the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 10 Oct 2024
Quantitative soil characterization using frequency domain electromagnetic induction method in heterogeneous fields
Abstract. The frequency domain electromagnetic induction (FDEM) method is a widely used tool for geophysical soil exploration. Field surveys using FDEM provide apparent electrical conductivity (ECa), which is typically used for qualitative interpretations. Quantitative estimations of soil properties remain challenging, especially in heterogeneous fields. Quantitative approaches are either based on deterministic or stochastic modeling. While the deterministic approach faces limitations related to instrumental drift, data calibration, inversion, and pedophysical modeling, the stochastic approach requires developing a local model, which involves extensive field sampling.
This study aims to evaluate the effectiveness of the FDEM modelling based on either a deterministic or stochastic approach, identify its limitations, and search for optimal field protocols. We provide practical guidelines for end-users to quantitatively predict soil water content, bulk density, clay content, cation exchange capacity, and water EC in heterogeneous fields.
Two field surveys were conducted in Belgium, where FDEM data was collected using Dualem-421S and Dualem-21HS sensors, along with data taken from electrical resistivity tomography (ERT) measurements and an impedance moisture probe, and soil sampling.
A comprehensive sensitivity analysis revealed that deterministic modeling procedures could not predict water content more accurately than a mean value approximation (negative R2). This analysis also highlighted the sensitivity of the minimization method used in FDEM data inversion and the applied pedophysical model. Stochastic modeling, which does not require FDEM data calibration or inversion, outperformed the deterministic approach. However, its prediction accuracy is limited, particularly if soil sample depth is not considered.
- Preprint
(1759 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2024-2693', Jacopo Boaga, 24 Oct 2024
The paper deals with the use of FDEM data for the quantitative characterisation of the first subsoil. The topic is of interest to the journal and the case study presented is a relevant example of an advance in the use of the EMI technique. Results are supported by the data presented and I encourage the publication. I suggest some minor revisions to be addressed before:
- Ln 72 a ratio of the same quantity is by definition a-dimensional, so please remove ppm
- some part of the paper cite ppm, other ppt (fig2), please homogenise it
- All the equations labels have layout issue that do not agree with the journal editing rules
- Eq1 QP means the ratio of the imaginary part of primary and secondary fields ? Please specify
- Ln 202-205 please provide references for the different polynomial approaches
- Ln 232 please provide the used open-source software reference
-Fig.5 To help reading I suggest to invert ideal and FDEM inverted EC columns, so reader can appreciate similarity between the former and the probe
- Ln 304 missing dot before 'However' ?
- Ln 300-308 all this paragraph is not clear and should be re-written
Citation: https://doi.org/10.5194/egusphere-2024-2693-RC1 -
AC1: 'Reply on RC1', Gaston Matias Mendoza Veirana, 15 Nov 2024
RC1:
The paper deals with the use of FDEM data for the quantitative characterisation of the first subsoil. The topic is of interest to the journal and the case study presented is a relevant example of an advance in the use of the EMI technique. Results are supported by the data presented and I encourage the publication. I suggest some minor revisions to be addressed before:
Comments:
RC1.1
- Ln 72 a ratio of the same quantity is by definition a-dimensional, so please remove ppm
Response: thanks for your suggestion. Despite indeed a ratio of the same quantity is a-dimensional, ‘ppm’ refers to parts per million (1*10-6) and does not refer to a unit.
RC1.2
- some part of the paper cite ppm, other ppt (fig2), please homogenise it
Response: Right, this was corrected to ppm.
RC1.3
- All the equations labels have layout issue that do not agree with the journal editing rules
Response: all the equations were adapted to the journal’s format
RC1.4
- Eq1 QP means the ratio of the imaginary part of primary and secondary fields ? Please specify
Response: this is now defined in the previous sentences. QP represents the imaginary component of the ratio between secondary and primary field.
RC1.5
- Ln 202-205 please provide references for the different polynomial approaches
Response: we reformulated this paragraph, now it reads:
‘Three distinct approaches to polynomial development were utilized. A first approach, named “Layers Together” (ST-LT) consisted of combining data from different soil depths, so that no differentiation was made between top- and subsoil samples for model development. Secondly, these sample sets were considered separately in an approach whereby different polynomials were developed for each soil layer (“Layers Separate”(ST-LS)). In this modelling approach, the same polynomial degree was maintained for both top – and subsoil data. Finally, the ST-LS2 approach was like ST-LS but permitted different polynomial degrees for the models of each layer.’
RC1.6
- Ln 232 please provide the used open-source software reference
Response: The reference was added.
Mendoza Veirana, G., & Philippe De Smedt. (2024c). orbit-ugent/Pedophysics: First release 0.1 (Version 0.1) [Computer software]. Zenodo. https://doi.org/10.5281/ZENODO.13465700RC1.7
-Fig.5 To help reading I suggest to invert ideal and FDEM inverted EC columns, so reader can appreciate similarity between the former and the probe
Response: Fig 5 is updated
RC1.8
- Ln 304 missing dot before 'However' ?
Response: changed as suggested
RC1.9
- Ln 300-308 all this paragraph is not clear and should be re-written
Response: changed as suggested:
‘When the best performing models for 𝜃 prediction are implemented using the entire dataset (Figure 7) – using both training and test data – this outperforms the modelling presented in Figure 6, where only test data are incorporated in error assessment. While this is a common approach, we want to highlight this is improper practice to critically evaluate model performance as the inclusion of training data in error estimation results in an overestimation of model performance (Altdorff et al., 2017; Lipinski et al., 2008; Tibshirani et al., 2001). In other words, implemented model errors should not be confused with actual expected accuracy of target property predictions.
To evaluate the influence of other soil properties in 𝜃 prediction, the residuals of the implemented stochastic models were correlated with other soil properties, but these were not significant.’
Citation: https://doi.org/10.5194/egusphere-2024-2693-AC1
-
AC1: 'Reply on RC1', Gaston Matias Mendoza Veirana, 15 Nov 2024
-
RC2: 'Comment on egusphere-2024-2693', Anonymous Referee #2, 03 Dec 2024
Dear authors,
I have read, your paper titled “Quantitative soil characterization using frequency domain electromagnetic induction method in heterogeneous fields” with interest.
The authors present a study, where frequency domain electromagnetic induction (FDEM), ERT and soil sampling measurements are used to evaluate the feasibility of FDEM modelling based on either a deterministic or stochastic approach to quantitative derive soil properties.
Overall, the paper is well structured and has a good readability. The topic is suitable for Hydrology and Earth System Sciences.
For enhances readability it would be helpful to add some tables or numeration as some points throughout the manuscript, see specific comments.
The methodology section needs refinement. There is an overwhelming amount of steps and I feel like there is some information on the test sites missing. Figure 2 is somewhat helpful but the steps shown in don’t quite align with the text, e.g. the LIN transformation is missing from the text. Or use the same wording (maybe also in the section headers).
SPECIFIC COMMENTS
Line(s) 72: ppm – no abbreviation explanation (I think it is not necessary to explain but in Line 143, you explain ppt, so maybe be consistent)
Line(s) 85 – 90: What are the land used for these test sites? Are the agriculturally used? Does a plow horizon exist? Was there any crop present while the measurements were performed? I feel like in Figure 1 A, there can be two different field be disguised? The northwest part of the site seem the show a higher rECa compared to the part south of it.
Line(s) 93: When you say driving speed. Was the device pulled with a tractor, quad? How did you ensure the distance from the ground?
Line(s) 94: By crossline density you mean crossline sampling density? Maybe use consistent wording here.
Line(s) 96: Is PRP the same as VCP (vertical coplanar loops)? If yes I think VCP is more commonly used.
Line 96 – 99: May it is worth putting this into a table for a better overview?
Line(s) 101 – 102: So there were multiple measurement? Could elaborate on that, when where they measurement and how often. The exact would be important to get a glimpse of the overall water content (summer cs. Winter)
Line(s) 121 – 122: Maybe use an enumeration? E.g., i.), ii.)?
Line(s) 122: Does the pedophysical model have a certain name?
Line(s) 142: The combining of the ERT and FDEM data is done on which step in the flow chart in Figure 2?
Line(s) 144: LIN was introduced earlier already. Is LIN approximation the same as LIN transformation (Figure 2)
Line(s) 146: How does shortening the profiles have an influence on the number of profiles?
Line(s) 155: This figure already shows results but the section is located in the methodology section. May be move to results and discussion section?
Line(s) 199-201: But aren’t soil probes multiple times represented? Maybe elaborate a little on this, for readers who are not too familiar with this method.
Line(s)208: abbreviation CEC has not been mentioned before
Line 270: This is only a suggestion. Maybe it is easier to compare the different ways to derive the EC when they are in figure. You could split for the two soil layers and then use different Markers for the analysis types?
Line(s) 275 – 280: Is valid also throughout the manuscript. I would recommend using the wording of stochastic and deterministic modeling in connection with the different ECs. To remind the reader that this is the overarching goal to compare the two and to derive the feasibility.
Line(s) 273: why not show this in the figure?
Line(s) 300. In the lower left plot the numbers overlapping with the bars are very hard to read. Move them above the bars, as the other subplots.
Line(s) 310: Why does it say in the legend ‘Layer 10 cm’ & ‘Layer 50 cm’? Do you mean topsoil and subsoil.
Citation: https://doi.org/10.5194/egusphere-2024-2693-RC2 -
RC3: 'Comment on egusphere-2024-2693', Damien Jougnot, 03 Dec 2024
The manuscript “Quantitative soil characterization using frequency domain electromagnetic induction method in heterogeneous fields” by Mendoza Veirana et al., under open discussion for a publication in HESS, deals with the difficult task of providing a quantitative use of frequency domain electromagnetic (FDEM) method. The author propose a novel and interesting framework to achieve this goal and show fairly convincing results that are, in my opinion, worthy of publication in HESS. My main comment of this work is the lack of discussions regarding the methods’ resolution/footprint. The rest of the comments are rather minor and detailed below.
Damien Jougnot
CNRS Senior researcher
Sorbonne UniversityList of comments:
Abstract: it is more usual to the abstract in a single (or two maximum) paragraph.
Introduction: In my opinion, one limitation which not mentioned (or not explicitly enough) is related to the method footprints/resolutions. It is a tricky problem to solve, but when comparing sample characteristic to geophysical field-data, one always have to deal with the curse of scale (see, for example, the seminal paper of Day-Lewis et al., 2005). Every researcher looking for quantitatively interpreting field measurements is facing this issue of comparing things at different scales (from sample to the geophysical pixel, but also within a tomogram itself). I’m not pretending that I have a way to tackle that (I did try to study these mesoscopic heterogeneities at a very controlled scale with the highest possible resolution, I could only show that in heterogeneous media, petrophysical relationship are not always valid and should be considered with care, e.g., Jougnot et al. 2018), but I think that this kind of limitation/issues should also be mentioned in the introduction and later discussed in the paper.
Line 46-47: I suggest additional reviews dedicated to soil electrical properties: Samouëlian et al. (2005) and/or Friedman (2005).
Methodology: I really enjoyed the methodology chosen by the authors, I find it very robust and state-of-the-art.
Line 80: It could be nice to provide examples/references of the fact that many soils violate the low induction number requirement. Showing that such framework is useful, especially when dealing with soils containing a clay fraction and trying to estimate this fraction.
Line 105: Why not measuring the electrical conductivity of the samples ? Since the authors analysed a lot of interesting and relevant properties, it would have been fun to test petrophysical relationship (like the one used later in the manuscript).
Line 134-135: ResIPy is a great tool and the authors could provide a link in addition to the reference, especially for the Jupyter Notebook.
Figure 3: the authors should provide letters for the subplots.
Figure 4: the fact that rECa is systematically above the forward ERT model could be discussed, it is not clear to me why (possibly later in the paper)?
Line 314: the subsection number should be 3.4
Subsection 2.3.4: I am more of a deterministic guy but the stochastic approach described here is clearly interesting.
Line 232: The authors should provide a reference and a link for the Pedophysics open-source software.
Section 4: As mentioned before, another strong limitation that should, in my opinion, be discussed, is related to the footprint differences between the methods (sampling, probe, ERT and FDEM). The water saturation or the clay content defined at the sample scale are likely to be different from the probe or geophysical method footprint scale (even without considering the smoothing related to inversion procedure). I think that this should be discussed as a limitation.
Refrences:
Day‐Lewis, F. D., Singha, K., & Binley, A. M. (2005). Applying petrophysical models to radar travel time and electrical resistivity tomograms: Resolution‐dependent limitations. Journal of Geophysical Research: Solid Earth, 110(B8).
Friedman, S. P. (2005). Soil properties influencing apparent electrical conductivity: a review. Computers and electronics in agriculture, 46(1-3), 45-70.
Jougnot, D., Jiménez-Martínez, J., Legendre, R., Le Borgne, T., Méheust, Y., & Linde, N. (2018). Impact of small-scale saline tracer heterogeneity on electrical resistivity monitoring in fully and partially saturated porous media: Insights from geoelectrical milli-fluidic experiments. Advances in Water Resources, 113, 295-309.
Samouëlian, A., Cousin, I., Tabbagh, A., Bruand, A., & Richard, G. (2005). Electrical resistivity survey in soil science: a review. Soil and Tillage research, 83(2), 173-193.
Citation: https://doi.org/10.5194/egusphere-2024-2693-RC3
Data sets
Data set. Quantitative soil characterization using frequency domain electromagnetic induction method in heterogeneous fields Gaston Matias Mendoza Veirana et al. https://doi.org/10.5281/zenodo.13465721
Model code and software
FDEM_quantitative_soil Public Gaston Matias Mendoza Veirana et al. https://doi.org/10.5281/ZENODO.13385389
Interactive computing environment
FDEM_quantitative_soil Public Gaston Matias Mendoza Veirana et al. https://doi.org/10.5281/ZENODO.13385389
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 198 | 72 | 23 | 293 | 4 | 3 |
- HTML: 198
- PDF: 72
- XML: 23
- Total: 293
- BibTeX: 4
- EndNote: 3
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1