Dear authors,
This is my second review of your paper. Although this version is largely improved, I have the feeling you misinterpreted some of my comments so that they are not entirely integrated in the new version. Through the clarifications you made, I understand this is not so crucial, as you only use the ERT data to derive estimated correlation length, but it is important that everything written is correct and that the discussion is clarified, since reviewer comments are also accessible along the published article.
I would therefore like to come back on some of my previous comments and your answers to them.
1. You are not using the ERT-data as a direct constraint to the facies. I would however classify approaches which did it as more advanced compared to what you do. I therefore think it is important this is highlighted more in the conclusion, and certainly as a perspective of the work. Such approach would make a true difference in the estimation of the facies (see 10.5194/hess-22-3351-2018, 10.5194/hess-22-5485-2018, examples with TEM data, but basically similar, and Hermans et al. (2015), cited but without a thorough discussion). This is really crucial to put your work in perspective of the existing literature.
2. You write: However, the resistivity anomalies (defining hydrofacies units up to 20 m depth in each profile) are very well defined, with clear resistivity contrasts between them that align with vertical lithological contacts observed in the wells. This result is very important to consider, because it allows us to confidently establish resistivity thresholds for each hydrofacies within the first 20 m depth, without uncertainty.
Writing “without uncertainty” is a very strong and, I am afraid, wrong statement. You should remember that any estimation you made is based on an inversion that distorts the true resistivity value. In addition, your hydrofacies are observed in boreholes that are not co-located. As a result, you don’t know where the interfaces are actually lying on your profile. It is impossible to state that it lies at a specific contourline. Even if you knew the exact location of the interface, It would likely not lie at the same contourline everywhere, because of the limitation of the inversion. This is actually confirmed as your interpretation involves different value for the transition between facies. It means that there is some subjectivity in the selection of these contourlines.
3. You write. Thus, we differ with reviewer’s point of view of implicitly suggesting that uncertainty analysis would show each hydrofacies having a probability of occurrence for any resistivity value, implying no distinct resistivity thresholds exist. From a mathematical perspective, this approach appears unsustainable given the physical and chemical properties of porous media under an electric field. For instance, between clays and gravel, resistivity thresholds can be clearly defined based on well-understood physical characteristics (grain size, pore geometry, grain density, tortuosity) and chemical properties (mineral composition, CEC), along with saturation conditions and pore fluid salinity.
This statement is wrong as demonstrated by Hermans and Irving (2017), I really encourage you to read that paper in details. Indeed, you might be able to discrimante a low conductivity clay from a resistive gravel, but your geological settings also has intermediate sand facies. Your error comes from the fact that you are confusing true resistivity values with inverted resistivity values. I invite you to plot the inverted resistivity versus the true resistivity for your synthetic case (Figure 4), and you would likely observe an overlap of inverted resistivity between most facies. This all comes from the limitation of regularized inversion. Eventually, it might not make a big difference for the estimation of hydrofacies characteristic length, but it should be clear that any length you estimate from ERT can be impacted by the inversion process. Indeed, the countour lines you use certainly do not correspond to the real interfaces. And this is the case even if if the choice of your inversion parameters (smoothness constraint inversion with R2) is properly done. What is needed to understand this is present in Figure 4 and can be easily included in the text.
4. You write: In fact, it effectively resolves the geometric characteristics of hydrofacies with minimal residual error, meaning it fits the synthetic resistivity data very well.
You should be careful with what you do with the synthetic case. The only thing you can do is confirm that the proposed geometry in the synthetic case can lead to a similar resistivity distribution as observed after inversion. It does not mean that this interpretation is actually correct. There are plenty of other possibilities that could lead to similar results. This is the definition of a non-unique solution.
I have some suggestions to further improve the manuscript.
1. L64. “Unregulated” or “excessive” instead of “irresponsible”. The latter implies decision despite the knowledge it could be harmful, while I assume it might not be the case.
2. L76. I would add “as conditioning data” and cite some of the papers where it has been done like Hermans et al. (2015, aready cited) or Barfod et al. (2018, 10.5194/hess-22-3351-2018, 10.5194/hess-22-5485-2018). This is really important to properly guide the reader where more advanced conditioning has been proposed. I would also come back on this in your summary/conclusion since there is no discussion section.
3. L127-131/ See my comments above on your response. You should highlight that the choice of the contour line bears some subjectivity as you don’t know the real interface location (borehole are projected) and that you interpret the gradual transition from high to low resistivity. You cannot be certain about that and it is important you acknowledge it explicitly in the manuscript even if the impact is likely limited in your methodology as you are not using ERT as conditioning data (that can also be underlined in the manuscript).
4. L132. Space between “are” and “using”.
5. L140. Space between “data” and “and”.
6. L161. Refer to Table 1 when mentioning proportions.
7. L199-200. If a high resistivity is next to a low resistivity, then you will always see a transition to intermediate, even in the absence of the intermediate facies. This is why relying on a single contour line or limited number of boreholes might be misleading. In figure 4, are you therefore always sure there is Gsc between G and Sgcs? I would not be.
8. L201. Replace “strong continuity” by “good/satisfactory/acceptable consistency”.
9. L212. I presume that the 0.1 is in the log scale and not the resistivity scale in ohm.m.
10. L212. A kriging interpretation requires a variogram model, that should then be mentioned. I assume however a simple linear interpolation would do the same job (this is basically a downscaling approach).
11. L214-216. It is still not fully clear to me. As I understand, the boreholes are projected onto the profiles, there is quite some uncertainty about these transitions. It is ok for your purpose that you define some threshold, but these should not be overinterpreted, and the amount of subjectivity involved should be acknowledged. As mentioned before, I really doubt that you have the exact same isocontour in all boreholes, even if you had co-located data. If it was the case, this would really be a coincidence that would not be validated if you had more boreholes available. Stating otherwise would be in strong contradiction with the existing literature. This is actually implicitly acknowledged in line 200-222 since you have different values of the contour line.
12. L234-236. It would be nice to show some of these tests, so that the reader can have a better feeling of the methodology.
13. L247-248. This is also linked to their thicknesses.
14. L255-256. you don't know. You just show that your synthetic model after inversion gives something similar, it does not mean that the true model resistivity is that one. But indeed, the discrimination potential of ERT is high close to the surface (Hermans and Irving, 2017).
15. L266-267. I am pretty sure one could fine a synthetic model with a continuous lens to represent this. Maybe it would involve variation in resisitivity values like you used for the different gravel lenses. Synthetic modelling tells you this situation is possible, not that other geometries are impossible.
16. L290-291 + conclusion (point 1/3). It could also indicate that transition probability is not sufficient to capture the spatial patterns ? Maybe methods which better deal with multiple-point statistics are needed.
17. L309-310. “Two validation boreholes in R8 achieved 100 % prediction accuracy (Fig. 6b), both located in the northwestern part of the study area.” It is not relevant, realizations are equiprobable and should not be ranked based on the validation.
18. L329. What does “and any potential ambiguities from synthetic modelling at greater depths” mean? What are the ambiguities from synthetic modelling?
19. L365-369. I fear you are here interpreting variations that are not statistically relevant. Have you run a statistical test?
20. L385-393. Here, you should mention that more advanced approaches exist to constrain geostatistical models with resistivity, where resistivity values are used as a soft constraint. There is no doubt such an approach, if resistivity values are broadly available, would result in better identification of facies.
21. L401-402. I don't know what robust means in this sentence... ERT is just use to adapt the length of the facies, so there is not reason why TPROGS would fail taking this into account.
22. L408-409. Sentence about R8 is not relevant, see above, especially with such a low number of realizations.
23. L416-417. Suggesting a coarse grid which exceeds the size of identified geometry is strange. I would not do that, especially given the small differences in obtained distributions. In addition, eventually, the size would rather be linked to the requirement of groundwater modelling.
Sincerely,
Thomas Hermans |
Dear authors,
I read with interests this paper entitled: “Integrated approach for characterizing aquifer heterogeneity in alluvial plains”. In this article, a methodology is proposed to integrate geophysical data as a constraint for geostatistical simulations meant for generating realistic realizations of alluvial aquifer heterogeneity. The topic is relevant, as these aquifers are amongst the most complex to characterize, while there are often exploited for drinking water production and highly contaminated in and around cities due to industrial activity. I support any effort related to a better characterization of these complex systems. In that sense, the paper is interesting, but from my point of view, the reader is left with a feeling of unfulfilled expectations. My major concerns are described below:
Specific comments:
References
Baines D., Smith D.G., Froese D.G., Bauman P. and Nimeck G. 2002. Electrical resistivity ground imaging (ERGI): a new tool for mapping the lithology and geometry of channel-belts and valley-fills. Sedimentology 49(3), 441–449.
Barfod, A.A.S., Vilhelmsen, T.N., Jørgensen, F., Christiansen, A.V., Høyer, A.-S., Straubhaar, J., Møller, I., 2018. Contributions to uncertainty related to hydrostratigraphic modeling using multiple-point statistics. Hydrol. Earth Syst. Sci. 22, 5485–5508. https://doi.org/10.5194/hess-22-5485-2018.
Bersezio R., Giudici M. and Mele M. 2007. Combining sedimentological and geophysical data for high resolution 3D mapping of fluvial architectural elements in the Quaternary Po Plain (Italy). Sedimentary Geology 202, 230–248
Bowling J., Harry D., Rodriguez A. and Zheng C. 2007. Integrated geophysical and geological investigation of a heterogeneous fluvial aquifer in Colombus Mississippi. Journal of Applied Geophysics 65, 58–73
Bowling J., Rodriguez A., Harry D. and Zheng C. 2005. Delineating alluvial aquifer heterogeneity using resistivity and GPR data. Ground Water 43(6), 890–903.
Caterina, D., Beaujean, J., Robert, T., Nguyen, F., 2013. A comparison study of different image appraisal tools for electrical resistivity tomography. Near Surface Geophysics 11, 639–657. https://doi.org/10.3997/1873-0604.2013022
Day-Lewis, F.D., Lane, J.W.Jr., 2004. Assessing the resolution-dependent utility of tomograms for geostatistics. Geophysical Research Letters 31, L07503. https://doi.org/10.1029/2004GL019617
Doetsch J., Linde N., Coscia I., Greenhalgh S.A. and Green A.G. 2010. Zonation for 3D aquifer characterization based on joint inversions of multimethod crosshole geophysical data. Geophysics 75(6), G53G64.
Doetsch J., Linde N., Pessognelli M., Green A.G. and Günther T. 2012a. Constraining 3-D electrical resistance tomography with GPR reflection data for improved aquifer characterization. Journal of Applied Geophysics 78, 68–76.
Enemark, T., R. B. Madsen, T. O. Sonnenborg, L. T. Andersen, P. B. E. Sandersen, J. Kidmose, I. Moller, T. M. Hansen, K. H. Jensen, and A.-S. Hoyer (2024). “Incorporating interpretation uncertainties from deterministic 3D hydrostratigraphic models in groundwater models”. In: Hydrology and Earth System Sciences 28.3, pp. 505–523. doi: 10.5194/hess-28-505-2024. url: https://hess.copernicus.org/articles/28/505/2024/.
Gottschalk, I.P., Hermans, T., Knight, R., Caers, J., Cameron, D.A., Regnery, J., McCray, J.E., 2017. Integrating non-colocated well and geophysical data to capture subsurface heterogeneity at an aquifer recharge and recovery site. Journal of Hydrology 555, 407–419. https://doi.org/10.1016/j.jhydrol.2017.10.028
Hermans, T., Irving, J., 2017. Facies discrimination with ERT using a probabilistic methodology: effect of sensitivity and regularization. Near Surface Geophysics 15, 13–25.
Hermans, T., Nguyen, F., Caers, J., 2015. Uncertainty in training image-based inversion of hydraulic head data constrained to ERT data: Workflow and case study. Water Resources Research 51, 5332–5352. https://doi.org/10.1002/2014WR016460
Isunza Manrique, I., Caterina, D., Nguyen, F., Hermans, T., 2023. Quantitative interpretation of geoelectric inverted data with a robust probabilistic approach. GEOPHYSICS 88, KS73–KS88. https://doi.org/10.1190/geo2022-0133.1
Mariethoz, G., Renard, P., Straubhaar, J., 2010. The Direct Sampling method to perform multiple-point geostatistical simulations. Water Resources Research 46. https://doi.org/10.1029/2008WR007621.
Mastrocicco M., Vignoli G., Colombani N. and Zeid N.A. 2010. Surface electrical resistivity tomography and hydrogeological characterization to constrain groundwater flow modeling in an agricultural field site near Ferrara (Italy). Environmental Earth Sciences 61(2), 311–322.
Strebelle, S., 2002. Conditional Simulation of Complex Geological Structures Using Multiple-Point Statistics. Mathematical Geology 34, 1–21.