the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 01 Nov 2022
Induced Electromagnetic prospecting for the characterization of the European southernmost glacier: the Calderone Glacier, Apennines, Italy
Abstract. The increasing rate of glacier retreat in recent decades is well documented and represents a great loss for the paleoclimate studies. In this framework, Ice Memory project aims to extract and analyze ice cores from worldwide glacier regions and then storage them in Antarctica as heritage for future generations. Ice coring projects usually require a focused geophysical investigation, often based on the Ground Penetrating Radar (GPR) technique and the active seismic prospection, in order to assess the most suitable drilling positions. As novelty, in the Calderone Glacier, we integrated the GPR results with a Frequency Domain Electro-Magnetic (FDEM) prospection which is not commonly applied in the glacial environment. A separated-coils FDEM instrument has been used to characterize the glacier up to several tens of meters of depth. The acquired FDEM datasets were inverted and compared to the GPR data and borehole information. The results demonstrate the ability of the FDEM instrument to correctly define the structure of the glacier and therefore its potential to be applied in frozen subsoils studies. All this opens new perspectives for the use of FDEM technique to characterize glacial or periglacial environments as rock glaciers, where the GPR acquisition logistic is limited by the rock blocky surface and affected by the scattering from surface debris.
- Preprint
(2199 KB) - Metadata XML
-
Supplement
(6311 KB) - BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on tc-2022-190', Anonymous Referee #1, 22 Nov 2022
Review comments for “Induced Electromagnetic prospecting for the characterization of the European southernmost glacier: the Calderone Glacier, Apennines, Italy” by Pavoni et al. manuscript 2022-190 for TC
Summary:
This paper has clear motivations and objectives. It is refreshing to see multiple geophysical methods being used in a glaciated environment. However, there are some major comments which should be addressed before considered for publication.
Major comments:
- This paper would greatly benefit from including the results and a log from the borehole observations mentioned in Line xx. The authors repeatability refer to the validation of the geophysical methods by borehole observations yet do not include the borehole results in the study. There is no location on the base map in Figure 1 of the borehole location. The paper would be much more complete with the addition of this observation.
- It is commonly known that electromagnetic methods are highly non-unique. Why do the authors not constrain the FDEM inversion with the GPR depths (snow and ice) and a range of expected conductivities?
- What is the uncertainty in the pseudo 2D FDEM inversion? Are there areas where the uncertainty is larger, for example in the ice, or at depth? A cross-section showing the uncertainty range/standard deviation would be useful to enable reliable assessment of the 2D section.
- I would consider changing the title. The words “prospecting for the characterization” do not work well together and FDEM should be in the title as this is the more novel technique used in this glaciated environment. I would suggest something along the lines of: “Characterisation of the Calderone Glacier, Apennines, Italy using GPR, FDEM and borehole observations”.
- The main text needs to be proof read thoroughly.
Minor comments:
- Figure A3 in the appendix, should be in the main text the depth of investigation is important in understanding the limitations of the electromagnetic inversion results.
- In the inverted FDEM sections, can you explain in more detail what is going on under the ice? The values seam to be more conductive than if it was bedrock.
Specific comments:
L98-99: did you get any CMP gathers to estimate the velocity of the snow and ice layer?
L130: “dozens” should be changed to “tens”
L134-136: the last sentence in this paragraph doesn’t make sense.
L194-196: Need to explain this sentence in more detail. What are the instrumental resolution limits?
L201-202: where did you measure the several meters of snow cover during the data acquisition? I would mark these locations on the base map in Figure 1. Was this by a snow pit?
I would change the use of the word “subsoil” to “subsurface” as you are not working in a soil environment.
L221-222 This sentence doesn’t make sense.
L226: “exciting” ? Do you mean “existing”?
L237-8: The ice ends at “x=30m”, where did it end before and what is this value relative to. The retreat of the glacier should be discussed in terms of “The ice mass has retreated xx m since 2015”.
L247-8: this borehole observation should be added to this paper.
L255: where is this measurement, show on Figure 1.
L263-267: This last paragraph is important for the FDEM inversion. I would have this in the main text before the inversion results, including figure A3.
L269: The paper needs to back up this concluding sentence with the borehole observations and a more detailed map in Figure 1 of where the snow depth measurements were acquired.
L274-275: do you mean time-laspe geophysical surveys?
L292: Repetition of the drilling/borehole observations with no results detailed in the paper.
L294-297: this should be in the discussion.
Table 1: could be moved to the appendix.
Figure 2: Nice figure. what is the measured apparent conductivity used to create these depth ranges? Do you not have to input a conductivity to estimate the depth of each coil separation?
Table 2: This is just a personnel preference, however many recent papers using electromagnetic methods in a glaciological environment work in terms of resistivity, in ohm.m, the inverse of conductivity. It might be easier for your readers to follow and directly compare with other studies if you worked in terms of resistivity.
Figure 5: This is also another personnel preference, however most recent papers using electromagnetic methods in a glaciological environment have hot colours (like red and purple) representing high resistivity/low conductivity and cold colours (e.g. blue) representing low resistivity/high conductivity. Again, this might make it easier for your readers to follow and compare with other studies.
Figure 8 and 9: Consider having one figure for 8 and one for 9, merging A and B, with the GPR transparently overlaid on the conductivity plot.
Figure A3: Have this in the main text. Add a dotted line at the depth of investigation. To me it looks like you only have sensitivity to 20 m depth as the sensitivity curves come close together after 20 m?
Citation: https://doi.org/10.5194/tc-2022-190-RC1 -
AC1: 'Reply on RC1', Mirko Pavoni, 15 Dec 2022
We truly thank the anonymous Reviewer for very constructive comments and suggestions about our manuscript.
Reply point by point:
- This paper would greatly benefit from including the results and a log from the borehole observations mentioned in Line xx. The authors repeatability refers to the validation of the geophysical methods by borehole observations yet do not include the borehole results in the study. There is no location on the base map in Figure 1 of the borehole location. The paper would be much more complete with the addition of this observation.
Reply: We fully agree with this comment and we will include in the manuscript the information about the borehole stratigraphy and the position in the map/models.
- It is commonly known that electromagnetic methods are highly non-unique. Why do the authors not constrain the FDEM inversion with the GPR depths (snow and ice) and a range of expected conductivities?
Reply: The Reviewer is right, as the result of any kind of geophysical method inversion process, we know that our FDEM models are a non-unique solution to the problem. Nevertheless, the aim of the FDEM survey on the Calderone Glacier was to verify the capability of the separated-coils FDEM device to characterize an environment with ice fraction occurring in the investigated subsurface. Consequently, it was important for us to evaluate the inversion results without using any prior structural information, as defined by the GPR profiles. In this way, we had the opportunity to understand if it is possible to use the separated-coils FDEM device in other similar environments, as rock glaciers, without the need to combine it with GPR surveys, which are challenging to perform with a coarse-blocky surface and less resolutive since the ice is highly mixed with debris (signal scattering problem). Considering the results obtained on the Calderone Glacier, we think that we achieved our goal. In the summer of 2022 we tested the separated-coils FDEM device in several rock glaciers and achieved very promising results.
- What is the uncertainty in the pseudo-2D FDEM inversion? Are there areas where the uncertainty is larger, for example in the ice, or at depth? A cross-section showing the uncertainty range/standard deviation would be useful to enable a reliable assessment of the 2D section.
Reply: Concerning the sensitivity of the inverted models, the used code EMagPy does not provide a sensitivity 2D section. However, the software provides the percentage value of RRMSE (Relative Root Mean Squared Error) to evaluate the accuracy of the inverted models. Therefore, we will insert this information in the manuscript for both investigation lines. Nevertheless, as we highlighted in the manuscript, the sensitivity of the FDEM measurements is higher in the near subsurface and decreases with the depth till it reaches (approximately) zero at a depth of about 30 meters (see Fig.A3, which will be removed from the Appendix and insert in the main text of the manuscript), consequently, we can consider the uncertainty of the pseudo-2D FDEM models in the same way.
- I would consider changing the title. The words “prospecting for the characterization” do not work well together and FDEM should be in the title as this is the more novel technique used in this glaciated environment. I would suggest something along the lines of: “Characterisation of the Calderone Glacier, Apennines, Italy using GPR, FDEM, and borehole observations”.
Reply: We agree with the Reviewer and this useful suggestion, we will modify the title and we will include “FDEM” inside it, e.g. Combination of Ground Penetrating Radar and Frequency-Domain Electromagnetic methods for the characterization of the Calderone Glacier (Gran Sasso d’Italia, Italy).
Even if the borehole result is important to validate our geophysical results, we don’t think that “borehole” could be appropriate to insert in the title, since only the geophysical surveys were carried out to characterize the structure of the glacier, and this way define the most suitable position to drill and extract the ice core sample, which was the target of the Ice Memory Project team.
- The main text needs to be proof read thoroughly
Reply: The reviewer is right, we will do it.
- Figure A3 in the appendix should be in the main text. The depth of investigation is important in understanding the limitations of the electromagnetic inversion results.
Reply: The suggestion is correct, we can move Figure A3 in the main text, particularly in Chapter 4.2 (FDEM inversion results).
- In the inverted FDEM sections, can you explain in more detail what is going on under the ice? The values seem to be more conductive than if it was bedrock
Reply: According to values that we found in literature, and that we observed ourselves in ERT surveys in the North-Eastern Alps areas, we assigned a representative value of 2E-1 mS/m to the calcareous bedrock. In the inverted and calibrated model Line 1 (Fig.8A), the boundary with the bedrock is well defined and it is very similar to the synthetic model (Fig.7A), suggesting the presence of the bedrock as confirmed also by the GPR profile and the borehole. On the other hand, in the inverted and calibrated model of Line 2 (Fig.9A), it is true that the conductivity values at the bottom of the section are higher than expected for the bedrock, particularly in the western direction. In this area, it is plausible that the ice layer is not directly in contact with the calcareous bedrock but instead with a lateral moraine, which obviously has higher conductivity values than bedrock. We will add these considerations in the main text, in particular in the chapter of Discussion.
- Specific comments:
- L98-99: did you get any CMP gathers to estimate the velocity of the snow and ice layers?
Reply: GSSI Sir4000 is a monostatic digital GPR antenna, therefore it was not possible to acquire CMP gathers.
- L130: “dozens” should be changed to “tens”
Reply: The Reviewer is right, it will be changed.
- L134-136: the last sentence in this paragraph doesn’t make sense: “Despite this, FDEM methods proved to be efficiently applicable in high resistive environments, considering in a relative way the inverted conductibility profile (e.g. Boaga et al. 2020; Pavoni et al. 2021).”
Reply: As the FDEM devices produced by GF Instruments (and other manufacturers) have not the capability to detect conductivity variations lower than 0.1 mS/m (instrumental limit resolution), we cannot pretend to measure the real conductivity values of the ice layers, which have conductivities << 0.1 mS/m. Consequently, in these low conductive environments, FDEM models should not be interpreted on the basis of the inverted conductivity values, but a calibration based on the forward modeling is needed to define a value representative for the frozen layers (e.g. Pavoni et al. 2021 evaluated values < 1 mS/m for the frozen layer in dolomitic rock glaciers, which was confirmed by the comparison with ERT results). We will re-phrase the sentence explaining better this concept in the manuscript.
- L194-196: Need to explain this sentence in more detail. What are the instrumental resolution limits?
Reply: The FDEM devices produced by GF Instruments have not the capability to detect conductivity variations lower than 0.1 mS/m, which is therefore the instrumental limit resolution.
- L201-202: where did you measure the several meters of snow cover during the data acquisition? I would mark these locations on the base map in Figure 1. Was this by a snow pit?
Reply: The Reviewer is absolutely right, we have to show in the map and in the models the point where the snow cover thickness has been measured. Yes, it was a snow pit. For the measurement, we used an extendable rod and the measured point is approximately at x~40 m in Line 1. We will specify this in the text.
- L221-222 This sentence doesn’t make sense: “Calibration intends to explore if exist a constant correction factor to be applied to the inversion results of the field datasets, in order to have the same conductivity scale of the synthetic model.”
Reply: We compared the inverted models obtained from the field datasets (Fig.5 – from now on field dataset inverted models) and the ones obtained from the synthetic datasets (Fig.7 - from now on synthetic models), searching for a correction factor that can be applied to the inversion results of the field datasets and allows to have a model with the same conductivity range as found in the inverted synthetic models (from 0 to 1 mS/m – see Fig.7). The defined correction factor of 1E-2 mS/m has been applied to the inversion results of field datasets (Line 1 and Line 2). In this way, as you can see in Fig.8 and Fig.9 (FDEM inverted and calibrated sections), the range of conductivity in the field dataset models span from 0 to 1 mS/m, and the ice layer boundaries can be defined with the same values found in the synthetic models (0-0.1 for the ice-rich layer and 0.1-0.2 for the ice-debris mixture, see chapter 4.3). We will modify chapter 4.4 and explain better the calibration of the FDEM results.
- L226: “exciting”? Do you mean “existing”?
Reply: Yes, sorry for the typo. We will correct it.
- L237-8: The ice ends at “x=30m”, where did it end before and what is this value relative to. The retreat of the glacier should be discussed in terms of “The ice mass has retreated xx m since 2015”.
Reply: Along the longitudinal Line 1, the ice-rich layer was easily detectable along the entire GPR profile measured in 2015 by Monaco & Scozzafava, but today seems to end at x≈30m. Therefore, in the last 7 years, between x = 0 and x = 30 meters of Line 1, we had (presumably) a loss of massive ice and an increase in the amount of ice-debris mixture.
- L247-8: this borehole observation should be added to this paper.
Reply: Yes, we will do it.
- L255: where is this measurement, show on Figure 1.
Reply: Yes, we will do it.
- L263-267: This last paragraph is important for the FDEM inversion. I would have this in the main text before the inversion results, including figure A3.
Reply: We really appreciate this suggestion. Our proposal is to add at the end of Chapter 3.2.1 (FDEM forward and inverse modeling) this sentence: “to define the bottom depth of the models, sensitivity profiles of the measurements have been calculated with EMagPy. In the current work, the inverted FDEM models are limited to the depths where the normalized sensitivity of the measurements reaches approximately zero.” We can show the sensitivity profiles in Chapter 4.2 (FDEM inversion results) and add this sentence: “the sensitivity of the measurements performed along Line 2 is presented (same results were found for Line 1). It is clear that sensitivities are higher in the near subsurface and decrease to (approximately) zero at a depth of about 30 meters. Consequently, we considered the uncertainty of the pseudo-2D FDEM inverted models in the same way, and we defined the bottom of the FDEM sections at a depth of 30 meters from the surface.” Finally, we can leave the following sentences in the discussion chapter: “It should be noted that both the FDEM inverted models have lower penetration depth (~30 m) than those predicted by the instrument manufacturer (see Fig.2). This is expected since the investigation depth decreases in low electrical conductivity environments (Hauck and Kneisell, 2008).”
- L269: The paper needs to back up this concluding sentence with the borehole observations and a more detailed map in Figure 1 of where the snow depth measurements were acquired.
Reply: Yes sure, we will do it.
- L274-275: do you mean time-lapse geophysical surveys?
Reply: Yes, it was the meaning of the sentence. The Reviewer is right and we can insert directly the reference to the time-lapse surveys (e.g. “A future development for the GPR measurements on the Calderone Glacier is to apply the method proposed by Santin et al. (2022), to estimate the debris content within the layer composed of ice-debris mixture. In case of periodic measurements in time-lapse configuration, this method could help to estimate the ice volume losses of the Calderone Glacier in the next future.”).
- L292: Repetition of the drilling/borehole observations with no results detailed in the paper.
Reply: It will be insert into the manuscript.
- L294-297: this should be in the discussion.
Reply: We respect the opinion of the Reviewer but it is our opinion that these sentences fit better in the conclusions of our work.
- Table 1: could be moved to the appendix.
Reply: We would prefer to leave it inside the main text, but we can move it to the Appendix if required.
- Nice figure. what is the measured apparent conductivity used to create these depth ranges? Do you not have to input a conductivity to estimate the depth of each coil separation?
Reply: Fig. 2 shows the nominal effective depth range influencing the measured apparent conductivities suggested by the manufacturer of the CMD-DUO device, GF Instruments. Instead of showing a Table with these effective depth ranges, we preferred to create a new intuitive figure expressly for this work.
- Table 2: This is just a personnel preference, however many recent papers using electromagnetic methods in a glaciological environment work in terms of resistivity, in ohm*m, the inverse of conductivity. It might be easier for your readers to follow and directly compare with other studies if you worked in terms of resistivity.
Reply: In our experience, we prefer to use conductivity (mS/m) when we work with EMI methods. As the Reviewer correctly mentioned, this is just our personnel preference. To be consistent, in the main text of the manuscript we will remove all the references to electrical resistivity and we will refer only to conductivity.
- Figure 5: This is also another personnel preference, however most recent papers using electromagnetic methods in a glaciological environment have hot colours (like red and purple) representing high resistivity/low conductivity and cold colours (e.g. blue) representing low resistivity/high conductivity. Again, this might make it easier for your readers to follow and compare with other studies.
Reply: As before for conductivity-resistivity, we personally prefer to use a color scale where the frozen/ice layer (with high electrical resistivities – low electrical conductivities) is presented with cold colors as blue, and the unfrozen subsoil/subsurface is defined by hot colors as red. In our opinion, as we verified in various geophysical conferences, it is more intuitive to associate the blue color with the ice, particularly for people that are not usually inside the cryosphere “world”. Again, as the Reviewer correctly mentioned, this is just our personnel preference.
- Figures 8 and 9: Consider having one figure for 8 and one for 9, merging A and B, with the GPR transparently overlaid on the conductivity plot.
Reply: We appreciate the suggestion of the Reviewer but it is our opinion that a comparison of separated models (FDEM and GPR) is the best solution to better appreciate the similarities between them. Moreover, we will insert in this Figure the stratigraphy of the borehole. Once again, this is just our personal preference.
- Figure A3: Have this in the main text. Add a dotted line at the depth of investigation. To me, it looks like you only have sensitivity to 20 m depth as the sensitivity curves come close together after 20 m?
We will move Fig.A3 into the main text, e.g. in Chapter 4.2. (FDEM inversion results). We could add a vertical dashed line that defines the value of zero sensitivity. In this way, it would be easier to appreciate that the sensitivity of the FDEM survey reaches zero at approximately 30 meters of depth. Between 20 and 30 meters of depth, the sensitivity of the measurements is still > 0, so that part of the subsurface is still contributing to the measured apparent conductivities. For this reason, we have limited our FDEM models to 30 meters of depth from the ground level.
Citation: https://doi.org/10.5194/tc-2022-190-AC1
-
RC2: 'Comment on tc-2022-190', Anonymous Referee #2, 20 Dec 2022
The authors present a multidisciplinary study to investigate the structure of a glacier ice body. They apply the methods of ground penetrating radar and induced electromagnetic prospecting. For the latter technique the authors utilize a CMD-DUO system. This system relies on the principle that an electromagnetic field is emitted by a Tx coil. In a conductive subsurface, secondary currents are induced and the superposition of the primary signal from the transmitter and the secondary signal from the induced eddy currenty in the subsurface are recorded in the Rx coil. This principle is inherently limited to applications where the subsurface is sufficiently conductive to induce eddy currents large enough to exceed the detection threshold of the recording system. The manufacturer of the used system gives this value at a limit of 10E-1 mS/cm, two orders of magnitude larger then the expected conductivity of massive ice. The authors have explained these limitations correctly in their manuscript.
Consequently, the recorded data show no evidence of sufficiently large induced signals, as evident from Figure 3. Here, the data scatter around zero, large parts of the data showing physically implausibel negative values. In a next step "Anomalous measurements filtering" is applied to remove outliers larger than two standard deviations. In the figure Figure 3C where polynomial interpolation smoothing is applied the data appear all positive. This cannot be explaind by the presented processing steps.
After applying an inverse modelling, the results do not match plausible values for glacier ice bodies. This can logically explained by improper input data. The authors confirm the unplausible data and apply an empiric shift of two orders of magnitude, justified by the misfit bewteen syntheic data and FDEM models from field data. No physical explanation for this appraoch is presented. This empiric shift of the data is by no means a "calibration", it is rather an adaptation to the expected values.
It is common practice for models obtained by inversion techniques to present i) the full recorded data, ii) the derived model, iii) the synthetic data, predicted by the model and iv) residuals of modelled and observed data. Ideally together with information about the performance/convergence of the inversion and the final data misfit e.g. as RMS. Based on this information, the reliability of the model can be evaluated. Such information would be mandatory for the manuscript. According to journal standards, the full dataset should be made avaialble as supplementary data, ideally together with the data processing algorithm, to give the reader transparent insight in the robustness of the approach.
The derived model in Figures 7&8 do not show any convincing correlation to the internal structure of the investigated ice body, the derived conductivities largely follow the topography and model boundaries. In Figure 8, the authors interpret the internal structures arbitrarily at the colour contours of 0.1, 0.2, 0.3, 0.4 mS/cm without any physical reasoning. In the introduction section the authors have introduced such units with conductivities orders of magnitude different.
I understand that the authors have invested a substantial field effort to acquire and process this dataset and compile this manuscript. But on the basis of i) the inconsistent approach in applying the technique of FDEM to an environment of such highly resistive ground, ii) the shortcomings in the data processing, iii) the incomplete information of the inversion and iv) the inconsistent interpretation, I cannot recomment this study for publication in The Cryosphere. I see the limitations of the manuscript so fundamental, that even major revisions will not bring the manuscript to a publishable standard and thus recomment rejection, but leave the final verdict to the editor.
Citation: https://doi.org/10.5194/tc-2022-190-RC2 -
AC2: 'Reply on RC2', Mirko Pavoni, 13 Jan 2023
We respect the comments of the ANONYMOUS reviewer, even if we partially disagree with some of her/his conclusions. Here below is a point-by-point reply.
1) The authors present a multidisciplinary study to investigate the structure of a glacier ice body. They apply the methods of ground penetrating radar and induced electromagnetic prospecting. For the latter technique, the authors utilize a CMD-DUO system. This system relies on the principle that an electromagnetic field is emitted by a Tx coil. In a conductive subsurface, secondary currents are induced and the superposition of the primary signal from the transmitter and the second signal from the induced eddy currently in the subsurface are recorded in the Rx coil. This principle is inherently limited to applications where the subsurface is sufficiently conductive to induce eddy currents large enough to exceed the detection threshold of the recording system. The manufacturer of the used system gives this value at a limit of 10E-1 mS/m, two orders of magnitude larger than the expected conductivity of massive ice. The authors have explained these limitations correctly in their manuscript.
Consequently, the recorded data show no evidence of sufficiently large induced signals, as evident from Figure 3. Here, the data scatter around zero, with large parts of the data showing physically implausible negative values. In the next step "Anomalous measurements filtering" is applied to remove outliers larger than two standard deviations. In figure 3C where polynomial interpolation smoothing is applied, the data appear all positive. This cannot be explained by the presented processing steps.
Reply 1: As pointed out by the reviewer, in the manuscript we have widely highlighted the principles of the FDEM method and the instrumental limitation of the CMD-DUO device. We should better underline that the resolution capabilities are linked to the volt-meter installed, and it does not represent the range of applications. Therefore, all our interpretations are done in a relative way, as explained.
We must specify that Fig.3 was misinterpreted by the Reviewer, since no negative data values are presented, and we did not record data scattering around zero. Fig.3A shows the raw dataset of Line 1 acquired with coils separation of 40 meters and horizontal coils mode orientation. Figure 3B shows data scattering around zero and negative values after a detrend function was applied to the raw dataset shown in Fig.3A. Detrending (shifting) removes both offsets and linear trends, and was adopted just to remove outliers. For this reason, in Fig.3B negative values and scattering around zero are found (as expected). Once we removed the outliers from the detrended data, the saved measurements have been brought back to their initial raw values and then interpolated with a polynomial function, as shown in Fig.3C. We are sorry if this process was not clear to Reviewer 2 (on the contrary of Reviewer 1), and we will better explain the filtering steps in the revised version of the manuscript. Obviously, all the raw data will be available for TC readers in case they intend to replicate the processing.
It seems that most of the anonymous Reviewer criticisms are based on the assumption that EMI methods cannot be applied in highly resistive environments. It is true that few eddy currents can be induced in low conductive layers but, as underlined in chapter 3.2 and in Fig.2, the measured apparent conductivities derive from the contribution of the whole deposits that compose the subsurface. As we can see in Fig.3C, higher apparent conductivities are measured for x<40 m and lower values are found for x>40 m, suggesting that the induction of eddy currents is facilitated in the first part of the transect, where in fact the ice layer has a lower thickness (see Fig.6A). On the other hand, for x>40 m, the apparent conductivity values decrease since the thickness of the ice layer increase and hinders the induction of a large amount of eddy currents in the subsurface. The structure of ice layers explaining our relative interpretation is confirmed by independent GPR and borehole measurements. The instrument resolution limit is in fact not linked to the induction capabilities (as the Reviewer seems to assert) but to the ability of the instrument to detect the weak low voltage received. Even if generating eddy currents in low conductive environments is challenging, EMI methods (regardless of time or frequency domains) have been historically applied in glacial and periglacial environments with ice-rich layers in the subsurface, e.g.:
- Bucki, A., Echelmeyer, K., & MacInnes, S. (2004). The thickness and internal structure of Fireweed rock glacier, Alaska, U.S.A., as determined by geophysical methods. Journal of Glaciology, 50(168), 67-75.
- Cockx, Liesbet, et al. Prospecting frost‐wedge pseudomorphs and their polygonal network using the electromagnetic induction sensor EM38DD. Permafrost and Periglacial Processes 17.2 (2006): 163-168.
- Daniel Blatter, Kerry Key, Anandaroop Ray, Neil Foley, Slawek Tulaczyk, Esben Auken, Trans-dimensional Bayesian inversion of airborne transient EM data from Taylor Glacier, Antarctica, Geophysical Journal International, Volume 214, Issue 3, September 2018, Pages 1919–1936.
- Foged, N., et al. "Airborne and ground-based TEM mapping in polar regions—Antarctica cases. NSG2021 2nd Conference on Geophysics for Infrastructure Planning, Monitoring and BIM. Vol. 2021. No. 1. European Association of Geoscientists & Engineers, 2021.
- Harada, K., Wada, K. and Fukuda, M. (2000). Permafrost mapping by transient electromagnetic method. Permafrost and Periglacial Processes, 11, 71–84.
- Harada, K., Wada, K. and Fukuda, M. (2003). Detection of permafrost structure by transient electromagnetic method in Mongolia. Proceedings of the 8th International Conference on Permafrost, Zurich, Switzerland, Extended Abstracts Reporting Current Research and New Information, 53–54.
- Hauck, C., Guglielmin, M., Isaksen, K. and Vonder Muhll, D. 2001. Applicability of frequency domain and time domain electromagnetic methods. Permafrost Periglac. Process, 12(1), 39–52.
- Hauck, C., Mühll, D.V. (1999). Detecting alpine permafrost using electro-magnetic methods. In: Hutter, K., Wang, Y., Beer, H. (eds) Advances in Cold-Region Thermal Engineering and Sciences. Lecture Notes in Physics, vol 533. Springer, Berlin, Heidelberg.
- Hoekstra, Pieter, Paul V. Sellmann, and Al Delaney. Ground and airborne resistivity surveys of permafrost near Fairbanks, Alaska. Geophysics 40.4 (1975): 641-656.
- Grombacher, D., Auken, E., Foged, N., Bording, T., Foley, N., Doran, P. T., ... & Tulaczyk, S. (2021). Induced polarization effects in airborne transient electromagnetic data collected in the McMurdo Dry Valleys, Antarctica. Geophysical Journal International, 226(3), 1574-1583.
- Madsen, L. M., Bording, T., Grombacher, D., Foged, N., Foley, N., Dugan, H. A., ... & Auken, E. (2022). Comparison of ground-based and airborne transient electromagnetic methods for mapping glacial and permafrost environments: Cases from McMurdo Dry Valleys, Antarctica. Cold Regions Science and Technology, 199, 103578.
- Maurer, H., & Hauck, C. (2007). Geophysical imaging of alpine rock glaciers. Journal of Glaciology, 53(180), 110-120.
- Neil Foley, Slawek Tulaczyk, Esben Auken, Cyril Schamper, Hilary Dugan, Jill Mikucki, Ross Virginia, Peter Doran; Helicopter-borne transient electromagnetics in high-latitude environments: An application in the McMurdo Dry Valleys, Antarctica. Geophysics 2015; 81 (1): WA87–WA99.
- Petersen, E., Holt, J., Stuurman, C., Levy, J. S., Nerozzi, S., Paine, J. G., ... & Fahnestock, M. (2016, March). Sourdough Rock Glacier, Alaska: An analog to martian debris-covered glaciers. In 47th Lunar and Planetary Science Conference (Vol. 2535).
2) After applying an inverse modelling, the results do not match plausible values for glacier ice bodies. This can logically explain by improper input data. The authors confirm the implausible data and apply an empiric shift of two orders of magnitude, justified by the misfit between synthetic data and FDEM models from field data. No physical explanation for this approach is presented. This empiric shift of the data is by no means a "calibration", it is rather an adaptation to the expected values.
Reply 2: We agree that we wrong term speaking about “calibration” and we thank the anonymous Reviewer for this comment. The term “correction” is more suitable since we introduced just a fixed shifting correction factor (1E-2 mS/m). The latter has been defined by comparing the inverted models from the field datasets (Fig.5) and the ones obtained from the synthetic datasets (Fig.7). Considering the instrumental limit resolution, from the inversion of field datasets we cannot expect to find conductivity values in the same range of table 2. Therefore, the synthetic forward modeling process was computed to verify the obtained results. The fixed correction factor is applied to the inversion results of the field datasets and allows to have models with the same conductivity range as in the inverted synthetic models (from 0 to 1 mS/m – see Fig.7). In this way, as you can see in Fig.8 and Fig.9 (FDEM inverted and corrected sections), the range of conductivity in the field dataset models span from 0 to 1 mS/m, and the ice layer boundaries can be defined with the same values found in the synthetic models (0-0.1 for the ice-rich layer and 0.1-0.2 for the ice-debris mixture, see chapter 4.3). We are asserting that real data fit the synthetic ones, suggesting the same subsoil structure with conductivities just shifted to higher values. Note that, we applied the same “correction factor” procedure to the result of FDEM surveys performed in rock glacier environments, and the resulting subsoil structures were confirmed by the ERT surveys carried out on the same investigation lines.
3) It is common practice for models obtained by inversion techniques to present i) the full recorded data, ii) the derived model, iii) the synthetic data, predicted by the model, and iv) residuals of modeled and observed data. Ideally together with information about the performance/convergence of the inversion and the final data misfit e.g. as RMS. Based on this information, the reliability of the model can be evaluated. Such information would be mandatory for the manuscript. According to journal standards, the full dataset should be made available as supplementary data, ideally together with the data processing algorithm, to give the reader transparent insight in the robustness of the approach.
Reply 3: We will insert in chapter 4.2 (FDEM inversion results) the RRMSE (Relative Root Mean Squared Error) to evaluate the accuracy of both the inverted models (line 1 and line 2) and we will create a public GitHub repository where we will insert FDEM and GPR datasets (not possible in the pre-print version). As we highlighted in the manuscript, EMagPy is a published Python-based open-source software to process FDEM forward and inverse modeling. Therefore, there is no algorithm that we need to share since it is already easily and freely available to download (see McLachlan et al., 2021).
4) The derived model in Figures 7&8 do not show any convincing correlation to the internal structure of the investigated ice body, the derived conductivities largely follow the topography and model boundaries. In Figure 8, the authors interpret the internal structures arbitrarily at the colour contours of 0.1, 0.2, 0.3, 0.4 mS/m without any physical reasoning. In the introduction section the authors have introduced such units with conductivities orders of magnitude different.
Reply 4: We respect the comment of the anonymous Reviewer, but we do not agree with her/his opinion. Our FDEM sections define a subsurface structure very similar to the GPR models, collected independently by another research group, as we highlighted in Fig.8 and Fig.9. To confirm our findings (e.g. the ice layer boundaries) we performed the forward modeling process (see chapter 4.3 FDEM forward modeling results). As we underlined in the conclusion chapter of the manuscript, the results of the FDEM forward modeling, which does not consider any instrumental limit, demonstrate that in these low-conductive environments, it is not possible to retrieve the real conductivity values of the layers (compare values of table 2 and Fig.7), even applying the Maxwell full solution in the inversion of synthetic datasets, however, we can retrieve the subsoil structure. Consequently, a low conductive environment does not exclude the use of EMI methods if the results are interpreted in a relative way to define the subsurface structure. In our case, the interpretation of the ice layer boundaries (<0.1 mS/m ice-rich layer, 0.1-0.2 mS/m ice, and debris) has been defined by the synthetic modeling (Fig 7), while subsoil geometries were confirmed from independent GPR data.
The model boundaries tend to follow the topography, both for FDEM and GPR models, since the glacier itself influences the morphology of the surface. Obviously, in the GPR model, the thickness variations of the ice layer are more appreciable, as expected given the higher resolution of the method. FDEM technique has clearly less capability to detect these lateral variations since it is a 1D vertical survey, and the CMD-DUO device has a very large coil separation (10-20-40 m). Moreover, the FDEM data inversion is still 1D, this means that the results are vertical conductivity profiles that can be interpolated in pseudo-2D conductivity section, as discussed in this work. Moreover, FDEM datasets have been acquired in challenging condition and it was complicated to guarantee the perfect coils orientation, height, and separation during the measurements. For these reasons we applied to the datasets the smoothing presented in Fig.3C, which may have hidden some information as small lateral conductivity variations.
5) I understand that the authors have invested a substantial field effort to acquire and process this dataset and compile this manuscript. But on the basis of i) the inconsistent approach in applying the technique of FDEM to an environment of such highly resistive ground, ii) the shortcomings in the data processing, iii) the incomplete information of the inversion and iv) the inconsistent interpretation, I cannot recommend this study for publication in The Cryosphere. I see the limitations of the manuscript so fundamental, that even major revisions will not bring the manuscript to a publishable standard and thus recommend rejection, but leave the final verdict to the editor.
Reply 5: We respect the opinion of the anonymous reviewer, but we do not agree with her/his conclusions that seem based more on a prejudice respect the EMI methods than on a careful reading of our case study. i) Induction methods can be used in periglacial and glacial environments (see literature above); ii) we adopted a solid inversion process already published in several papers; iii) our findings are confirmed by the comparison with GPR and borehole measurements. In the corrected manuscript we will better specify the data filtering and processing, as the Reviewer suggested. We hope that the Editor can appreciate the revised version of the manuscript, far from any methodological prejudice, since we think it represents an interesting application for the cryosphere community.
-
AC2: 'Reply on RC2', Mirko Pavoni, 13 Jan 2023
Status: closed
-
RC1: 'Comment on tc-2022-190', Anonymous Referee #1, 22 Nov 2022
Review comments for “Induced Electromagnetic prospecting for the characterization of the European southernmost glacier: the Calderone Glacier, Apennines, Italy” by Pavoni et al. manuscript 2022-190 for TC
Summary:
This paper has clear motivations and objectives. It is refreshing to see multiple geophysical methods being used in a glaciated environment. However, there are some major comments which should be addressed before considered for publication.
Major comments:
- This paper would greatly benefit from including the results and a log from the borehole observations mentioned in Line xx. The authors repeatability refer to the validation of the geophysical methods by borehole observations yet do not include the borehole results in the study. There is no location on the base map in Figure 1 of the borehole location. The paper would be much more complete with the addition of this observation.
- It is commonly known that electromagnetic methods are highly non-unique. Why do the authors not constrain the FDEM inversion with the GPR depths (snow and ice) and a range of expected conductivities?
- What is the uncertainty in the pseudo 2D FDEM inversion? Are there areas where the uncertainty is larger, for example in the ice, or at depth? A cross-section showing the uncertainty range/standard deviation would be useful to enable reliable assessment of the 2D section.
- I would consider changing the title. The words “prospecting for the characterization” do not work well together and FDEM should be in the title as this is the more novel technique used in this glaciated environment. I would suggest something along the lines of: “Characterisation of the Calderone Glacier, Apennines, Italy using GPR, FDEM and borehole observations”.
- The main text needs to be proof read thoroughly.
Minor comments:
- Figure A3 in the appendix, should be in the main text the depth of investigation is important in understanding the limitations of the electromagnetic inversion results.
- In the inverted FDEM sections, can you explain in more detail what is going on under the ice? The values seam to be more conductive than if it was bedrock.
Specific comments:
L98-99: did you get any CMP gathers to estimate the velocity of the snow and ice layer?
L130: “dozens” should be changed to “tens”
L134-136: the last sentence in this paragraph doesn’t make sense.
L194-196: Need to explain this sentence in more detail. What are the instrumental resolution limits?
L201-202: where did you measure the several meters of snow cover during the data acquisition? I would mark these locations on the base map in Figure 1. Was this by a snow pit?
I would change the use of the word “subsoil” to “subsurface” as you are not working in a soil environment.
L221-222 This sentence doesn’t make sense.
L226: “exciting” ? Do you mean “existing”?
L237-8: The ice ends at “x=30m”, where did it end before and what is this value relative to. The retreat of the glacier should be discussed in terms of “The ice mass has retreated xx m since 2015”.
L247-8: this borehole observation should be added to this paper.
L255: where is this measurement, show on Figure 1.
L263-267: This last paragraph is important for the FDEM inversion. I would have this in the main text before the inversion results, including figure A3.
L269: The paper needs to back up this concluding sentence with the borehole observations and a more detailed map in Figure 1 of where the snow depth measurements were acquired.
L274-275: do you mean time-laspe geophysical surveys?
L292: Repetition of the drilling/borehole observations with no results detailed in the paper.
L294-297: this should be in the discussion.
Table 1: could be moved to the appendix.
Figure 2: Nice figure. what is the measured apparent conductivity used to create these depth ranges? Do you not have to input a conductivity to estimate the depth of each coil separation?
Table 2: This is just a personnel preference, however many recent papers using electromagnetic methods in a glaciological environment work in terms of resistivity, in ohm.m, the inverse of conductivity. It might be easier for your readers to follow and directly compare with other studies if you worked in terms of resistivity.
Figure 5: This is also another personnel preference, however most recent papers using electromagnetic methods in a glaciological environment have hot colours (like red and purple) representing high resistivity/low conductivity and cold colours (e.g. blue) representing low resistivity/high conductivity. Again, this might make it easier for your readers to follow and compare with other studies.
Figure 8 and 9: Consider having one figure for 8 and one for 9, merging A and B, with the GPR transparently overlaid on the conductivity plot.
Figure A3: Have this in the main text. Add a dotted line at the depth of investigation. To me it looks like you only have sensitivity to 20 m depth as the sensitivity curves come close together after 20 m?
Citation: https://doi.org/10.5194/tc-2022-190-RC1 -
AC1: 'Reply on RC1', Mirko Pavoni, 15 Dec 2022
We truly thank the anonymous Reviewer for very constructive comments and suggestions about our manuscript.
Reply point by point:
- This paper would greatly benefit from including the results and a log from the borehole observations mentioned in Line xx. The authors repeatability refers to the validation of the geophysical methods by borehole observations yet do not include the borehole results in the study. There is no location on the base map in Figure 1 of the borehole location. The paper would be much more complete with the addition of this observation.
Reply: We fully agree with this comment and we will include in the manuscript the information about the borehole stratigraphy and the position in the map/models.
- It is commonly known that electromagnetic methods are highly non-unique. Why do the authors not constrain the FDEM inversion with the GPR depths (snow and ice) and a range of expected conductivities?
Reply: The Reviewer is right, as the result of any kind of geophysical method inversion process, we know that our FDEM models are a non-unique solution to the problem. Nevertheless, the aim of the FDEM survey on the Calderone Glacier was to verify the capability of the separated-coils FDEM device to characterize an environment with ice fraction occurring in the investigated subsurface. Consequently, it was important for us to evaluate the inversion results without using any prior structural information, as defined by the GPR profiles. In this way, we had the opportunity to understand if it is possible to use the separated-coils FDEM device in other similar environments, as rock glaciers, without the need to combine it with GPR surveys, which are challenging to perform with a coarse-blocky surface and less resolutive since the ice is highly mixed with debris (signal scattering problem). Considering the results obtained on the Calderone Glacier, we think that we achieved our goal. In the summer of 2022 we tested the separated-coils FDEM device in several rock glaciers and achieved very promising results.
- What is the uncertainty in the pseudo-2D FDEM inversion? Are there areas where the uncertainty is larger, for example in the ice, or at depth? A cross-section showing the uncertainty range/standard deviation would be useful to enable a reliable assessment of the 2D section.
Reply: Concerning the sensitivity of the inverted models, the used code EMagPy does not provide a sensitivity 2D section. However, the software provides the percentage value of RRMSE (Relative Root Mean Squared Error) to evaluate the accuracy of the inverted models. Therefore, we will insert this information in the manuscript for both investigation lines. Nevertheless, as we highlighted in the manuscript, the sensitivity of the FDEM measurements is higher in the near subsurface and decreases with the depth till it reaches (approximately) zero at a depth of about 30 meters (see Fig.A3, which will be removed from the Appendix and insert in the main text of the manuscript), consequently, we can consider the uncertainty of the pseudo-2D FDEM models in the same way.
- I would consider changing the title. The words “prospecting for the characterization” do not work well together and FDEM should be in the title as this is the more novel technique used in this glaciated environment. I would suggest something along the lines of: “Characterisation of the Calderone Glacier, Apennines, Italy using GPR, FDEM, and borehole observations”.
Reply: We agree with the Reviewer and this useful suggestion, we will modify the title and we will include “FDEM” inside it, e.g. Combination of Ground Penetrating Radar and Frequency-Domain Electromagnetic methods for the characterization of the Calderone Glacier (Gran Sasso d’Italia, Italy).
Even if the borehole result is important to validate our geophysical results, we don’t think that “borehole” could be appropriate to insert in the title, since only the geophysical surveys were carried out to characterize the structure of the glacier, and this way define the most suitable position to drill and extract the ice core sample, which was the target of the Ice Memory Project team.
- The main text needs to be proof read thoroughly
Reply: The reviewer is right, we will do it.
- Figure A3 in the appendix should be in the main text. The depth of investigation is important in understanding the limitations of the electromagnetic inversion results.
Reply: The suggestion is correct, we can move Figure A3 in the main text, particularly in Chapter 4.2 (FDEM inversion results).
- In the inverted FDEM sections, can you explain in more detail what is going on under the ice? The values seem to be more conductive than if it was bedrock
Reply: According to values that we found in literature, and that we observed ourselves in ERT surveys in the North-Eastern Alps areas, we assigned a representative value of 2E-1 mS/m to the calcareous bedrock. In the inverted and calibrated model Line 1 (Fig.8A), the boundary with the bedrock is well defined and it is very similar to the synthetic model (Fig.7A), suggesting the presence of the bedrock as confirmed also by the GPR profile and the borehole. On the other hand, in the inverted and calibrated model of Line 2 (Fig.9A), it is true that the conductivity values at the bottom of the section are higher than expected for the bedrock, particularly in the western direction. In this area, it is plausible that the ice layer is not directly in contact with the calcareous bedrock but instead with a lateral moraine, which obviously has higher conductivity values than bedrock. We will add these considerations in the main text, in particular in the chapter of Discussion.
- Specific comments:
- L98-99: did you get any CMP gathers to estimate the velocity of the snow and ice layers?
Reply: GSSI Sir4000 is a monostatic digital GPR antenna, therefore it was not possible to acquire CMP gathers.
- L130: “dozens” should be changed to “tens”
Reply: The Reviewer is right, it will be changed.
- L134-136: the last sentence in this paragraph doesn’t make sense: “Despite this, FDEM methods proved to be efficiently applicable in high resistive environments, considering in a relative way the inverted conductibility profile (e.g. Boaga et al. 2020; Pavoni et al. 2021).”
Reply: As the FDEM devices produced by GF Instruments (and other manufacturers) have not the capability to detect conductivity variations lower than 0.1 mS/m (instrumental limit resolution), we cannot pretend to measure the real conductivity values of the ice layers, which have conductivities << 0.1 mS/m. Consequently, in these low conductive environments, FDEM models should not be interpreted on the basis of the inverted conductivity values, but a calibration based on the forward modeling is needed to define a value representative for the frozen layers (e.g. Pavoni et al. 2021 evaluated values < 1 mS/m for the frozen layer in dolomitic rock glaciers, which was confirmed by the comparison with ERT results). We will re-phrase the sentence explaining better this concept in the manuscript.
- L194-196: Need to explain this sentence in more detail. What are the instrumental resolution limits?
Reply: The FDEM devices produced by GF Instruments have not the capability to detect conductivity variations lower than 0.1 mS/m, which is therefore the instrumental limit resolution.
- L201-202: where did you measure the several meters of snow cover during the data acquisition? I would mark these locations on the base map in Figure 1. Was this by a snow pit?
Reply: The Reviewer is absolutely right, we have to show in the map and in the models the point where the snow cover thickness has been measured. Yes, it was a snow pit. For the measurement, we used an extendable rod and the measured point is approximately at x~40 m in Line 1. We will specify this in the text.
- L221-222 This sentence doesn’t make sense: “Calibration intends to explore if exist a constant correction factor to be applied to the inversion results of the field datasets, in order to have the same conductivity scale of the synthetic model.”
Reply: We compared the inverted models obtained from the field datasets (Fig.5 – from now on field dataset inverted models) and the ones obtained from the synthetic datasets (Fig.7 - from now on synthetic models), searching for a correction factor that can be applied to the inversion results of the field datasets and allows to have a model with the same conductivity range as found in the inverted synthetic models (from 0 to 1 mS/m – see Fig.7). The defined correction factor of 1E-2 mS/m has been applied to the inversion results of field datasets (Line 1 and Line 2). In this way, as you can see in Fig.8 and Fig.9 (FDEM inverted and calibrated sections), the range of conductivity in the field dataset models span from 0 to 1 mS/m, and the ice layer boundaries can be defined with the same values found in the synthetic models (0-0.1 for the ice-rich layer and 0.1-0.2 for the ice-debris mixture, see chapter 4.3). We will modify chapter 4.4 and explain better the calibration of the FDEM results.
- L226: “exciting”? Do you mean “existing”?
Reply: Yes, sorry for the typo. We will correct it.
- L237-8: The ice ends at “x=30m”, where did it end before and what is this value relative to. The retreat of the glacier should be discussed in terms of “The ice mass has retreated xx m since 2015”.
Reply: Along the longitudinal Line 1, the ice-rich layer was easily detectable along the entire GPR profile measured in 2015 by Monaco & Scozzafava, but today seems to end at x≈30m. Therefore, in the last 7 years, between x = 0 and x = 30 meters of Line 1, we had (presumably) a loss of massive ice and an increase in the amount of ice-debris mixture.
- L247-8: this borehole observation should be added to this paper.
Reply: Yes, we will do it.
- L255: where is this measurement, show on Figure 1.
Reply: Yes, we will do it.
- L263-267: This last paragraph is important for the FDEM inversion. I would have this in the main text before the inversion results, including figure A3.
Reply: We really appreciate this suggestion. Our proposal is to add at the end of Chapter 3.2.1 (FDEM forward and inverse modeling) this sentence: “to define the bottom depth of the models, sensitivity profiles of the measurements have been calculated with EMagPy. In the current work, the inverted FDEM models are limited to the depths where the normalized sensitivity of the measurements reaches approximately zero.” We can show the sensitivity profiles in Chapter 4.2 (FDEM inversion results) and add this sentence: “the sensitivity of the measurements performed along Line 2 is presented (same results were found for Line 1). It is clear that sensitivities are higher in the near subsurface and decrease to (approximately) zero at a depth of about 30 meters. Consequently, we considered the uncertainty of the pseudo-2D FDEM inverted models in the same way, and we defined the bottom of the FDEM sections at a depth of 30 meters from the surface.” Finally, we can leave the following sentences in the discussion chapter: “It should be noted that both the FDEM inverted models have lower penetration depth (~30 m) than those predicted by the instrument manufacturer (see Fig.2). This is expected since the investigation depth decreases in low electrical conductivity environments (Hauck and Kneisell, 2008).”
- L269: The paper needs to back up this concluding sentence with the borehole observations and a more detailed map in Figure 1 of where the snow depth measurements were acquired.
Reply: Yes sure, we will do it.
- L274-275: do you mean time-lapse geophysical surveys?
Reply: Yes, it was the meaning of the sentence. The Reviewer is right and we can insert directly the reference to the time-lapse surveys (e.g. “A future development for the GPR measurements on the Calderone Glacier is to apply the method proposed by Santin et al. (2022), to estimate the debris content within the layer composed of ice-debris mixture. In case of periodic measurements in time-lapse configuration, this method could help to estimate the ice volume losses of the Calderone Glacier in the next future.”).
- L292: Repetition of the drilling/borehole observations with no results detailed in the paper.
Reply: It will be insert into the manuscript.
- L294-297: this should be in the discussion.
Reply: We respect the opinion of the Reviewer but it is our opinion that these sentences fit better in the conclusions of our work.
- Table 1: could be moved to the appendix.
Reply: We would prefer to leave it inside the main text, but we can move it to the Appendix if required.
- Nice figure. what is the measured apparent conductivity used to create these depth ranges? Do you not have to input a conductivity to estimate the depth of each coil separation?
Reply: Fig. 2 shows the nominal effective depth range influencing the measured apparent conductivities suggested by the manufacturer of the CMD-DUO device, GF Instruments. Instead of showing a Table with these effective depth ranges, we preferred to create a new intuitive figure expressly for this work.
- Table 2: This is just a personnel preference, however many recent papers using electromagnetic methods in a glaciological environment work in terms of resistivity, in ohm*m, the inverse of conductivity. It might be easier for your readers to follow and directly compare with other studies if you worked in terms of resistivity.
Reply: In our experience, we prefer to use conductivity (mS/m) when we work with EMI methods. As the Reviewer correctly mentioned, this is just our personnel preference. To be consistent, in the main text of the manuscript we will remove all the references to electrical resistivity and we will refer only to conductivity.
- Figure 5: This is also another personnel preference, however most recent papers using electromagnetic methods in a glaciological environment have hot colours (like red and purple) representing high resistivity/low conductivity and cold colours (e.g. blue) representing low resistivity/high conductivity. Again, this might make it easier for your readers to follow and compare with other studies.
Reply: As before for conductivity-resistivity, we personally prefer to use a color scale where the frozen/ice layer (with high electrical resistivities – low electrical conductivities) is presented with cold colors as blue, and the unfrozen subsoil/subsurface is defined by hot colors as red. In our opinion, as we verified in various geophysical conferences, it is more intuitive to associate the blue color with the ice, particularly for people that are not usually inside the cryosphere “world”. Again, as the Reviewer correctly mentioned, this is just our personnel preference.
- Figures 8 and 9: Consider having one figure for 8 and one for 9, merging A and B, with the GPR transparently overlaid on the conductivity plot.
Reply: We appreciate the suggestion of the Reviewer but it is our opinion that a comparison of separated models (FDEM and GPR) is the best solution to better appreciate the similarities between them. Moreover, we will insert in this Figure the stratigraphy of the borehole. Once again, this is just our personal preference.
- Figure A3: Have this in the main text. Add a dotted line at the depth of investigation. To me, it looks like you only have sensitivity to 20 m depth as the sensitivity curves come close together after 20 m?
We will move Fig.A3 into the main text, e.g. in Chapter 4.2. (FDEM inversion results). We could add a vertical dashed line that defines the value of zero sensitivity. In this way, it would be easier to appreciate that the sensitivity of the FDEM survey reaches zero at approximately 30 meters of depth. Between 20 and 30 meters of depth, the sensitivity of the measurements is still > 0, so that part of the subsurface is still contributing to the measured apparent conductivities. For this reason, we have limited our FDEM models to 30 meters of depth from the ground level.
Citation: https://doi.org/10.5194/tc-2022-190-AC1
-
RC2: 'Comment on tc-2022-190', Anonymous Referee #2, 20 Dec 2022
The authors present a multidisciplinary study to investigate the structure of a glacier ice body. They apply the methods of ground penetrating radar and induced electromagnetic prospecting. For the latter technique the authors utilize a CMD-DUO system. This system relies on the principle that an electromagnetic field is emitted by a Tx coil. In a conductive subsurface, secondary currents are induced and the superposition of the primary signal from the transmitter and the secondary signal from the induced eddy currenty in the subsurface are recorded in the Rx coil. This principle is inherently limited to applications where the subsurface is sufficiently conductive to induce eddy currents large enough to exceed the detection threshold of the recording system. The manufacturer of the used system gives this value at a limit of 10E-1 mS/cm, two orders of magnitude larger then the expected conductivity of massive ice. The authors have explained these limitations correctly in their manuscript.
Consequently, the recorded data show no evidence of sufficiently large induced signals, as evident from Figure 3. Here, the data scatter around zero, large parts of the data showing physically implausibel negative values. In a next step "Anomalous measurements filtering" is applied to remove outliers larger than two standard deviations. In the figure Figure 3C where polynomial interpolation smoothing is applied the data appear all positive. This cannot be explaind by the presented processing steps.
After applying an inverse modelling, the results do not match plausible values for glacier ice bodies. This can logically explained by improper input data. The authors confirm the unplausible data and apply an empiric shift of two orders of magnitude, justified by the misfit bewteen syntheic data and FDEM models from field data. No physical explanation for this appraoch is presented. This empiric shift of the data is by no means a "calibration", it is rather an adaptation to the expected values.
It is common practice for models obtained by inversion techniques to present i) the full recorded data, ii) the derived model, iii) the synthetic data, predicted by the model and iv) residuals of modelled and observed data. Ideally together with information about the performance/convergence of the inversion and the final data misfit e.g. as RMS. Based on this information, the reliability of the model can be evaluated. Such information would be mandatory for the manuscript. According to journal standards, the full dataset should be made avaialble as supplementary data, ideally together with the data processing algorithm, to give the reader transparent insight in the robustness of the approach.
The derived model in Figures 7&8 do not show any convincing correlation to the internal structure of the investigated ice body, the derived conductivities largely follow the topography and model boundaries. In Figure 8, the authors interpret the internal structures arbitrarily at the colour contours of 0.1, 0.2, 0.3, 0.4 mS/cm without any physical reasoning. In the introduction section the authors have introduced such units with conductivities orders of magnitude different.
I understand that the authors have invested a substantial field effort to acquire and process this dataset and compile this manuscript. But on the basis of i) the inconsistent approach in applying the technique of FDEM to an environment of such highly resistive ground, ii) the shortcomings in the data processing, iii) the incomplete information of the inversion and iv) the inconsistent interpretation, I cannot recomment this study for publication in The Cryosphere. I see the limitations of the manuscript so fundamental, that even major revisions will not bring the manuscript to a publishable standard and thus recomment rejection, but leave the final verdict to the editor.
Citation: https://doi.org/10.5194/tc-2022-190-RC2 -
AC2: 'Reply on RC2', Mirko Pavoni, 13 Jan 2023
We respect the comments of the ANONYMOUS reviewer, even if we partially disagree with some of her/his conclusions. Here below is a point-by-point reply.
1) The authors present a multidisciplinary study to investigate the structure of a glacier ice body. They apply the methods of ground penetrating radar and induced electromagnetic prospecting. For the latter technique, the authors utilize a CMD-DUO system. This system relies on the principle that an electromagnetic field is emitted by a Tx coil. In a conductive subsurface, secondary currents are induced and the superposition of the primary signal from the transmitter and the second signal from the induced eddy currently in the subsurface are recorded in the Rx coil. This principle is inherently limited to applications where the subsurface is sufficiently conductive to induce eddy currents large enough to exceed the detection threshold of the recording system. The manufacturer of the used system gives this value at a limit of 10E-1 mS/m, two orders of magnitude larger than the expected conductivity of massive ice. The authors have explained these limitations correctly in their manuscript.
Consequently, the recorded data show no evidence of sufficiently large induced signals, as evident from Figure 3. Here, the data scatter around zero, with large parts of the data showing physically implausible negative values. In the next step "Anomalous measurements filtering" is applied to remove outliers larger than two standard deviations. In figure 3C where polynomial interpolation smoothing is applied, the data appear all positive. This cannot be explained by the presented processing steps.
Reply 1: As pointed out by the reviewer, in the manuscript we have widely highlighted the principles of the FDEM method and the instrumental limitation of the CMD-DUO device. We should better underline that the resolution capabilities are linked to the volt-meter installed, and it does not represent the range of applications. Therefore, all our interpretations are done in a relative way, as explained.
We must specify that Fig.3 was misinterpreted by the Reviewer, since no negative data values are presented, and we did not record data scattering around zero. Fig.3A shows the raw dataset of Line 1 acquired with coils separation of 40 meters and horizontal coils mode orientation. Figure 3B shows data scattering around zero and negative values after a detrend function was applied to the raw dataset shown in Fig.3A. Detrending (shifting) removes both offsets and linear trends, and was adopted just to remove outliers. For this reason, in Fig.3B negative values and scattering around zero are found (as expected). Once we removed the outliers from the detrended data, the saved measurements have been brought back to their initial raw values and then interpolated with a polynomial function, as shown in Fig.3C. We are sorry if this process was not clear to Reviewer 2 (on the contrary of Reviewer 1), and we will better explain the filtering steps in the revised version of the manuscript. Obviously, all the raw data will be available for TC readers in case they intend to replicate the processing.
It seems that most of the anonymous Reviewer criticisms are based on the assumption that EMI methods cannot be applied in highly resistive environments. It is true that few eddy currents can be induced in low conductive layers but, as underlined in chapter 3.2 and in Fig.2, the measured apparent conductivities derive from the contribution of the whole deposits that compose the subsurface. As we can see in Fig.3C, higher apparent conductivities are measured for x<40 m and lower values are found for x>40 m, suggesting that the induction of eddy currents is facilitated in the first part of the transect, where in fact the ice layer has a lower thickness (see Fig.6A). On the other hand, for x>40 m, the apparent conductivity values decrease since the thickness of the ice layer increase and hinders the induction of a large amount of eddy currents in the subsurface. The structure of ice layers explaining our relative interpretation is confirmed by independent GPR and borehole measurements. The instrument resolution limit is in fact not linked to the induction capabilities (as the Reviewer seems to assert) but to the ability of the instrument to detect the weak low voltage received. Even if generating eddy currents in low conductive environments is challenging, EMI methods (regardless of time or frequency domains) have been historically applied in glacial and periglacial environments with ice-rich layers in the subsurface, e.g.:
- Bucki, A., Echelmeyer, K., & MacInnes, S. (2004). The thickness and internal structure of Fireweed rock glacier, Alaska, U.S.A., as determined by geophysical methods. Journal of Glaciology, 50(168), 67-75.
- Cockx, Liesbet, et al. Prospecting frost‐wedge pseudomorphs and their polygonal network using the electromagnetic induction sensor EM38DD. Permafrost and Periglacial Processes 17.2 (2006): 163-168.
- Daniel Blatter, Kerry Key, Anandaroop Ray, Neil Foley, Slawek Tulaczyk, Esben Auken, Trans-dimensional Bayesian inversion of airborne transient EM data from Taylor Glacier, Antarctica, Geophysical Journal International, Volume 214, Issue 3, September 2018, Pages 1919–1936.
- Foged, N., et al. "Airborne and ground-based TEM mapping in polar regions—Antarctica cases. NSG2021 2nd Conference on Geophysics for Infrastructure Planning, Monitoring and BIM. Vol. 2021. No. 1. European Association of Geoscientists & Engineers, 2021.
- Harada, K., Wada, K. and Fukuda, M. (2000). Permafrost mapping by transient electromagnetic method. Permafrost and Periglacial Processes, 11, 71–84.
- Harada, K., Wada, K. and Fukuda, M. (2003). Detection of permafrost structure by transient electromagnetic method in Mongolia. Proceedings of the 8th International Conference on Permafrost, Zurich, Switzerland, Extended Abstracts Reporting Current Research and New Information, 53–54.
- Hauck, C., Guglielmin, M., Isaksen, K. and Vonder Muhll, D. 2001. Applicability of frequency domain and time domain electromagnetic methods. Permafrost Periglac. Process, 12(1), 39–52.
- Hauck, C., Mühll, D.V. (1999). Detecting alpine permafrost using electro-magnetic methods. In: Hutter, K., Wang, Y., Beer, H. (eds) Advances in Cold-Region Thermal Engineering and Sciences. Lecture Notes in Physics, vol 533. Springer, Berlin, Heidelberg.
- Hoekstra, Pieter, Paul V. Sellmann, and Al Delaney. Ground and airborne resistivity surveys of permafrost near Fairbanks, Alaska. Geophysics 40.4 (1975): 641-656.
- Grombacher, D., Auken, E., Foged, N., Bording, T., Foley, N., Doran, P. T., ... & Tulaczyk, S. (2021). Induced polarization effects in airborne transient electromagnetic data collected in the McMurdo Dry Valleys, Antarctica. Geophysical Journal International, 226(3), 1574-1583.
- Madsen, L. M., Bording, T., Grombacher, D., Foged, N., Foley, N., Dugan, H. A., ... & Auken, E. (2022). Comparison of ground-based and airborne transient electromagnetic methods for mapping glacial and permafrost environments: Cases from McMurdo Dry Valleys, Antarctica. Cold Regions Science and Technology, 199, 103578.
- Maurer, H., & Hauck, C. (2007). Geophysical imaging of alpine rock glaciers. Journal of Glaciology, 53(180), 110-120.
- Neil Foley, Slawek Tulaczyk, Esben Auken, Cyril Schamper, Hilary Dugan, Jill Mikucki, Ross Virginia, Peter Doran; Helicopter-borne transient electromagnetics in high-latitude environments: An application in the McMurdo Dry Valleys, Antarctica. Geophysics 2015; 81 (1): WA87–WA99.
- Petersen, E., Holt, J., Stuurman, C., Levy, J. S., Nerozzi, S., Paine, J. G., ... & Fahnestock, M. (2016, March). Sourdough Rock Glacier, Alaska: An analog to martian debris-covered glaciers. In 47th Lunar and Planetary Science Conference (Vol. 2535).
2) After applying an inverse modelling, the results do not match plausible values for glacier ice bodies. This can logically explain by improper input data. The authors confirm the implausible data and apply an empiric shift of two orders of magnitude, justified by the misfit between synthetic data and FDEM models from field data. No physical explanation for this approach is presented. This empiric shift of the data is by no means a "calibration", it is rather an adaptation to the expected values.
Reply 2: We agree that we wrong term speaking about “calibration” and we thank the anonymous Reviewer for this comment. The term “correction” is more suitable since we introduced just a fixed shifting correction factor (1E-2 mS/m). The latter has been defined by comparing the inverted models from the field datasets (Fig.5) and the ones obtained from the synthetic datasets (Fig.7). Considering the instrumental limit resolution, from the inversion of field datasets we cannot expect to find conductivity values in the same range of table 2. Therefore, the synthetic forward modeling process was computed to verify the obtained results. The fixed correction factor is applied to the inversion results of the field datasets and allows to have models with the same conductivity range as in the inverted synthetic models (from 0 to 1 mS/m – see Fig.7). In this way, as you can see in Fig.8 and Fig.9 (FDEM inverted and corrected sections), the range of conductivity in the field dataset models span from 0 to 1 mS/m, and the ice layer boundaries can be defined with the same values found in the synthetic models (0-0.1 for the ice-rich layer and 0.1-0.2 for the ice-debris mixture, see chapter 4.3). We are asserting that real data fit the synthetic ones, suggesting the same subsoil structure with conductivities just shifted to higher values. Note that, we applied the same “correction factor” procedure to the result of FDEM surveys performed in rock glacier environments, and the resulting subsoil structures were confirmed by the ERT surveys carried out on the same investigation lines.
3) It is common practice for models obtained by inversion techniques to present i) the full recorded data, ii) the derived model, iii) the synthetic data, predicted by the model, and iv) residuals of modeled and observed data. Ideally together with information about the performance/convergence of the inversion and the final data misfit e.g. as RMS. Based on this information, the reliability of the model can be evaluated. Such information would be mandatory for the manuscript. According to journal standards, the full dataset should be made available as supplementary data, ideally together with the data processing algorithm, to give the reader transparent insight in the robustness of the approach.
Reply 3: We will insert in chapter 4.2 (FDEM inversion results) the RRMSE (Relative Root Mean Squared Error) to evaluate the accuracy of both the inverted models (line 1 and line 2) and we will create a public GitHub repository where we will insert FDEM and GPR datasets (not possible in the pre-print version). As we highlighted in the manuscript, EMagPy is a published Python-based open-source software to process FDEM forward and inverse modeling. Therefore, there is no algorithm that we need to share since it is already easily and freely available to download (see McLachlan et al., 2021).
4) The derived model in Figures 7&8 do not show any convincing correlation to the internal structure of the investigated ice body, the derived conductivities largely follow the topography and model boundaries. In Figure 8, the authors interpret the internal structures arbitrarily at the colour contours of 0.1, 0.2, 0.3, 0.4 mS/m without any physical reasoning. In the introduction section the authors have introduced such units with conductivities orders of magnitude different.
Reply 4: We respect the comment of the anonymous Reviewer, but we do not agree with her/his opinion. Our FDEM sections define a subsurface structure very similar to the GPR models, collected independently by another research group, as we highlighted in Fig.8 and Fig.9. To confirm our findings (e.g. the ice layer boundaries) we performed the forward modeling process (see chapter 4.3 FDEM forward modeling results). As we underlined in the conclusion chapter of the manuscript, the results of the FDEM forward modeling, which does not consider any instrumental limit, demonstrate that in these low-conductive environments, it is not possible to retrieve the real conductivity values of the layers (compare values of table 2 and Fig.7), even applying the Maxwell full solution in the inversion of synthetic datasets, however, we can retrieve the subsoil structure. Consequently, a low conductive environment does not exclude the use of EMI methods if the results are interpreted in a relative way to define the subsurface structure. In our case, the interpretation of the ice layer boundaries (<0.1 mS/m ice-rich layer, 0.1-0.2 mS/m ice, and debris) has been defined by the synthetic modeling (Fig 7), while subsoil geometries were confirmed from independent GPR data.
The model boundaries tend to follow the topography, both for FDEM and GPR models, since the glacier itself influences the morphology of the surface. Obviously, in the GPR model, the thickness variations of the ice layer are more appreciable, as expected given the higher resolution of the method. FDEM technique has clearly less capability to detect these lateral variations since it is a 1D vertical survey, and the CMD-DUO device has a very large coil separation (10-20-40 m). Moreover, the FDEM data inversion is still 1D, this means that the results are vertical conductivity profiles that can be interpolated in pseudo-2D conductivity section, as discussed in this work. Moreover, FDEM datasets have been acquired in challenging condition and it was complicated to guarantee the perfect coils orientation, height, and separation during the measurements. For these reasons we applied to the datasets the smoothing presented in Fig.3C, which may have hidden some information as small lateral conductivity variations.
5) I understand that the authors have invested a substantial field effort to acquire and process this dataset and compile this manuscript. But on the basis of i) the inconsistent approach in applying the technique of FDEM to an environment of such highly resistive ground, ii) the shortcomings in the data processing, iii) the incomplete information of the inversion and iv) the inconsistent interpretation, I cannot recommend this study for publication in The Cryosphere. I see the limitations of the manuscript so fundamental, that even major revisions will not bring the manuscript to a publishable standard and thus recommend rejection, but leave the final verdict to the editor.
Reply 5: We respect the opinion of the anonymous reviewer, but we do not agree with her/his conclusions that seem based more on a prejudice respect the EMI methods than on a careful reading of our case study. i) Induction methods can be used in periglacial and glacial environments (see literature above); ii) we adopted a solid inversion process already published in several papers; iii) our findings are confirmed by the comparison with GPR and borehole measurements. In the corrected manuscript we will better specify the data filtering and processing, as the Reviewer suggested. We hope that the Editor can appreciate the revised version of the manuscript, far from any methodological prejudice, since we think it represents an interesting application for the cryosphere community.
-
AC2: 'Reply on RC2', Mirko Pavoni, 13 Jan 2023
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 703 | 233 | 62 | 998 | 76 | 50 | 46 |
- HTML: 703
- PDF: 233
- XML: 62
- Total: 998
- Supplement: 76
- BibTeX: 50
- EndNote: 46
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1