the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 23 Sep 2026
A National Depth-Resolved Soil-Moisture-to-Electromagnetic Proxy Database for Hydrogeophysical Monitoring
Abstract. Soil moisture is a key control on drought, recharge, crop water availability, agricultural water management, and land-atmosphere exchange, but its depth-resolved spatial variability remains difficult to monitor over large areas. Non-invasive methods such as ground-penetrating radar (GPR), electrical resistivity, and electromagnetic surveys can support soil-moisture estimation, yet their interpretation depends strongly on soil texture and hydraulic-retention properties. This study develops a national six-depth soil-moisture-to-electromagnetic proxy database to support future GPR- and resistivity-based root-zone monitoring and decision support. We used multi-year soil-moisture, porosity, saturation, texture, and hydraulic-retention data from 117 monitoring stations across Hungary at 10, 20, 30, 45, 60, and 75 cm depth. Dielectric permittivity, EM-wave velocity, electrical conductivity, attenuation, and apparent resistivity were derived as proxy variables from observed soil moisture using established petrophysical relationships, including the Topp equation and an Archie-type formulation. The proxy database identified a consistent 30-45 cm buffering zone, where soil moisture increased to 22.25%, dielectric permittivity peaked at 12.29, EM-wave velocity reached a minimum of 0.0935 m ns⁻¹, and apparent resistivity decreased to 545 Ω m. Texture strongly shaped the translated EM response: sandy soils were drier and more resistive, whereas clayey soils retained more water and showed higher dielectric response. Hydraulic-retention variables provided calibration-relevant prior information for explaining why similar EM or resistivity signals may represent different true soil-moisture values across soil types. The framework provides a transferable national calibration layer for drought-monitoring agencies, irrigation planners, precision-agriculture users, and future drone-based geophysical surveys in regions with comparable monitoring and soil databases.
- Preprint
(3814 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 26 Nov 2026)
-
RC1: 'Comment on essd-2026-563', Anonymous Referee #1, 22 Oct 2026
reply
-
AC1: 'Reply on RC1', Diaa Sheishah, 27 Oct 2026
reply
Response to Reviewer 1
Manuscript essd-2026-563
We thank the reviewer for recognizing the value of the national depth-resolved dataset and for identifying areas requiring stronger documentation and more cautious interpretation. We have revised the submitted manuscript accordingly and provide a point-by-point response to each comment below.
Comment 1 The manuscript is based on a valuable national-scale dataset combining multi-year, depth-resolved soil-moisture observations from 117 monitoring stations with information on soil texture, porosity, hydraulic-retention properties and related environmental variables across six depths. In my view, this dataset represents the strongest contribution of the study. I therefore see clear potential for publication of the dataset. The underlying CSV data files are accessible, appear complete and are generally well structured and understandable. In this respect, I consider the dataset itself potentially suitable for publication as a data contribution. However for a data paper, I would expect a much stronger focus on data provenance, variable definitions, quality-control procedures, uncertainty and the distinction between measured, processed and model-derived quantities. The CSV files themselves are generally understandable and usable, but formal metadata are missing and the provided Python scripts are only sparsely documented. This makes independent reuse and reproduction of the processing workflow more difficult than necessary.
Response: We agree that the national dataset is the manuscript’s central contribution. Therefore, the revised manuscript now defines three product classes - measured/source-reported, processed, and model-derived - documents the source period and station/depth coverage, reports QC counts, states that invalid negative values were masked and missing soil moisture was not interpolated, and supplies a data dictionary and documented robustness workflow.
Change in the manuscript: Revised Abstract; Sects. 2, 3.1.1, and Data Availability; accompanying Metadata and Data Dictionary and Reproducibility folder.
Comment 2 A substantial part of the article focusses on interpreting model-derived hydrogeophysical proxies, integrated indices, spatial patterns and process-related conclusions. At this point I have two major concerns regarding the interpretation of the derived geophysical proxies and the proposed 30–45 cm buffering zone that require restructuring the manuscript and reanalysis of the underlying dataset. From my point of view an independent validation of the derived geophysical proxies would be essential if the authors want to use them as a central part of the manuscript. Dielectric permittivity, EM-wave velocity, electrical conductivity, attenuation and apparent resistivity are not independently measured but are derived from soil moisture and other soil properties through assumed petrophysical relationships – the authors state that but forming a transferable national calibration layer out of this is not appropriate in my view. The present analysis is a model-based translation from observed soil conditions to expected geophysical responses, but it does not demonstrate that these responses correspond to actual GPR, ERT, EM, dielectric or laboratory measurements. This distinction is fundamental and should be reflected much more clearly in the manuscript.
Response: We agree with this central distinction. The products are now consistently called conditional model translations or scenarios intended for hypothesis generation and survey design. The revised manuscript states explicitly that site-specific field or laboratory validation is required before quantitative operational use.
Change in the manuscript: Revised Abstract; Sects. 1, 3.1, 3.4–3.6, 4.1–4.6, 5.1–5.5, and Conclusion; revised figure captions.
Comment 3 A related major concern is the lack of a systematic uncertainty assessment. The derived proxy values depend on model choice and parameterization as well as on assumptions regarding pore-water properties, porosity, saturation, soil texture and other variables. These uncertainties are likely to vary considerably among sites and soil classes. Reporting single proxy values (e.g., with an accuracy of one decimal place for the resistivity values) without uncertainty or sensitivity ranges implies a degree of accuracy and transferability that is not supported by the underlying methodology. A detailed sensitivity and uncertainty analysis is needed to quantify how parameter uncertainty and model assumptions propagate into the derived electromagnetic quantities.
Response: We added a reproducible dielectric sensitivity analysis. Volumetric water content was perturbed by ±0.02 m³ m⁻³ before applying the Topp equation, yielding an overall εᵣ envelope of 10.09–13.17 and a v envelope of 0.0866–0.0984 m ns⁻¹. We also compared Topp and CRIM at each depth (mean differences approximately 0.4–2.2%). We did not report numerical electrical uncertainty bounds because the required pore-water and surface-conductivity measurements are unavailable; those outputs are instead flagged as conditional scenarios and excluded from quantitative clay/fine-soil claims.
Change in the manuscript: New Sect. 3.7; new Results text in Sect. 4.1; new Fig. S9; revised limitations in Sect. 5.3.
Comment 4 I’m concerned about the use of an Archie-type relationship for clayey and fine-textured soils. In my view, a classical Archie formulation should not be applied as a quantitative proxy model to soils with clay content, because surface conductivity associated with clay minerals violates a central assumption of the model. Under such conditions, electrical conductivity is no longer primarily controlled by pore-water conductivity and saturation. Assigning texture-specific Archie parameters doesn’t resolve this fundamental limitation. For clayey soils, the authors should use a petrophysical formulation that explicitly accounts for surface conductivity and clay-related charge transport or refrain from deriving quantitative proxies for these soil classes. This is important because the Archie-derived quantities are subsequently used to compare soil-texture classes, calculate integrated indices, identify spatial patterns and support broader hydrogeophysical interpretations. Any systematic bias introduced by an inappropriate petrophysical model is therefore propagated into several downstream analyses.
Response: We agree. Assigning texture-specific Archie exponents does not address the absence of a surface-conduction term. Because pore-water conductivity and clay surface conductivity measurements are unavailable, a clay-aware model cannot be calibrated defensibly. We retained the original electrical fields and figures only to preserve the submitted database/workflow and reproducibility, but now label ρa, σ, α, and Archie-dependent indices as conditional scenarios. We no longer use absolute clayey or fine-loam electrical values as quantitative evidence, and we state the measurements needed for future validation.
Change in the manuscript: Revised Sects. 3.1, 3.4, 4.2, 4.4–4.6, 5.2–5.3, and Conclusion; revised captions for Figs. 4–5, 7–12, S1, S6, and S7.
Comment 5 The interpretation of the 30–45 cm interval as a nationally consistent hydrogeophysical “buffering zone” appears insufficiently supported or at least I don’t understand it out of the manuscript. The proposed zone is primarily based on a relatively small maximum in measured soil moisture at the discrete 30 and 45 cm measurement depths. The associated dielectric, velocity, conductivity, attenuation and resistivity patterns are derived from this same soil-moisture signal. Therefore these geophysical proxies don’t provide independent confirmation of the existence of a distinct process-based layer. I find it surprising that the physical origin of this potentially very interesting vertical pattern is not explored in greater detail. The monitoring network is described as covering different land-use settings, but land use isn’t systematically included in the analysis. In agricultural soils, the 30–45 cm interval could coincide with the lower boundary of the cultivated layer, a compacted plough pan or other management-related structural changes affecting porosity, hydraulic conductivity, rooting and water retention. Furthermore the manuscript does not investiges whether the observed moisture maximum corresponds to pedological horizons, changes in bulk density, root distribution, soil structure or hydraulic properties. These explanations should be investigated before interpreting the observed pattern as a generic buffering mechanism. I suggest, the authors should test whether the mid-profile maximum persists when the data are separated by land-use class, for example cropland, grassland and other land uses. Where such information is available, the pattern could also be related to soil horizons, bulk density, rooting characteristics and management or tillage depth. It could also be useful to assess whether the maximum occurs consistently at individual stations or whether the apparent 30–45 cm zone partly emerges from averaging a heterogeneous national dataset.
Response: We performed the requested station and station-year tests. Only 32.5% of stations and 33.7% of station-years have their maximum at either 30 or 45 cm; only 29.9% of stations have a mid-profile mean above both upper and deep layers. The mid-minus-upper contrast is +0.63 percentage points (95% station-bootstrap CI 0.18 to 1.09), whereas mid-minus-deep is +0.14 (−0.40 to 0.68). We therefore no longer call the feature a nationally consistent or generic buffering zone. Harmonized data on land use, soil horizons, bulk density, rooting, and tillage depth are not available for all stations, so we report that limitation instead of inferring unsupported classes.
Change in the manuscript: New Sect. 3.7 and Fig. S8; revised Sects. 4.1, 4.5, 5.1, 5.5, Abstract, and Conclusion.
Comment 6 I had no closer look at the Machine-learning analysis as the above mentioned items required an enhanced revision from my point of view.
Response: We appreciate this clarification. This analysis has a useful and distinct role in a data paper: it tests whether source-reported retention, grain-size, and texture variables can reproduce the ranges of the model-derived targets without using SM or any SM-derived variable as a predictor. We have nevertheless narrowed the claims so that high R² is described as proxy-emulation skill, not independent geophysical validation. To strengthen the section with a measured target, we also added a supplementary repeated cross-validation analysis in which the same independent soil-property predictors estimate station-mean measured SM. This provides an out-of-sample data-reuse check without claiming validation of GPR, ERT, EM, or laboratory response. The source-verified 60 cm ρa value was also corrected from approximately 0.91 to 0.86.
Change in the manuscript: Retained and revised Sects. 3.6 and 4.6; added supplementary measured-SM repeated-CV analysis and Figs. S11–S15; revised Sect. 5.5; retained Figs. 10–11 and S4–S7 with clarified captions.
Comment 7 The figures are very difficult to read in their current form. Axis labels, tick labels, legends, annotations and in some cases panel labels are too small and should be revised.
Response: We agree. We audited all code files in the project and identified the active plotting scripts. Font sizes were increased in the manuscript figure generators, and every figure was re-exported at 600 dpi. Figure 5 remains one composite figure, as in the submitted manuscript, but was regenerated on a larger canvas with larger axis labels, tick labels, depth labels, titles, and panel letters, with spacing adjusted to avoid overlap. We also retained the original figures and added the reviewer-requested diagnostic and sensitivity figures alongside them in the Supplement (Figs. S8–S15).
Change in the manuscript: Revised all manuscript figures and active plotting scripts; retained Fig. 5 as one enlarged composite; added Figs. S8–S15; re-embedded all figures in the clean and tracked manuscripts.
Comment 8 In summary, I see a considerable value in the soil-moisture and soil-property dataset, but the hydrogeophysical interpretation requires substantial strengthening/revision. In my view, publication of the geophysical proxies requires independent validation against actual geophysical or laboratory measurements, together with or at least considering explicit propagation of model and parameter uncertainty. The use of Archie-type relationships for soils with clay content should be avoided. The 30–45 cm moisture maximum is an interesting finding, but its interpretation as a generic hydrogeophysical buffering zone should be supported by a more detailed analysis of land use, soil-profile architecture and site-specific vertical patterns.
Response: We have followed this direction throughout the revised manuscript. The restructuring was limited to a new provenance/QC subsection and a new robustness/sensitivity subsection, plus localized interpretive corrections. The revised article now foregrounds the dataset and transparently separates evidence from scenario-based interpretation.
Citation: https://doi.org/10.5194/essd-2026-563-AC1
-
AC1: 'Reply on RC1', Diaa Sheishah, 27 Oct 2026
reply
-
RC2: 'Comment on essd-2026-563', Luca Peruzzo, 30 Oct 2026
reply
Dear Authors, Dear Editor,
Thank you for your consideration. I have read the manuscript and comments from the first Reviewer. I hope my review will be adequate and helpful.
Comments
L27 "established petrophysical relationships", I agree but my perspective is that Topp and Archie are not necessarily suitable and likely require field-specific validation or calibration. This is clearly and critically the case for Archie, as highlighted by R1. However, Topp equation can lead to similar issues too, see already Dirksen and Dasberg, 1993.
Sometimes, these models are still applied beyond their expected suitable contexts, but 1) this is not a good idea or accepted and 2) may be defendable on a site-specific basis after validation or calibration. This work, however, explores a wide range of conditions because of natural variability and intrinsic general scope of a national proxy database. This conflicts with the necessary site-specific investigation and consequent lack of supporting rational, calibration or validation. In this sense the geophysical part seems to me an extensive forward model exercise based on well established and widely explored relationships, in which the set of variables is constrained and tuned to Hungary conditions.
I think the authors are already well aware of the above, see for example L21 "interpretation depends strongly on soil texture and hydraulic-retention properties". Yet, this fundamental issue remains and in my opinion strongly hinders the quality of the final geophysical datasets.
L36 "transferable national calibration layer", in line with the above, I think this is misleading: 1) the used relationships are not generally valid, 2) the geophysical results are not validated or calibrated, 3) even if that was the case, the results would help but would not automatically provide a reliable calibration layer. See also R1 comments on uncertainty. Also, the same electrical property would surely depend on the measurement technique (e.g., drone FDEM vs ERT), scale (incomprehensibility), etc. This also hinders the generalization and the definition of a common depth resolution; hence, I find this generally misleading and kind of glossy.
L75 "nonlinear and context-dependent", these paragraphs on soil hydrology read well and I found them are technically objective.
L86 "robust" Considering the above and the machine learning additional complexity, it seems a bit subjective to define such solutions robust. Nothing against AI, but at the same time AI solutions cannot shadow underlying petrophysical challenges.
L98 "The lack of a national, depth-resolved reference database...". My objection is that it would require actual geophysical measurements, combined with relevant soil characterizations, as the authors introduced. The two things need to go together to provide new and significant information.
L101 "Second, the role of texture and hydraulic retention in controlling water storage is still not sufficiently translated into practical calibration information for estimating true SM from non-invasive EM or resistivity measurements." I partially agree, although a lot of literature exists on the relationship between retention-curve quantities and electrical conductivity or permittivity, e.g., TDR and GPR corrections based on clay contents, or hydraulic quantities from IP measurements. I agree that both aspects are not solved, despite the decades of efforts, but in my opinion this reflects the challenges in 1) having and combining good and comparable electrical and soil data, 2) petrophysical challenges even when more complex petrophysical models and quantities are considered (e.g., complex conductivity). Ultimately, I agree on the stated lack, but this seems to hinder the attempts toward national level generalizations.
L119 My perspective is that soil variability and hydrological spatiotemporal dynamics would impact site geophysical measurements, representativeness of the collected soil data, and water dynamics (preferential pathways, seasonality, etc.), this is not sufficiently expressed. Even with good data, site knowledge, and experience, it is hard to predict what would be measured, it is often already difficult to explain what was measured (i.e., taking advantage of the measured electrical conductivity or permittivity). In practical terms, ERT, FDEM, GPR and TDR, would give different SM estimates, potentially all correct and yet different, and very hard to extrapolate in space and time.
L136 A part from the geophysical work, are there previous works on the soil hydrology part?
L210 “should be interpreted cautiously” I agree, but it is not clear to whom this is direct. The scientific community is aware of this, the public or the end-user of the database would likely not read this or would not know the implications. A user-oriented uncertainty would be more helpful. That said, my concerns would then shift to the uncertainty associated with the abstraction of a relationship that was meant for other contexts and to be used after calibration.
Eq. 4, I think az is the Winsauer fitting-tortuosity parameter, correct? While this has been widely used for fitting, its use here in predicting soil resistivity may be avoided. For example, this would add a third free-fitting parameter (a, m, n), in addition to porosity and saturation which are here taken from the soil data. The centrality of these aspects to the geophysical part would require a more detailed introduction (see for example Glover, 2016 and 2017 on the use of the Winsauer a parameter).
L208 the authors are likely aware that n and m (as for az above) do not depend only on the texture, but on all soil properties, which is why assigning a value without calibration does not make much sense in my opinion. I may have missed how these were estimated in this study. The retention information was not used to improve the prediction of geophysical quantities, right? For example, to constrain and improve the choice of these Archie parameters. I think the authors focused on the dependency of SM and HG quantities on the retention data, which is interesting (although I did not find significant novelty as the may driver was WHC, see comments below).
Section 3.4 I think the procedure is correct but I don’t think it adds much. Why would the propagation be relevant? One would, in the end, estimate SM from rhoa and permittivity, no? The vertical profile and their physical-methodological implications would be relevant if methodological aspects were included (e.g., FDEM sensitivity to the defined soil layers, GPR penetration - reflections, ERT resolution), but I would consider these outside the scope of this work. Similarly, inferring hydrological dynamics from these physical profiles would also be counter intuitive, as I would directly consider the SM and several other hydrological variables available.
Section 3.5 I have encountered similar interpolation when working with VLF EM wave propagation, but there the physics (i.e., length and depth of the EM waves) support such an ambitious task. I am not sure how much this can translate to this study, where the controlling variables and parameters vary at much smaller spatial scale, yet controlling the hydrological and geophysical quantities.
L349-358 “... indicates that the derived dielectric proxy primarily reflects retained water in the monitored profile. The inverse dielectric–velocity relationship was clearly expressed in the v profile (Fig. 3C). ...” Yes, I think this is intrinsically correct and expected from the models. I think this could be removed to better focus more on the final use of the database and its contextualization.
L359 “Electrical properties showed a stronger vertical contrast than SM and εᵣ.” “This reciprocal σ–ρa behavior indicates stronger conductive continuity in the middle profile, most likely associated with higher water retention and improved pore-water connectivity.” As for the previous paragraph, this part could also be shortened.
L366 I don’t see how the attenuation would contribute to the SM-oriented study. See also general comment on section 3.4, above.
Section 4.2 As R1 also highlighted, the soil hydrology and climatic data are very interesting and actionable information. The results and interpretation are in line with well established soil water dynamics. I think the network of sensors is very nice, particularly its good spatial-regional coverage (shown if Fig. 1). For example, I would be interested in the retention curves themselves and associated hydraulic conductivities and van Genuchten parameters.
Section 4.3 L 439 “These results demonstrate that increasing water-holding capacity systematically reduces v, confirming that retained water enhances dielectric polarization and slows wave propagation.” Although rarely framed in these terms, this is well established, both retention properties and geophysical methods strongly depend on the clay content, and at a lower degree on soil organic matter. Correct?
In light of my previous comments, I think that the underlying soil and hydrology datasets are very extensive and valuable, but I could not see the value in this geophysics-oriented manuscript.
Sincerely,
Luca Peruzzo
Citation: https://doi.org/10.5194/essd-2026-563-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 389 | 117 | 38 | 544 | 41 | 40 |
- HTML: 389
- PDF: 117
- XML: 38
- Total: 544
- BibTeX: 41
- EndNote: 40
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The manuscript is based on a valuable national-scale dataset combining multi-year, depth-resolved soil-moisture observations from 117 monitoring stations with information on soil texture, porosity, hydraulic-retention properties and related environmental variables across six depths. In my view, this dataset represents the strongest contribution of the study. I therefore see clear potential for publication of the dataset. The underlying CSV data files are accessible, appear complete and are generally well structured and understandable. In this respect, I consider the dataset itself potentially suitable for publication as a data contribution. However for a data paper, I would expect a much stronger focus on data provenance, variable definitions, quality-control procedures, uncertainty and the distinction between measured, processed and model-derived quantities. The CSV files themselves are generally understandable and usable, but formal metadata are missing and the provided Python scripts are only sparsely documented. This makes independent reuse and reproduction of the processing workflow more difficult than necessary.
A substantial part of the article focusses on interpreting model-derived hydrogeophysical proxies, integrated indices, spatial patterns and process-related conclusions. At this point I have two major concerns regarding the interpretation of the derived geophysical proxies and the proposed 30–45 cm buffering zone that require restructuring the manuscript and reanalysis of the underlying dataset.
Geophysical proxies
From my point of view an independent validation of the derived geophysical proxies would be essential if the authors want to use them as a central part of the manuscript. Dielectric permittivity, EM-wave velocity, electrical conductivity, attenuation and apparent resistivity are not independently measured but are derived from soil moisture and other soil properties through assumed petrophysical relationships – the authors state that but forming a transferable national calibration layer out of this is not appropriate in my view. The present analysis is a model-based translation from observed soil conditions to expected geophysical responses, but it does not demonstrate that these responses correspond to actual GPR, ERT, EM, dielectric or laboratory measurements. This distinction is fundamental and should be reflected much more clearly in the manuscript. A related major concern is the lack of a systematic uncertainty assessment. The derived proxy values depend on model choice and parameterization as well as on assumptions regarding pore-water properties, porosity, saturation, soil texture and other variables. These uncertainties are likely to vary considerably among sites and soil classes. Reporting single proxy values (e.g., with an accuracy of one decimal place for the resistivity values) without uncertainty or sensitivity ranges implies a degree of accuracy and transferability that is not supported by the underlying methodology. A detailed sensitivity and uncertainty analysis is needed to quantify how parameter uncertainty and model assumptions propagate into the derived electromagnetic quantities.
I’m concerned about the use of an Archie-type relationship for clayey and fine-textured soils. In my view, a classical Archie formulation should not be applied as a quantitative proxy model to soils with clay content, because surface conductivity associated with clay minerals violates a central assumption of the model. Under such conditions, electrical conductivity is no longer primarily controlled by pore-water conductivity and saturation. Assigning texture-specific Archie parameters doesn’t resolve this fundamental limitation. For clayey soils, the authors should use a petrophysical formulation that explicitly accounts for surface conductivity and clay-related charge transport or refrain from deriving quantitative proxies for these soil classes. This is important because the Archie-derived quantities are subsequently used to compare soil-texture classes, calculate integrated indices, identify spatial patterns and support broader hydrogeophysical interpretations. Any systematic bias introduced by an inappropriate petrophysical model is therefore propagated into several downstream analyses.
Interpretation of the 30–45 cm “buffering zone”
The interpretation of the 30–45 cm interval as a nationally consistent hydrogeophysical “buffering zone” appears insufficiently supported or at least I don’t understand it out of the manuscript. The proposed zone is primarily based on a relatively small maximum in measured soil moisture at the discrete 30 and 45 cm measurement depths. The associated dielectric, velocity, conductivity, attenuation and resistivity patterns are derived from this same soil-moisture signal. Therefore these geophysical proxies don’t provide independent confirmation of the existence of a distinct process-based layer. I find it surprising that the physical origin of this potentially very interesting vertical pattern is not explored in greater detail. The monitoring network is described as covering different land-use settings, but land use isn’t systematically included in the analysis. In agricultural soils, the 30–45 cm interval could coincide with the lower boundary of the cultivated layer, a compacted plough pan or other management-related structural changes affecting porosity, hydraulic conductivity, rooting and water retention. Furthermore the manuscript does not investiges whether the observed moisture maximum corresponds to pedological horizons, changes in bulk density, root distribution, soil structure or hydraulic properties. These explanations should be investigated before interpreting the observed pattern as a generic buffering mechanism. I suggest, the authors should test whether the mid-profile maximum persists when the data are separated by land-use class, for example cropland, grassland and other land uses. Where such information is available, the pattern could also be related to soil horizons, bulk density, rooting characteristics and management or tillage depth. It could also be useful to assess whether the maximum occurs consistently at individual stations or whether the apparent 30–45 cm zone partly emerges from averaging a heterogeneous national dataset.
I had no closer look at the Machine-learning analysis as the above mentioned items required an enhanced revision from my point of view.
The figures are very difficult to read in their current form. Axis labels, tick labels, legends, annotations and in some cases panel labels are too small and should be revised.
In summary, I see a considerable value in the soil-moisture and soil-property dataset, but the hydrogeophysical interpretation requires substantial strengthening/revision. In my view, publication of the geophysical proxies requires independent validation against actual geophysical or laboratory measurements, together with or at least considering explicit propagation of model and parameter uncertainty. The use of Archie-type relationships for soils with clay content should be avoided. The 30–45 cm moisture maximum is an interesting finding, but its interpretation as a generic hydrogeophysical buffering zone should be supported by a more detailed analysis of land use, soil-profile architecture and site-specific vertical patterns.