because no interpretation is added to the GPR section.
Author Response
Please, see the attachment.
Author Response File: Author Response.pdf
Reviewer 2 Report
To the Authors:
Thank you for the opportunity to review your submission. Your paper is an interesting application of ERT and EMI data that illustrates how different geophysical methods are sensitive to subsurface properties. I'm recommending that the paper be reconsidered for publication after major revision due to the extensive use of indirect language and run-on sentences in the manuscript. Furthermore, the introduction could use more information on the background of forward and inverse modeling and may benefit overall from less technical details.
I'm happy to openly discuss any of these comments/suggestions in the interest of stewarding an open review of the paper.
Sincerely,
Reviewer
Line 32: “Among them, Electrical Resistivity Tomography (ERT) and Electromagnetic Induction (EMI) methods are amongst the most commonly used to retrieve the medium electrical characteristics - mainly, the electrical resistivity.”
Perhaps this is more related to the field of operation rather than a general statement? The reviewer suggests rewording to be more direct, i.e., ERT and EMI can be used to map subsurface electrical resistivity.
Line 37 and 39: The forward model plays an important part in the inversion process, but there are other critical steps in inverting geophysical data that should not be trivialized. The reviewer suggests rewording these sentences to place more emphasis on inverse modeling and less on forward modeling, since inversion of the geophysical data is the topic.
Line 41: Change ERT’s to ERT surveys. Also maybe point out that time-lapse acquisition is much more reliable for ERT vs. EMI.
Lines 35-51: The reviewer suggests making these three separate paragraphs into one since they all discuss the advantages and disadvantages of EMI and ERT methods.
Figure 2: What is the significance of the yellow and blue circles. These are not described in the caption. The reviewer is assuming they represent data space and model space.
General Note 1: The reviewer suggests rewriting the introduction using more active and direct writing. The usage of parenthetical statements and indirect statements separated by commas makes it difficult to read. Also, the reviewer found the introduction far too technically abrupt, adding some more guiding topic sentences to the paragraphs, and perhaps a more thorough leading paragraph could help to outline the ideas presented in the introduction section. +Additionally, the use of generalized terms like ‘poor’ should be avoided; there is no reason why something more quantifiable cannot be used. Furthermore, the reviewer suggests the addition of a paragraph that introduces forward and inverse modeling.
General Note 2: Avoid the use of terms like ‘clearly’ and ‘in a natural way’ as it is not always clear to the reader or perceived what exactly it is that the author is referring to. In general, the use of these generalized adjectives throughout the paper should be removed (severe)
General Note 3: The goal of the research is to include data in the inversion? Seems trivial as this has been done for years now.
General Note 4: It’s not clear to the reviewer how this is different from treating other geophysical data as a reference model in the inversion, which is not a novel practice. In fact, the reviewer is of the impression that this is an outdated approach, as many practitioners have advanced to joint inversion techniques. It’s not clear how this approach is novel, how it compares to joint inversion of the data, or what information is gained by having EMI and ERT data.
Author Response
Please, see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Dear authors,
Your manuscript describes a novel and to my mind promising approach to increase the reliability of geophysical interpretation by a combined use of different methods mutually complementing each other. Contrary to the conventional joint inversion, your method does not require the acquisition in all methods to be carried out at all points. Such an approach is particularly feasible in case of 3D surveys , in which only fast collecting data methods, such as EMI, RMT/CSRMT, etc are applied on a 3D grid, while time consuming measurements (e.g. ERT, TDEM, etc) can be used in 2D or 1D acquisition schemes.
Also using independent information for regularization of the inversion instead of conventional oversimplified smoothness constraints makes the described overall interpretation being apparently most reliable with regard to the incurred costs.
These very strong points of your work would be, probably, sufficient for an internal technical report describing your method. However, it is not the case for a scientific publication intended for a wide audience. Unfortunately, your manuscript consists of so many unclear descriptions and, in general, is written so heavily for understanding, that I am afraid it will not be accepted by most of the readers and, particularly, by geophysical practitioners.
Attached please find your marked manuscript consisting of my numerous comments, most, if not all of which, concern with your unclear descriptions rather than with the geophysical content. Unfortunately, I was unable to provide reasonable corrections both because in many cases I was not sure I understood your point properly and also because it would mean to actually rewrite the entire manuscript.
Since your English is sufficiently good, I would not recommend to be assisted by a professional translator. I would strongly recommend you to attract a geophysical (more precisely geoeletrical/EM) professional natively speaking English for preparing the revised version of your manuscript.
Comments for author File: Comments.pdf
Author Response
Please, see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Review on „Migration of localized information across an entire 3D electrical resistivity volume via constrained EMI inversion based on a realistic prior distribution” – Revised version.
The authors presented an revised version of an interesting manuscript that aimed to demonstrate, how the EMI inversion could enhanced by incorporating additional information, a reference model or an ERT section along a line across the survey area. Instead of using lateral constraints in a deterministic inversion, they applied realistic prior models and a minimum search approach to derive EMI inversion results. Finally, they tried to verify their results against ground penetrating radar (GPR) measurements and borehole data.
General comments:
Although the revision of the first part of manuscript was good (except the missing “Materials” section), the second part (“Results” – particularly the application to field data – and “Conclusions”) is still not acceptable for publication without a further revision.
In the following, I will respond to the authors’ point-by-point response to Reviewer no. 1:
Points 1-3: OK.
Point 4: Not OK.
All data (incl. information on survey area) used within the manuscript should be introduced and described in a “Materials” section. The “Conclusions” presented are more a summary of the content than concluding remarks.
Point 5: Not OK.
The fonts are OK now, but I do not understand the authors’ response on the color bars: 3163 and 10000 Ωm are too close to be able to highlight the features of the different models/results in EMI studies.
Point 6: Remote Sensing/MDPI should decide.
Point 7: (OK).
I am not able (and willing) to check all the references in detail, but I am wondering whether e.g. [11-12] are really the best publications describing systems. Are all the (numerous) self-citations necessary?
Point 8: OK.
Point 9: Not OK.
I recommend describing all data used in a “Materials” section. The authors referred to the data before they introduced it, e.g. Figure 2, survey area, ERT results, EMI survey, boreholes, GPR in “Methods” section. These data sets must be described in detail as well as the geological set-up of the survey area. The editor should decide on this point.
Points 10-18: OK.
Point 19: Not OK.
I do not understand why the authors did not present the formula of the forward operator. It is just one line and some explanation (for further details, the references are OK).
Point 20: (OK).
Why did you choose 5%? It that the noise level of the EMI field data? BTW: Add information on EMI field data noise level where you introduced the EMI field data.
The error bars shown in Fig. 5 b,d,f seem to be much larger than 5-10 % - explanation necessary!
Point 21: (OK).
It would be very helpful for the reader if a sketch of the case “unconstrained inversion” is added to Fig. 2 (as case a). Then, the new sentence with added further information can be easily replaced by “(Figure 2a)”.
Point 22: Not OK.
I still do not understand how an unconstrained inversion of (noisy) synthetic data derived for a model with a resistive cover above a medium resistive half-space can produce more or less the opposite. At least far from the pipe (3D effect), I would expect inversion models that show the principal structure of the original model (if the relative error level is far less than the data level, the EMI forward operator is correct and the prior is reasonable). A detailed explanation is mandatory, why a more or less consistent opposite of the original model was derived.
For the purpose of comparison, I recommend adding conventional (standard deterministic) inversion results – at least for the synthetic test.
Point 23: Not OK.
Only the situation close to the pipe (3D) and far from the pipe (homogeneous) are interesting. The editor should decide on this point.
Point 24: OK.
Point 25 (cf. Point 28): Not OK.
I do not understand the message of the authors concerning this point. Did they really want to state that a) EMI could not discriminate high resistivities at all (why did they then use different upper bounds for the color bars?) and b) the conductive cover prevents any differentiation below (why did they then compare the EMI models with boreholes results?).
What is the reason for the conductive cover in the EMI inversion models? The ERT shows a resistive cover! Furthermore, the thickness of the EMI conductive cover corresponds to the water table, which normally appears as an interface between high above lower resistivities (and not vice versa). BTW: The information on the water tables should be provided in a “Materials” section, not in the “Discussion” section.
A comparison with boreholes is only meaningful near their location in order to demonstrate that “the lithologies found in the boreholes are perfectly compatible with the EMI results”. As I do not see this match (the figures are too small close to the boreholes), I recommend plotting the last 10-15 % of the model profiles and GPR sections shown in Fig. 10&11 besides the boreholes (at a vertical scale used for the boreholes in the upper part of the figure). Those interfaces in the borehole logs, which could have the greatest influence with respect to resistivity changes, should be marked (and transferred to the models/GPR section).
A comparison with GPR sections without an interpretation on top cannot serve as reference. Therefore, the GPR section is not appropriate for a comparison and the first 85-90 % of the profiles are not necessary.
Point 26: Not OK.
If the authors want to demonstrate the effect of low data noise levels, they should do that with the help of synthetic data. That could be a good proof of their approach: The use of noise-free data should result in inversion models close to the original model. Otherwise, there is something wrong.
I recommend skipping Fig. 12. The editor should decide on this point.
Points 27-28: Not OK.
Yes, the manuscript is already long, but it could be shortened, if the figures focus on essential results. Not all of the figures are necessary (a paper in remote sensing is not a report!). On the other hand, an acceptance of the approach presented (not only by the reviewers, but also by the readers) requires a sound comparison with other approved methods and/or results.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 2 Report
To the Authors,
Thank you for spending some time to revise your manuscript. It has been improved significantly, but there are still a couple issues that I feel need to be addressed.
1. Please avoid using the term 'migration' in the title as this has a very specific meaning in the context of geophysics that has nothing to do with EMI or ERT.
2. Please add some references to support your claims that ERT inversions are commonly 2D and 3D and EMI inversions are typically 1D. (lines 35-39)
3. Why are ERT surveys troublesome? Why not be more direct and state what you actually mean? I think what you want to say here is that it takes a considerable amount of time to cover a large area with ERT vs EMI. The fact of the matter is that both methods have their idiosyncrasies, but these are all things that every practitioner considers when selecting a method. (line 40)
4. The sentence on lines 58-60 could be more direct, for example, 'Inversion of the observed data to produce subsurface electrical property distributions requires the use of a forward model as well as an inverse method, e.g., Tikhonov inversion."
5. The reviewer still finds several examples of convoluted sentence structure in the manuscript that results in indirect communication of ideas. For example, on line 67-68 the authors state "However, often, 3D response are much more demanding to be calculated" but the sentence structure if more direct if you reword to "However, calculating the 3D responses are often much more demanding."
Kind Regards
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
