the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
The QuantiSlakeTest, dynamic weighting of soil under water to measure soil structural stability
Frédéric M. Vanwindekens
Brieuc F. Hardy
Abstract. We evaluated the performance of a new, simple test to evaluate soil structural stability. The QuantiSlakeTest (QST) consists in a quantitative approach of the slake test, a dynamic weighting of a dried structured soil sample once immersed in water. The objective of this work was threefold: we aimed to (i) derive indicators from QST curves to evaluate soil structural stability regarding the underlying mechanisms of soil disaggregation; (ii) establish the relationship between soil properties and QST indicators; and (iii) assess how QST indicators respond to contrasting soil management practices. To meet these goals, we sampled the soil of 35 plots from three long-term field trials in the silt loam region of Belgium dealing with contrasting organic matter inputs, tillage treatments and P-K fertilisation, respectively. For each plot, QST curves were compared to the three tests of Le Bissonnais, targeting specific mechanisms of soil disaggregation.
Shortly after immersion in water, soil mass increases due to the rapid replacement of air by water in soil porosity. Then soil mass reaches a maximum before decreasing, once mass loss by disaggregation exceeds mass gain by air loss. Our results confirmed that the early mass loss under water is mainly related to slaking, whereas after a longer time period, clay dispersion becomes the dominant process of soil disaggregation. The overall soil structural stability was positively correlated to the soil organic carbon (SOC) content and negatively correlated to the clay content of soil. Accordingly, the SOC:clay ratio was closely related to QST indicators. Nevertheless, for a similar carbon (C) input, green manure and crop residues were more efficient in decreasing clay dispersivity whereas farmyard manure promoted SOC storage and was more efficient against slaking. QST curves had a strong discriminating power between reduced tillage and ploughing regardless of the indicator, as reduced tillage increases both total SOC content and root biomass in the topsoil.
The QST has several advantages. It is rapid to run, doesn't require expensive equipment or consumables and provides a high density of information on both specific mechanisms of soil disaggregation and the overall soil structural stability. As an open access program for QST data management is currently under development, the test has a strong potential for adoption by a widespread community of end-users.
- Preprint
(2541 KB) - Metadata XML
- BibTeX
- EndNote
Frédéric M. Vanwindekens and Brieuc F. Hardy
Status: closed
-
RC1: 'Comment on egusphere-2022-1092', Nicolas P.A. Saby, 24 Nov 2022
The manuscript presents a very interesting study on a new laboratory method to measure soil structural stability. The subject is well introduced, the statistical analysis is thoroughly executed using a nice and robust dataset.
The manuscript is very well written, well organised, interesting and very clear. There is also a very interesting discussion on the mechanisms controlling soil disaggregation.
the paper addresses a relevant scientific question within the scope of SOIL by providing new soil health indicators determined by a new and simple method. The results are compared with results obtained using a reference method (MWD).
Moreover, the authors paid particular attention to the publication of FAIR data and codes.
I have very little comments
In the summary, l8, it is not clear that you use some indicators based on the curves to the derive information on soil structural stability. Maybe you could add one sentence to specify this point.
L89, ‘in field conditions’, it is not clear to me. SLAKES mobile application should be use in a lab or a room.
L100 there is a repetition of the definition of the QST acronym?
Is Table 1 published on a data repository? If yes, add the DOI in the caption and the text.
L230-240, the choice of the different indicators could appear arbitrary and in particular for the time (ii) or the threshold value (iii). I don’t think it is discussed. More generally, the use of the QST curves could be discussed in terms of statistics. You may use functional data analysis or curve modelling instead. Fajardo et al. use a Gompertz model and interpret the parameters of the curves. Maybe this point could be mentioned for future research. I think there is only one shourt sentence in the conclusion (l495).
Figure 4 and 5 I think we are missing some results on the MWD here to be able to compare.
L286 and Fig 3, yes there is a clear relationship but also there is a residual variability. I think this should be acknowledged in the text.
L344 some QST indicators. It is possible to be more specific?
L386. Add reference to figure 3 for clarity?
L410: something wrong with the reference.
L450 the use of a large soil volume facilitates the soil sample preparation compare to MWD.
L500 this sentence is right only for soils of central Belgium for the moment.
Citation: https://doi.org/10.5194/egusphere-2022-1092-RC1 -
AC1: 'Reply on RC1', Frédéric Vanwindekens, 19 Dec 2022
[N.B. We copied/pasted the comments and questions and added our answers after each points - main questions of referee are in bold in the text below]
The manuscript presents a very interesting study on a new laboratory method to measure soil structural stability. The subject is well introduced, the statistical analysis is thoroughly executed using a nice and robust dataset.
The manuscript is very well written, well organised, interesting and very clear. There is also a very interesting discussion on the mechanisms controlling soil disaggregation.
the paper addresses a relevant scientific question within the scope of SOIL by providing new soil health indicators determined by a new and simple method. The results are compared with results obtained using a reference method (MWD).
Moreover, the authors paid particular attention to the publication of FAIR data and codes.
• We appreciate the kind feedback of Mr Saby and we are very grateful for his time and helpful comments.
I have very little comments
• In the summary, l8, it is not clear that you use some indicators based on the curves to the derive information on soil structural stability. Maybe you could add one sentence to specify this point.
◦ The term "indicators" was mentioned earlier in the paragraph, in the objectives of the paper, but we added details also in l8, for more clearly point that these are calculated from the curve at some typical dynamic step:
◦ "For each plot, QST indicators derived from curves (e.g. relative increase or decrease in soil mass, disaggregation speed, time to reach a threshold value of mass loss …) were compared to the results of the three tests of Le Bissonnais, targeting specific mechanisms of soil disaggregation."
• [X] L89, ‘in field conditions’, it is not clear to me. SLAKES mobile application should be use in a lab or a room.
◦ The strength of SLAKES application relies on the rapid acquisition of data on soil structure with little equipment needed (the measurement can be performed inside but also outside in absence of rain or wind) à we propose to remove “in field conditions”:
◦ “Recently, the SLAKES mobile application provided encouraging results as a tool for rapid data acquisition on soil structure. The test relies on image recognition to measure…”
• [X] L100 there is a repetition of the definition of the QST acronym?
◦ Yes, a mistake, we adapted:
◦ "In this work, we evaluated the performance of a new, simple test to evaluate soil structural stability, named QuantiSlakeTest (QST), a quantitative approach of the slake test."
• [X] Is Table 1 published on a data repository? If yes, add the DOI in the caption and the text.
◦ The data were available via the "gitlab" repository (https://frdvnw.gitlab.io/qst-openscience/). But not in a real "data repository". We added this table in the data repository linked to the "SlakingLab - users, data and codes for running QuantiSlakeTests" community on Zenodo. Reference and doi : Vanwindekens, Frédéric M., & Hardy, Brieuc F. (2022). QST - open data of the article Vanwindekens & Hardy (2022) - table 1 soil properties (1.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.7405113
◦ As suggested, we will add the doi link in the text, just after the ref to the table and in the caption of the table 1.
• [X] L230-240, the choice of the different indicators could appear arbitrary and in particular for the time (ii) or the threshold value (iii). I don’t think it is discussed. More generally, the use of the QST curves could be discussed in terms of statistics. You may use functional data analysis or curve modelling instead. Fajardo et al. use a Gompertz model and interpret the parameters of the curves. Maybe this point could be mentioned for future research. I think there is only one short sentence in the conclusion (l495).
◦ This is a very good point, Thank you for this comment. Actually, curve modelling has been tested (we used double exponential models) but the increase in soil mass occurring at the start of the test makes modelling complicated and model parameters correlated poorly with MWDs and soil properties. A perspective that we see is to measure the mass of soil leaving the basket additionally to the mass of soil in the basket. This second measurement would have the advantage to get rid of the interference of air bubbles leaving the basket and therefore limiting the number of factors controlling soil mass evolution under water. At the moment our device doesn’t allow such a measurement. We believe that such a development has the potential to improve greatly curve interpretation regarding the underlying mechanisms of soil disaggregation
• [X] Figure 4 and 5 I think we are missing some results on the MWD here to be able to compare.
◦ We think reviewer point figures: 4 and 6 (same kind of figures, with boxplot). We structured the results of our paper in three main parts: i) comparison QST <-> Le Bissonnais ii) indicators against soil properties (using QST & LeBissonnais) iii) then comparing the treatments of the LTE against soil structural stability, but with a focus on the QST indicators. The two first parts were there to evaluate the QST curves in comparison with recognized ones (Le Bissonnais), while for the comparison of contrasting farming practices, we used only the QST as the central approach of the paper. That's why it did not come into our mind to add boxplot with MWD indicators.
◦ We think also that the 2 graphs are already complete. We already made a "sub-selection" of most relevant QST indicators.
◦ Nevertheless, we found the proposition of the referee and we propose to add 3 graphs in supplementary materials. We will add a reference to these graphs in the text of this part.
• [X] L286 and Fig 3, yes there is a clear relationship but also there is a residual variability. I think this should be acknowledged in the text.
◦ In this part of the results, we focus the presentation of our data on correlation coefficients between variables from soil properties and from indicators. Our aim is not to produce a predictive approach with a model. We observe a standard deviation in the repeated test of a same plot, that could be explained by local soil conditions of the sampling sites (eg. slopes, presence of roots, or earthworms' galleries). We will acknowledge that in the new version of the text.
• [X] L344 some QST indicators. It is possible to be more specific?
◦ We will specify in the text:
◦ "Therefore, we think that that QST indicators Slope~max-60~ might be more specific to slaking than the fast-wetting test of Le Bissonnais."
• [X] L386 Add reference to figure 3 for clarity?
◦ Added:
◦ "It is important to underline that the close linear relationship found between W~max~-W~t0~ and the SOC:clay ratio (see Fig. 3)) has probably no general character and was obtained here because cropland soils of the current study were sampled under standardised conditions of seeding and cover (winter wheat)."
• [X] L410: something wrong with the reference.
◦ Yes, a mistake in the manuscript (omission of a LaTeX command). We adapted:
◦ "… allowed to maintain SOC content to the initial level (about 10 g kg−1; Buysse et al., 2013b). A smaller SOC storage for a similar C input means a higher rate of mineralisation. The formation of water-stable aggregates under the effect of microbial decomposition of root biomass is a known process (Dufey et al., 1986). In the present study, the more microbially active, labile"
• [X] L450 the use of a large soil volume facilitates the soil sample preparation compare to MWD.
◦ We agree that the soil sample preparation of the QST is easier compare to MWD, but we don't think that the sample size is the cause. We will add in the paper, in the beginning of this paragraph:
◦ "The soil sample preparation is far easier than approaches based on MWD as, after removed from the steel cylinder, the sample is dried “as is” and used “as is” in the QST."
◦ If soil sample preparation is shorter, we don’t see it as critical because clods crumbling by hand and soil sieving to 3-5 mm is neither complicated nor time consuming. To our point, the main advantage is the absence of bias compared to sieving that selects aggregates of a certain size, neglecting particles < 3mm that can represent a large fraction of a soil that is frequently tilled
• [X] L500 this sentence is right only for soils of central Belgium for the moment.
◦ We agree (and we are very interested to test the approach in other pedological context). We will adapt this paragraph of the conclusion as following:
◦ "For cropland soils of the silt loam region of central Belgium, we show that the early mass loss under water is mainly related to slaking, whereas after soil saturation with water, clay dispersion becomes the dominant process of soil disaggregation. "Citation: https://doi.org/10.5194/egusphere-2022-1092-AC1
-
AC1: 'Reply on RC1', Frédéric Vanwindekens, 19 Dec 2022
-
RC2: 'Comment on egusphere-2022-1092', Anonymous Referee #2, 18 Feb 2023
First of all, apologies for the delayed response.
This article presents a relevant, novel and worthwhile adaptation of an existing soil structural stability test, and applies it to different case studies, demonstrating its potential for routine evaluations within the context of soil (structural) quality assessments. It certainly deserves to be published. Nevertheless, in my opinion, the introduction needs to be revised, a number of issues have to be clarified (see details below) and additional indicators derived from the QST curves could be tested (without requiring additional measurements).
Introduction
The introduction is quite lengthy. There is a lot of interesting information, but it does not fully serve the purpose of the paper. Part of the difficulty stems from the fact that the authors were not able to decide whether the main focus of the paper is about a new tool or about understanding the effects of various practices on their novel indicators. To me, the paper is first and foremost about a new technique to quantify soil disintegration during immersion, which is then applied to a variety of soil samples from different cropping systems from loess soils in central Belgium. The focus of the introduction should therefore be on existing techniques and why another method like the QST is relevant (L78 and subsequent). L23-49 do not really seem relevant for introducing the QST.
It is also not obvious to me how relevant it is to discuss the hierarchical levels of soil structure, except that the samples should be sufficiently large to encompass all levels so as to be representative of the soil. The fact that the authors do not revert back to these concepts in their discussion clearly indicates that this information is of limited relevance for this paper.
In English, the usual term is ‘aggregate stability’, which in the case of Le Bissonnais is indeed what is being measured (3-5 mm aggregates). It may be worthwhile discussing the distinction the authors make between aggregate stability and structural stability (if any). I would argue that QuantiSlake = structural stability, whereas Le Bissonnais = aggregate stability …
Specific comments
L9 : this is not fully correct. see comments further down on the Le Bissonnais test.
L32 I suppose liming is also an important practice in these systems, that has contributed to changes in pH and base saturation. Liming does not fall under 'organic and mineral fertilizer'. It is a soil amendment.
L35-37 please add reference
L42 add reference for the ‘critical value of SOC content’
L47 ‘…. According to farmers ….’ and scientists !
Starting L46 (till L65), the text becomes somewhat confusion. The paragraph starts about conservation tillage, then switches to erosion, then switches to aggregate stability and the hierarchical structure of aggregates : these are a lot of different ideas for a single paragraph. In general, 1 paragraph = 1 idea.
L57 Ca++ … and Mg++ (as a rule, divalent cations are much more effective than monovalent cations)
L68 The mechanisms listed here are relevant for soil sealing / soil erosion studies. I would argue that in a broader context, mechanical breakdown also occurs as a result of mechanical stresses exerted during tillage operations or traffic.
L86-87 : the method of Koestel is not really about measuring aggregate stability; Likewise, the use of VIS-NIR does not allow to measure stability : it is based on correlations, and therefore will always require reference methods.
L90-93 : there is some confusion here in the text between ‘aggregate size distribution’ (or level of aggregation) and ‘aggregate stability’. Furthermore, the “profil cultural” (and other methods listed here) do not seek to measure aggregate stability.
L95 : not sure I understand why topography and climate would influence the choice of method; please clarify
L108-116 : can be deleted, as it will be repeated in materials and methods / discussion sections
L121 explain abbreviation : CRA-W
L123 : round off to the nearest mm; the decimal really doesn't provide useful information
L148 latin names are provided here but were not given in the previous section when the crops were first named; please correct
L153 “Complete random block with split plot design” : doesn’t sound quite right. ‘split-plot’ is not compatible with ‘completely randomized’ because by definition, at least one treatment (the ‘split’) is not fully randomized. Please clarify.
L165 sampling occurred in April 2019 but no fertilizer was added since 2016 ? Please check
L171 I’m not aware that chlorides can affect soil structure in soils dominated by permanent charge, but potassium (monovalent cation) definitely does!
L198 : did you replicate measurements for each plot ? I believe Le Bissonnais recommends 3-5 replicates per plot (and per test).
L200 : the mechanisms involved in the test are not presented correctly; see comment L211 (below)
L203 : the whole point of rewetting the soil using alcohol prior to shaking in water is also to minimize slaking (and swelling and dispersion)
L203 : No sieving at 50 µm during the immersion phase in ethanol ?
L205 : Start a new paragraph at ‘Two main indicators …’
L209 : how strongly correlated are the MWD and MA indicators ? Is it worth considering both ?
L211-212 : this way of presenting the tests is not strictly correct. rapid wetting involves slaking, dispersion, but also differential swelling! Slow wetting involves differential swelling and dispersion. The mechanical breakdown test seeks to minimize slaking, swelling and dispersion, but it is not obvious that this really mimics drop impact. From the discussion, it appears that the authors are aware of all this, so why present the three test in such a caricatural way ?
Table 1 : please specify the upper and lower limits for the different particle size fractions, as this can be different from one country to another
L225 not very clear. According to L220, the first measurement should be within less than 1 second after plunging the sample into the water. What does ‘right after buoyancy’ refer to ? How do you determine the time to buoyancy? Maybe this could be illustrated graphically.
L229: start new paragraph at ‘Several ….’
L229 If I understand well, this maximum mass is going to depend on the relative rate of disintegration vs. relative rate of wetting, as well as mass of the sample (corrected for buoyancy based on the volume of the solid phase). Doesn’t this also contain information?
L234-236 this has to be better explained. The sentence seems to indicate that it is the slope of the curve at t= 30 sec (with t0 taken as tmax), but this is not what Fig 2 shows. What Fig 2 shows is not the slope of the curve, but the mass loss over a certain time interval (so the slope between t0 and t30, and not the slope at t30).
L240 what about testing (Wmax-Wend) ?
L240 indicators are presented as ‘mass at the end’ and ‘area under the curve’. So units should be ‘g’ and ‘g.sec’. But in later graphs, Wend is presented as unitless and AUC is in 1/sec. So it seems that both indicators have been normalized. This must be explained more clearly.
L241 on line 219, it says the experiment is run for approximately 1000 sec. The AUC is evidently going the depend on the length of the experiment. So is a fixed duration used for all samples ?
The curves are normalized by Wmax, but not by Wend (use Wmax as upper limit and Wenfd as lower limit for normalization): isn’t that introducing some sort of bias ? Also, in the AUC, a large part of the value may come from Wend * duration (the bottom, rectangular area in Fig 2). So a large chunk of the AUC contains the same info as Wend. Normalizing the curves by Wmax and Wend would allow to have an AUC that is independent of Wend.
L243 check this sentence
Caption of Figure 1 : What does 'managing QST laboratory' mean ? Or do you mean 'for managing and graphically displaying QST laboratory data '?
L245 why only the roots remaining in the cage ? Isn’t this a biased estimate, as smaller root fragments may have fallen through the mess during sample breakdown ? Wouldn’t it be more appropriate to pass the entire soil again through a 2-mm mesh sieve to recover all roots?
L256 Although the choice of threshold is to some extent arbitrary, using a 10% probability threshold is somewhat unusual in statistics. Using a higher threshold can sometimes be justified based on the dataset characteristics, but what is the reason for this choice here ? Please justify.
L268 there is no such Fig. 2. Fig. 2 corresponds to the QST and Fig. 3 to another correlation. Perhaps you mean Table 2?
- 3.1 it is also worthwhile noting that correlations are usually stronger with MWD1 than MWD2 (which seems to make sense)
Table 2 : It seems that slope values are expressed in negative values, which was not immediately obvious to me after reading the materials and method section. (I was expecting a higher slope to be negatively correlated with SOC content; the positive correlation stems from the fact that negative values are used). This should be clarified.
L287 : the study of Johannes et al., 2017, did not investigate soil structural stability but soil structural quality (which are two different concepts)! Visual assessment methods don’t allow to assess stability.
In section 3.2, I’m missing information regarding the correlation among indicators from the same test. How are slopes, txx, AUC, … correlated in the QST test ? Is it worth considering all these indicators? Same for MWDs and MAs in Le Bissonnais tests : are they correlated ? This comment also relates to L301-304 : what was the basis for this selection of criteria? Correlation analysis would help justify the selection.
Fig. 4 Wmax – Wt0 are given as unitless, even though in materials and methods the authors talk about mass loss. This is confusing. The same problem arise for slopes, Wend, AUC. The confusion seems to stem from the fact that the indicators are not well explained in the M&M section (one speaks of ‘mass loss’ whereas it should be ‘relative mass loss’).
L336-337 : as mentioned above, it is not correct to attribute single mechanisms to each of the three tests of Le Bissonnais. In the FW-fast wetting test (MWD1), all mechanisms are involved (slaking, dispersion, diff. swelling) except mechanical breakdown. In the SW-slow wetting test (MWD2), dispersion and diff. swelling can be present. So the fact that the QST indicators of L335 are better correlated with FW than SW does not necessarily indicate that slaking is the main mechanisms, but merely that slaking significantly contributes to breakdown in the QST, making the correlation stronger with MWD1. This is a subtle but important difference.
L337-340 : the ‘problem’ with t50, t75, etc… is that these indicators inherently also include what happens during the early times. Wouldn’t it be much more relevant to consider the time it takes for the sample to loose 25% of its mass, then the time from 25 to 50%, then the time from 50 to 75, etc … ? In this way, the various indicators can be expected to be much more independent, and interpretation of processes would also be facilitated. If indeed dispersion and diff. swelling are the main mechanisms at later stages, then one would expect even stronger correlations between MWD2 and, say, the time needed to lose 75-90% of the mass than between MWD2 and t90.
L343 : indeed ! … and from differential swelling. If the authors are aware of this, then why present the Le Bissonnais tests in the wrong way in materials and methods ?
L345 I think the authors misunderstand the FW test of Le Bissonnais. Slaking plays an important role in this test, but it is not the only mechanism at play. It is the difference between the SW and FW tests that predominantly reflects the effect of slaking (though slaking may actually also facilitate dispersion, so the effects are not simply additive). One cannot say that the FW test alone reflects the sole effect of slaking.
L348 indeed, the continuous measurements of QST would allow to better discriminate between mechanism on a single sample, something that cannot be done with the le Bissonnais test. Nevertheless, as suggested above, using time intervals (t0-25, t25-50,t50-75, …) should allow to better discriminate between processes than t0-25, t0-50, t0-75, etc. Furthermore, the difference between MWD1 and MWD2 of Le Bissonnais should better reflect the sole effect of slaking than MWD1 alone, so testing correlations between QST results (early times) and (MWD2-MWD1) might be of interest.
The same reasoning would be applied to the slopes : slope max-60, 60-300, 300-600.
L385 : again, it is not correct to state that VESS assesses structural stability (it is a measures of quality)
L421 : this is rather speculative. It is not certain at all that this would be the case. FYM is very different from green manures or crop residues (much lower C/N ratio right from the start) and therefore will not interact with microbial life in the same way.
L424-425 : I think I sort of understand what the authors mean, but the sentence is awkward and must be clarified
L438-440: Again, I’m not aware that chloride can have such effects on soils with permanent charge. There is no mention of the effect of chloride on stability in the paper of Paradelo. It is related to the monovalent cation (K). This discussion must be revised.
L450 aggregate stability tests such as Le Bissonnais or drop impact tests (Imeson and Vis) where developed in view of relating the results to soil erodibility or even as a way of parameterizing soil detachment during rainfall in erosion models. In that respect, working with aggregates seems fairly relevant (particularly in case of cultivated tilled soils). As a means to assess soil quality, the larger sample size used in the QST seems relevant. Whether QST results could be used to assess soil erodibility remains to be proven, even though the good correlations with Le Bissonnais results are encouraging.
L460 : differential swelling ?
L461 see suggestions above on how to possibly improve the indicators
L463-465 : again, there is a bit of a mixup here between ‘stability’ and ‘quality’
L468 : indeed, some overlap is expected with the current indicators, but this could possibly be improved (see suggestions above)
L484 I may be wrong (those documents are not easily accessible), but I believe that the equation developed in Laon does not relate to detachment by drop impact per se (but rather to the aggregate stability tests of Henin and Monnier as a measure of the soil sensitivity to crusting). To be checked.
L494-497 : future developments are not to be mentioned in the conclusion
L503 : I do not believe this was really demonstrated in this paper (as a matter of fact, linear correlations – as demonstrated in this paper – are not compatible with thresholds) , and it should therefore not appear here.
See additional annotations in the attached pdf file
-
AC2: 'Reply on RC2', Frédéric Vanwindekens, 19 Apr 2023
First of all, apologies for the delayed response.
This article presents a relevant, novel and worthwhile adaptation of an existing soil structural stability test, and applies it to different case studies, demonstrating its potential for routine evaluations within the context of soil (structural) quality assessments. It certainly deserves to be published. Nevertheless, in my opinion, the introduction needs to be revised, a number of issues have to be clarified (see details below) and additional indicators derived from the QST curves could be tested (without requiring additional measurements).
Authors : Dear Reviewer,
First we would like to thank you about the time spent for a serious, in depth revision of our work. Most comments appear to be relevant and will help to strengthen the manuscript. [SEE the pdf and work files attached for a better formatting)
Introduction
- The introduction is quite lengthy. There is a lot of interesting information, but it does not fully serve the purpose of the paper. Part of the difficulty stems from the fact that the authors were not able to decide whether the main focus of the paper is about a new tool or about understanding the effects of various practices on their novel indicators. To me, the paper is first and foremost about a new technique to quantify soil disintegration during immersion, which is then applied to a variety of soil samples from different cropping systems from loess soils in central Belgium. The focus of the introduction should therefore be on existing techniques and why another method like the QST is relevant (L78 and subsequent). L23-49 do not really seem relevant for introducing the QST.
Authors : Thank you for this comment. We agree that the main originality of the work stands in the new method of aggregate/structural stability measurement by soil immersion under water. Nevetherless, it is also clear that a new method can hardly be evaluated properly without an appropriate case study that offers a range of conditions and outputs for the test. The whole manuscript, from introduction to discussion, is written accordingly, with a balance between 1) QST method description and evaluation and 2) application of the QST to silt loam Luvisols (having a fragile structure) subject to contrasting soil management practices. The introduction respects this duality, with one part presenting the soil and agricultural context of the loess belt of central Belgium and the other one focussing on soil aggregation and aggregate stability measurement.
We don’t want to lose this duality because it represents the guideline and the genesis of our work, and that limiting the text to the QST method only would be a substantial loss of information, important to evaluate the method itself. We believe that the agronomic outputs of the manuscript can be of interest for many readers of SOIL and will not prevent other readers to put the focus on the methodology.
- It is also not obvious to me how relevant it is to discuss the hierarchical levels of soil structure, except that the samples should be sufficiently large to encompass all levels so as to be representative of the soil. The fact that the authors do not revert back to these concepts in their discussion clearly indicates that this information is of limited relevance for this paper.
Authors : We understand the point raised by reviewer 2 but we believe that reminding some key elements of soil aggregation is not useless at all for such a paper. First, some critical advances of research on soil aggregation date from the eighties and even earlier. Therefore, several milestone papers are not easily accessible online. Second, the ultimate goal of the test is to provide a practical, cheap and effective solution for soil stability measurement for a wide community of end-users, including those with little expertise in soil analysis. We are convinced that many of the readers of this text and final users of the test will greatly benefit from such information, in order to provide them a clearer idea of what stands behing the concept of soil aggregation (actually the feedbacks that we have received on the paper by now support this point).
- In English, the usual term is ‘aggregate stability’, which in the case of Le Bissonnais is indeed what is being measured (3-5 mm aggregates). It may be worthwhile discussing the distinction the authors make between aggregate stability and structural stability (if any). I would argue that QuantiSlake = structural stability, whereas Le Bissonnais = aggregate stability …
Authors : We thank you for this important remark. Indeed we think that it is useful to stick to “soil structural stability” rather than to “aggregate stability” because we work on soil volumes exceeding the size of soil aggregates. This information will be added in the M & M in the revised version of the text. We propose the following complement of information l. 243 :
“ Whereas Le Bissonnais (1996) and many reference methods measure the stability of soil aggregates, the QST rather measures soil structural stability, as it works on 100 cm³ soil volumes rather than on soil aggregates. Therefore we will stick to “soil structural stability” when referring to QST measurements.”
We will also check carefully to avoid the use of “soil structural stability” for other reference methods measuring “soil aggregate stability”
Specific comments
- L9 : this is not fully correct. see comments further down on the Le Bissonnais test.
Authors : Agreed, thank you for this remark. Will be revised all over the text cf. Comment l.211-212.
- L32 I suppose liming is also an important practice in these systems, that has contributed to changes in pH and base saturation. Liming does not fall under 'organic and mineral fertilizer'. It is a soil amendment.
Authors : will be replaced by “organic and mineral fertilizers and amendments”
- L35-37 please add reference
Authors : reference : Goidts & Van Wesemael (2007)
- L42 add reference for the ‘critical value of SOC content’
Authors : reference : Meersmans et al (2011)
- L47 ‘…. According to farmers ….’ and scientists !
Authors : we propose to remove "According to local farmers” in the revised version so the statement is more general
- Starting L46 (till L65), the text becomes somewhat confusion. The paragraph starts about conservation tillage, then switches to erosion, then switches to aggregate stability and the hierarchical structure of aggregates : these are a lot of different ideas for a single paragraph. In general, 1 paragraph = 1 idea.
Authors : This is where the transition between 1) the soil and agricultural context and 2) soil aggregation and soil aggregate stability measurement occurs. The order of information “soil erosion” > “soil aggregate stability” > “soil aggregation theory” doesn’t sound inappropriate to us, as i) the issue of soil erosion is the driver of our research; ii) Soil aggregate stability measurement is the most common proxy to assess soil erodibility and iii) the theory of soil aggregation is key to understand the properties of soil aggregates and mechanisms of soil aggregation and disaggregation.
In the revised version, we propose to make the transition smoother and to reorganize the paragraphs (l. 53):
“..., soil aggregate stability is often used as an indicator of soil erodibility (Barthès & Roose, 2002).
The process of soil aggregation is critical to understand the factors controlling soil aggregate stability. The theory of aggregate hierarchy of Hadas (1987)...”
- L57 Ca++ … and Mg++ (as a rule, divalent cations are much more effective than monovalent cations)
Authors : the sentence will be completed accordingly
- L68 The mechanisms listed here are relevant for soil sealing / soil erosion studies. I would argue that in a broader context, mechanical breakdown also occurs as a result of mechanical stresses exerted during tillage operations or traffic.
Authors : Thank you for this. “The resistance of soil to mechanical breakdown also improves resistance to soil compaction due to traffic on the field” will be added l. 71.
- L86-87 : the method of Koestel is not really about measuring aggregate stability; Likewise, the use of VIS-NIR does not allow to measure stability : it is based on correlations, and therefore will always require reference methods.
Authors : We agree with this remark, we propose to split the information, with spectroscopic techniques providing information on soil structure mentioned in a separated paragraph:
“Recently, the potential of some non-destructive methods on the evaluation of soil structure and aggregation has been revealed, such as...”
Accordingly, the information about the SLAKES application (l. 88-90) will be moved up and attached to the previous paragraph
- L90-93 : there is some confusion here in the text between ‘aggregate size distribution’ (or level of aggregation) and ‘aggregate stability’. Furthermore, the “profil cultural” (and other methods listed here) do not seek to measure aggregate stability.
Authors : We propose to remove from l.91 to 93 about the field methods for the evaluation of soil structure, for the sake of clarity.
- L95 : not sure I understand why topography and climate would influence the choice of method; please clarify
Authors : The idea is that topography and climate are key controls of erosion risk. The amount, frequency and intensity of precipitations and length and steepness of the slope will control erosion risks as much as soil intrinsic erodibility. Therefore, in a flat area in the temperate zone, soil aggregate stability in wet conditions may provide useful information on the risks of soil sealing and crusting or compaction whereas in other contexts the mechanical strength of soil, or a rainfall simulator approach may bring more relevant information.
- L108-116 : can be deleted, as it will be repeated in materials and methods / discussion sections
Authors : Actually we believe that l. 108-112 are absolutely necessary as they present the general approach of the work, in line with the objectives presented l. 104 – 108. We propose to remove the text from l. 113 to 116 (indeed somehow redundant with information within M&M)
- L121 explain abbreviation : CRA-W
Authors : Abbreviation is defined at first use (l. 109). For information, CRA-W refers to “Centre wallon de Recherches Agronomiques”, with letters in a confusing order due for historical reasons …
- L123 : round off to the nearest mm; the decimal really doesn't provide useful information
Authors : We fully agree, this will be done (3 significant digits)
- L148 latin names are provided here but were not given in the previous section when the crops were first named; please correct
Authors : we will check carefully that the Latin name is given for each plant species at first use in the revised version.
- L153 “Complete random block with split plot design” : doesn’t sound quite right. ‘split-plot’ is not compatible with ‘completely randomized’ because by definition, at least one treatment (the ‘split’) is not fully randomized. Please clarify.
Authors :
Some sources indicate that the split-splot design is compatible with randomized complete blocks : Dagnelie P. (2012), Principes d’experimentation. Planification des expériences et analyse de leurs résultats. See. Chap 7.
After verification, the design of the tillage trial is more correctly described as a Latin Square with one studied factor (the four tillage treatments) and one controlled factor (the 4 blocks) without repetition (sensu stricto). We therefore propose the following formulation:
“The trial includes four tillage treatments repeated four times, following a Latin square design with the blocks aligned in a row.”
- L165 sampling occurred in April 2019 but no fertilizer was added since 2016 ? Please check
Authors : P-K fertilisers are applied once per rotation, before the sugar beet, which is the starter of the three-year rotation. The last application before sampling occurred after the 2016 harvest of barley in summer 2016. The next application occurred after the 2019 harvest, so after sampling.
- L171 I’m not aware that chlorides can affect soil structure in soils dominated by permanent charge, but potassium (monovalent cation) definitely does!
Authors : we propose the following adaptation: “on the potential effect of contrasting levels of KCl application on soil structural stability (cf Paradelo et al 2016)”
- L198 : did you replicate measurements for each plot ? I believe Le Bissonnais recommends 3-5 replicates per plot (and per test).
Authors : Indeed 3 to 5 replicates are recommended by Le Bissonnais et al. (1996) but here we didn’t make any replication within one single plot because the soil sampling area was very limited (1 m²) and because we relied on field repetitions (true repetitions decrease the relevance of pseudo-replication). Moreover, since the Le Bissonnais method works on 5-10 g of aggregates 3-5 mm in diameter, the result is already an average value for many small aggregates.
- L200 : the mechanisms involved in the test are not presented correctly; see comment L211 (below)
Authors : We agree that across the document we sometimes make the following shortcut : fast-wetting = slaking. This statement must indeed be qualified and we will revisit the text accordingly. Nevertheless, in the methodology, the initial formulation seems correct (fast wetting of dry soil aggregates in water aims to test first and foremost their resistance to slaking, even if clay dispersion and differential swelling may also play a disaggregating role after 10 min under water).
We propose the following formulation: l. 200 “The first test consists in fast-wetting soil aggregates in water, exacerbating the effect of slaking”
About description of the third test, see next comment
- L203 : the whole point of rewetting the soil using alcohol prior to shaking in water is also to minimize slaking (and swelling and dispersion)
Authors : Indeed. We propose to complete the sentence l. 203 “...to test their mechanical strength while minimizing slaking, differential swelling and dispersion”
- L203 : No sieving at 50 µm during the immersion phase in ethanol ?
Authors : we followed rigorously the norm ISO FDIS 10930:2011. After each treatment, remaining aggregates were transferred to the 50 µm sieve. The < 50 µm fraction is not recuperated for any of the three test (it is calculated by difference between initial weight and fractions remaining on the sieves)
- L205 : Start a new paragraph at ‘Two main indicators …’
Authors : We agree and will adapt the manuscript in this way.
- L209 : how strongly correlated are the MWD and MA indicators ? Is it worth considering both ?
Authors : Both indicators are recommended by the norm. They are generally positively correlated for test 1 and 3 but not much for test 2. MA 3 is generally more (negatively) correlated to QST indicators than MWD3 (see Table 2). We propose to add the correlation matrix in supporting information. We propose to remove MAs from Table 2 since we make little use of it. Please fin the correlation matrix here below
- L211-212 : this way of presenting the tests is not strictly correct. rapid wetting involves slaking, dispersion, but also differential swelling! Slow wetting involves differential swelling and dispersion. The mechanical breakdown test seeks to minimize slaking, swelling and dispersion, but it is not obvious that this really mimics drop impact. From the discussion, it appears that the authors are aware of all this, so why present the three test in such a caricatural way ?
Authors : As stated earlier, we agree with this comment and we are aware of this. The text will be revised accordingly (see comment l. 200). The link between the three tests and the mechanisms of soil disaggregation will be qualified in the introduction and we will stick to the names of the three tests, “fast wetting”, “slow wetting” and “mechanical breakdown” in the other sections.
- Table 1 : please specify the upper and lower limits for the different particle size fractions, as this can be different from one country to another
Authors : We used the following limits based on NF ISO 11464 :
- Sand (coarse, > 200μm)
- Sand (fine, 50 μm - 200 μm)
- Silt (coarse, 20 μm - 50 μm)
- Silt (fine, 2 μm - 20 μm)
- Clay (< 2 μm)
- As fine/coarse sand and silt data are not used in the manuscript we will simplify the table by removing them. In the revised version of the manuscript will only stand total sand, silt and clay:
- Sand (50 – 2000 μm)
- Silt (2 μm - 50 μm)
- Clay (< 2 μm)
- L225 not very clear. According to L220, the first measurement should be within less than 1 second after plunging the sample into the water. What does ‘right after buoyancy’ refer to ? How do you determine the time to buoyancy? Maybe this could be illustrated graphically.
Authors : Indeed an illustration speaks from itself. Actually the graphs as shown on figure 2 and figure 5 only show soil mass under water after Archimedes upward buoyant force (once soil sample is completely immersed) --> if you keep soil mass data before and during immersion, you see the big drop from Archimedes buoyancy.
We will produce a didactic graph in the new proposal. Here a rapid “raw” graph. We use simple mathematical approaches based on derivatives (‘primary’, ‘secondary’) for finding the first minima. This point id considered as the “T0” and points before it are deleted (soil sample in the air, or partly in the air). The final graph that we will propose will complete the fig. 2 of the submitted version (graph with indicators).
- L229: start new paragraph at ‘Several ….’
Authors : We agree and will adapt the manuscript in this way.
- L229 If I understand well, this maximum mass is going to depend on the relative rate of disintegration vs. relative rate of wetting, as well as mass of the sample (corrected for buoyancy based on the volume of the solid phase). Doesn’t this also contain information?
Authors : As explained in our previous comment, we deleted information before and during immersion (the buoyancy). Soil mass is then normalized according to the maximum value reached (WMax=1) after the considered T0 (first minimum), so that soil mass is expressed as a relative mass.
If you now look at figure 2, you see how soil mass generally behave for the soils of this study: soil mass increases due to wetting (due to water filling in porosity) and reaches a maximum (WMax = 1) before decreasing (once mass loss due to disaggregation becomes dominant compared to mass gain by wetting).
Going back to Archimedes buoyancy: indeed, mass drop due to immersion certainly contains extra information that we didn’t exploit in the present manuscript.
To clarify, we propose the following reformulation :
l.225 - … : “soil mass drops due to Archimedes upward buoyant force (data not shown). The first value of soil mass under water is defined as the time 0 (t0) of the QST test (soil mass before and during immersion is removed from the graph and not shown here). The graph on Figure 2 illustrates how soil mass behaved under water for the soils of the present study. In the initial phase, soil mass generally increases due to water filling porosity. After a few seconds or minuts, the soil mass reaches a maximum (Wmax) before decreasing, once mass loss due to disaggregation becomes dominant compared to mass gain by wetting. Soil mass was normalised according to Wmax, so that mass value vary between 0 and 1. …"
l. 231 --> 241: “mass loss” will be changed to “relative mass loss”
Authors : We don’t fully understand the question. I will try to clarify. All our graphs and indicators are based on a relative weight. As we used the Wmax as reference (denominator), the relative weight at “Max” is 1 (see also answer for l240).
- L234-236 this has to be better explained. The sentence seems to indicate that it is the slope of the curve at t= 30 sec (with t0 taken as tmax), but this is not what Fig 2 shows. What Fig 2 shows is not the slope of the curve, but the mass loss over a certain time interval (so the slope between t0 and t30, and not the slope at t30).
Authors : We agree with the comment. But, we computed in the new manuscript new indicators based on the proposal done in comment linked to L337-340. We computed 2 new kinds of indicators: local slope (slope between max and max+30sec, slope between max+30 and max+60,..) and the “delta - t” (ex delta-t50-75, time between 50% loss and 75% weight loss). After analyses, we proposed to switch to the new version of “local slopes”, which are relevant and show interesting results. When presenting these new slopes, we will be as clear as possible to avoid misunderstanding.
- L240 what about testing (Wmax-Wend) ?
Authors : As we consider the relative soil mass (Wmax = 1), Wmax-Wend won’t provide more information than Wend itself.
- L240 indicators are presented as ‘mass at the end’ and ‘area under the curve’. So units should be ‘g’ and ‘g.sec’. But in later graphs, Wend is presented as unitless and AUC is in 1/sec. So it seems that both indicators have been normalized. This must be explained more clearly.
Authors : As explained above (see l. 228-229 of the initial version of the text), soil mass is a “relative soil mass”, without unit. The ‘y-axis’ has no unit (relative soil mass, normalized according to Wmax [g/g] = [-]) --> Wend and all other “weight” indicators (Wmax, Wt0,…) have no unit.
As a consequence the unit of slopes are only [1 / sec] = [sec-1] and AUC [1 x sec] = [sec].
It was expressed in L228-229, but we will present that more clearly (see earlier proposition of reformulation).
- L241 on line 219, it says the experiment is run for approximately 1000 sec. The AUC is evidently going the depend on the length of the experiment. So is a fixed duration used for all samples ?
Authors: Thank you for this comment. for the sake of comparability, we fixed the time to 900s for each sample, because a few of them didn't last until 1000 s. This 900 s timestep is our reference time for AUC calculation. For QST that had a duration time < 900s because of reaching a steady state before that, the curves were artificially extended to 900 for a comparable AUC between all experiences.
- The curves are normalized by Wmax, but not by Wend (use Wmax as upper limit and Wend as lower limit for normalization): isn’t that introducing some sort of bias ? Also, in the AUC, a large part of the value may come from Wend * duration (the bottom, rectangular area in Fig 2). So a large chunk of the AUC contains the same info as Wend. Normalizing the curves by Wmax and Wend would allow to have an AUC that is independent of Wend.
Authors:
As suggested, the redundancy analysis of QST indicators revealed a strong (r=0.98) positive correlation between AUC and Wend. We made a try to split the AUC into two (the rectangular area delimited by Wend and the area between the curve and Wend. Nevertheless, the curve-dependent fraction of the AUC correlates poorly to MWDs and soil properties. As the interpretation of this indicator is not clear to us, we propose not to use it.
We agree that the final time of measurement is very important for the calculation of the AUC because the longer the time considered, the more the rectangular area controls the AUC... At the moment we didn’t look to optimize the time considered for AUC calculation with any objective criterion (such as, e.g., sensitivity to soil management practices...).
Calculation of the AUC between Wmax and Wend would provide a completely different information from the current AUC. Current AUC decreases with total relative mass loss by disaggregation; the proposed AUC (calculated between Wmax and Wend) would generally increase with mass loss but would also depend on the kinetics of mass stabilization. It would also be somehow redundant with Wend.
The contribution of the rectangular area delimited by Wend is important for samples with a good structural stability (elevated Wend) but decreases with soil structural instability (low Wend).
- L243 check this sentence
Authors: Sorry for this mistake, sentence will be removed.
- Caption of Figure 1 : What does 'managing QST laboratory' mean ? Or do you mean 'for managing and graphically displaying QST laboratory data '?
Authors: The `slaker` application is used also during the QST, for
- adding sample identification and useful metadata
- Defining experience parameters (max time, …)
- starting the recording of data,
- Checking the good processing of the data collection
- Stopping the experience (computer / data part)
- We agree the sentence is not clear, we propose to adapt : “application for parametrizing and driving the experience”
- L245 why only the roots remaining in the cage ? Isn’t this a biased estimate, as smaller root fragments may have fallen through the mess during sample breakdown ? Wouldn’t it be more appropriate to pass the entire soil again through a 2-mm mesh sieve to recover all roots?
Authors: The idea raised during the experience, seeing roots in the basket. We think you are right and, in further work, we will adapt the protocol for a more rigorous quantification of root content. We certainly underevaluated the root biomass, in the 2 tested treatments of the Tillage LTE. We propose to move the root biomass figure to supporting information.
- L256 Although the choice of threshold is to some extent arbitrary, using a 10% probability threshold is somewhat unusual in statistics. Using a higher threshold can sometimes be justified based on the dataset characteristics, but what is the reason for this choice here ? Please justify.
Authors: We will adapt the tables and the text considering only the significatively differences between treatments at a 5% probability threshold, more usual in statistical analyses
- L268 there is no such Fig. 2. Fig. 2 corresponds to the QST and Fig. 3 to another correlation. Perhaps you mean Table 2?
Authors: Yes, we should have written Table 2. We will correct the wrong reference in the revised version, the mistake occurs 8 times in the text
- 3.1 it is also worthwhile noting that correlations are usually stronger with MWD1 than MWD2 (which seems to make sense)
Authors: This is generally true, except for t75 to t95 which also makes sense, as the end of the curve is not expected to relate to slaking but rather to differential swelling and physico-chemical dispersion.
- Table 2 : It seems that slope values are expressed in negative values, which was not immediately obvious to me after reading the materials and method section. (I was expecting a higher slope to be negatively correlated with SOC content; the positive correlation stems from the fact that negative values are used). This should be clarified.
Authors: Slopes (max-30, …) and also new proposed “local slopes” (slope-30-60, …) are all evaluated in the decreasing phase of the QST (with Wmax as the starting point), so with negative values.
We will be more explicit int the materials and methods section: we propose to add information at l. 236: “...taking tmax as the starting point (...). The steepest slopes have therefore the most negative values.”
Higher values of SOC are associated with slope closer to 0. This is illustrated in the following graph given as an example
- L287 : the study of Johannes et al., 2017, did not investigate soil structural stability but soil structural quality (which are two different concepts)! Visual assessment methods don’t allow to assess stability.
Authors : We agree with the precision, we will correct the paragraph in that way:
stability changed to quality
- In section 3.2, I’m missing information regarding the correlation among indicators from the same test. How are slopes, txx, AUC, … correlated in the QST test ? Is it worth considering all these indicators? Same for MWDs and MAs in Le Bissonnais tests : are they correlated ? This comment also relates to L301-304 : what was the basis for this selection of criteria? Correlation analysis would help justify the selection.
Authors : Thank you for this comment. Indeed we need a clearer decision rule to select indicators from curves to compare the treatments of the trials. After calculation of the new indicators suggested by Reviewer 2 (delta slopes, delta t to be more specific to a certain part of the curve than initial slopes and t), we calculated the correlation matrix for QST indicators (below) and propose to add it as a supporting information.
From this matrix we observe that several indicators are highly redundant (r > 0.9):
- Wend, AUC Slope 300 & Slope 600
- Slopes 30 & Slope 60
- Tmax, t25 & t50
- T50 & t75, t75 & t90, t90 & t 95
- Delta t & t...
We propose to select four indicators to compare the different treatments of the three trials based on the following criteria:
- Avoid highly redundant indicators (r < 0.7), and if arbitration is necessary, choose the conceptually simplest one
- Having one indicator of each type/part of the curve :
- Start of the curve
- Local slopes (e.g. slope 30-60) for the early to intermediate mass loss
- Delta t (e.g. t50-t75) for the later mass loss
- One “global” indicator
- Select the most discriminant indicators between treatments.
According to this decision rules, we propose to keep:
- Tmax
- Slope 30-60 (for the tillage & PK trial) and slope 60-300 for the SOM trial
- t50-t75
- Wend
A more exhaustive selection but with already some arbitration (delta slopes & delta t) will be kept for comparison to MWDs & soil properties in Table 2.
We will adapt the result section accordingly.
- Fig. 4 Wmax – Wt0 are given as unitless, even though in materials and methods the authors talk about mass loss. This is confusing. The same problem arise for slopes, Wend, AUC. The confusion seems to stem from the fact that the indicators are not well explained in the M&M section (one speaks of ‘mass loss’ whereas it should be ‘relative mass loss’).
Authors : Indeed, as clarified earlier, all mass in our graphs and indicators are “relative” to the maximum mass (Wmax) reached after immersion. We will clarify in MM section and be more explicit when naming the indicators linked to max.
- So our indicator Wmax – Wt0 is indeed unitless
- More details in the answer of comment here above (l.225, l.240)
- L336-337 : as mentioned above, it is not correct to attribute single mechanisms to each of the three tests of Le Bissonnais. In the FW-fast wetting test (MWD1), all mechanisms are involved (slaking, dispersion, diff. swelling) except mechanical breakdown. In the SW-slow wetting test (MWD2), dispersion and diff. swelling can be present. So the fact that the QST indicators of L335 are better correlated with FW than SW does not necessarily indicate that slaking is the main mechanisms, but merely that slaking significantly contributes to breakdown in the QST, making the correlation stronger with MWD1. This is a subtle but important difference.
Authors : Thank you for this point. We propose to reformulate accordingly l. 336 : “... , which indicates that slaking significantly contributes to the initial stage of the QST.”
- L337-340 : the ‘problem’ with t50, t75, etc… is that these indicators inherently also include what happens during the early times. Wouldn’t it be much more relevant to consider the time it takes for the sample to loose 25% of its mass, then the time from 25 to 50%, then the time from 50 to 75, etc … ? In this way, the various indicators can be expected to be much more independent, and interpretation of processes would also be facilitated. If indeed dispersion and diff. swelling are the main mechanisms at later stages, then one would expect even stronger correlations between MWD2 and, say, the time needed to lose 75-90% of the mass than between MWD2 and t90.
Authors : We welcome this remark and the idea of new indicators more independent from each other. We integrated these new indicators in our code and analysed the results. It provided new insights, and we prefer to use them rather than former Slopes & t’s that were removed, to avoid highly redundant indicators (see earlier comment on redundancy analysis). We have to rewrite the result section accordingly in the revised version. (see also our answer to comment L234-236).
- L343 : indeed ! … and from differential swelling. If the authors are aware of this, then why present the Le Bissonnais tests in the wrong way in materials and methods ?
Authors : Agreed, see above the proposed changes
- L345 I think the authors misunderstand the FW test of Le Bissonnais. Slaking plays an important role in this test, but it is not the only mechanism at play. It is the difference between the SW and FW tests that predominantly reflects the effect of slaking (though slaking may actually also facilitate dispersion, so the effects are not simply additive). One cannot say that the FW test alone reflects the sole effect of slaking.
Authors : We take note of these important remarks. We propose the following reformulation l. 343 – 349 : “... from the fast and slow wetting tests of Le Bissonnais. It is also worth to mention that time of wetting of the soils of our study was relatively short (< 2 min), as indicated by the release of air bubbles from soil. We therefore advocate that indicators from the initial stage of the curve provide information much more specific to slaking than the fast wetting test of Le Bissonnais, lasting 10 minutes, which largely exceeds the time during which slaking occurs.”
- L348 indeed, the continuous measurements of QST would allow to better discriminate between mechanism on a single sample, something that cannot be done with the le Bissonnais test. Nevertheless, as suggested above, using time intervals (t0-25, t25-50,t50-75, …) should allow to better discriminate between processes than t0-25, t0-50, t0-75, etc. Furthermore, the difference between MWD1 and MWD2 of Le Bissonnais should better reflect the sole effect of slaking than MWD1 alone, so testing correlations between QST results (early times) and (MWD2-MWD1) might be of interest.
Authors : Thank you very much for these remarks providing new insights and perspectives of improvement. As answered here above, we computed now all the new proposed indicators and analysed them and we will integrate them in the text. We also explored the difference between MWD1 and MWD2. This will be commented in the new version of the manuscript.
- The same reasoning would be applied to the slopes : slope max-60, 60-300, 300-600.
Authors : We fully agree, deltas for slopes were calculated and will be analysed, too
- L385 : again, it is not correct to state that VESS assesses structural stability (it is a measures of quality)
Authors : “stability” changed to “quality”
- L421 : this is rather speculative. It is not certain at all that this would be the case. FYM is very different from green manures or crop residues (much lower C/N ratio right from the start) and therefore will not interact with microbial life in the same way.
Authors : We agree to remove this speculative statement about the timing of FYM application.
- L424-425 : I think I sort of understand what the authors mean, but the sentence is awkward and must be clarified
Authors : Proposition of reformulation l. 423: “...QST indicator (Fig. 6). However, indicators from the late part of QST curves (slope max 300 and slope max 600) and global indicators (Wend, AUC) tend to discriminate better between tillage treatments.”
- L438-440: Again, I’m not aware that chloride can have such effects on soils with permanent charge. There is no mention of the effect of chloride on stability in the paper of Paradelo. It is related to the monovalent cation (K). This discussion must be revised.
Authors : Proposition of revision l. 437 ff : “... the working assumption that KCl application might decrease soil structural stability (Paradelo et al. 2016) was not verified. This might be due to a relatively short-lived destructuring effect of KCl, since the last application occurred in the summer of 2016, almost three years before soil sampling. The beneficial effect of K fertilization on crop production and restitution of organic matter to soil might also have counteracted a potentially negative short-term effect.”
- L450 aggregate stability tests such as Le Bissonnais or drop impact tests (Imeson and Vis) where developed in view of relating the results to soil erodibility or even as a way of parameterizing soil detachment during rainfall in erosion models. In that respect, working with aggregates seems fairly relevant (particularly in case of cultivated tilled soils). As a means to assess soil quality, the larger sample size used in the QST seems relevant. Whether QST results could be used to assess soil erodibility remains to be proven, even though the good correlations with Le Bissonnais results are encouraging.
Authors : Again, we agree with the view of the reviewer. The link between QST and soil erodibility need new experiments/data. We propose to add the following complement of information l. 452:
“...inferior or superior equivalent diameter). Nevertheless, the relevance of QST curves to assess soil erodibility needs to be verified.”
- L460 : differential swelling ?
Authors : proposition of adaptation: “... three main mechanisms … (slaking, physico-chemical dispersion and differential swelling)”
- L461 see suggestions above on how to possibly improve the indicators
Authors : We computed already and we will adapt the manuscript accordingly
- L463-465 : again, there is a bit of a mixup here between ‘stability’ and ‘quality’
Authors : “stability” changed to “quality” in the revised version
- L468 : indeed, some overlap is expected with the current indicators, but this could possibly be improved (see suggestions above)
Authors : Same answer as above, we agree and we will adapt this statement in the revised version
- L484 I may be wrong (those documents are not easily accessible), but I believe that the equation developed in Laon does not relate to detachment by drop impact per se (but rather to the aggregate stability tests of Henin and Monnier as a measure of the soil sensitivity to crusting). To be checked.
Authors : Unfortunately, this soil sealing/crusting index is calculated routinely in the services of soil analysis for farmers in Belgium and we know the equation that is applied but we were unable to access the original article, so we were unable to check this point. We therefore propose the following modification l. 482:
“ However, soil resistance to sealing and crusting is routinely estimated by pedotransfer functions relating using pH in water, SOC content and clay content as input variables (Remy & Marin-Laflèche 1974), which appears complementary with the information offered by the QST.”
- L494-497 : future developments are not to be mentioned in the conclusion
Authors : Accepted, we will remove these lines in the revised version
- L503 : I do not believe this was really demonstrated in this paper (as a matter of fact, linear correlations – as demonstrated in this paper – are not compatible with thresholds) , and it should therefore not appear here.
Authors : we propose to remove the following information from the text l. 503-504:“...with the threshold value of 0.1 being a reasonable target for SOM management at field and farm scale...”
- See additional annotations in the attached pdf file
Authors : We thank you very much for this, we will take the annotations into account while preparing our revised manuscript for the next submission.
Citation: https://doi.org/10.5194/egusphere-2022-1092-RC2
Status: closed
-
RC1: 'Comment on egusphere-2022-1092', Nicolas P.A. Saby, 24 Nov 2022
The manuscript presents a very interesting study on a new laboratory method to measure soil structural stability. The subject is well introduced, the statistical analysis is thoroughly executed using a nice and robust dataset.
The manuscript is very well written, well organised, interesting and very clear. There is also a very interesting discussion on the mechanisms controlling soil disaggregation.
the paper addresses a relevant scientific question within the scope of SOIL by providing new soil health indicators determined by a new and simple method. The results are compared with results obtained using a reference method (MWD).
Moreover, the authors paid particular attention to the publication of FAIR data and codes.
I have very little comments
In the summary, l8, it is not clear that you use some indicators based on the curves to the derive information on soil structural stability. Maybe you could add one sentence to specify this point.
L89, ‘in field conditions’, it is not clear to me. SLAKES mobile application should be use in a lab or a room.
L100 there is a repetition of the definition of the QST acronym?
Is Table 1 published on a data repository? If yes, add the DOI in the caption and the text.
L230-240, the choice of the different indicators could appear arbitrary and in particular for the time (ii) or the threshold value (iii). I don’t think it is discussed. More generally, the use of the QST curves could be discussed in terms of statistics. You may use functional data analysis or curve modelling instead. Fajardo et al. use a Gompertz model and interpret the parameters of the curves. Maybe this point could be mentioned for future research. I think there is only one shourt sentence in the conclusion (l495).
Figure 4 and 5 I think we are missing some results on the MWD here to be able to compare.
L286 and Fig 3, yes there is a clear relationship but also there is a residual variability. I think this should be acknowledged in the text.
L344 some QST indicators. It is possible to be more specific?
L386. Add reference to figure 3 for clarity?
L410: something wrong with the reference.
L450 the use of a large soil volume facilitates the soil sample preparation compare to MWD.
L500 this sentence is right only for soils of central Belgium for the moment.
Citation: https://doi.org/10.5194/egusphere-2022-1092-RC1 -
AC1: 'Reply on RC1', Frédéric Vanwindekens, 19 Dec 2022
[N.B. We copied/pasted the comments and questions and added our answers after each points - main questions of referee are in bold in the text below]
The manuscript presents a very interesting study on a new laboratory method to measure soil structural stability. The subject is well introduced, the statistical analysis is thoroughly executed using a nice and robust dataset.
The manuscript is very well written, well organised, interesting and very clear. There is also a very interesting discussion on the mechanisms controlling soil disaggregation.
the paper addresses a relevant scientific question within the scope of SOIL by providing new soil health indicators determined by a new and simple method. The results are compared with results obtained using a reference method (MWD).
Moreover, the authors paid particular attention to the publication of FAIR data and codes.
• We appreciate the kind feedback of Mr Saby and we are very grateful for his time and helpful comments.
I have very little comments
• In the summary, l8, it is not clear that you use some indicators based on the curves to the derive information on soil structural stability. Maybe you could add one sentence to specify this point.
◦ The term "indicators" was mentioned earlier in the paragraph, in the objectives of the paper, but we added details also in l8, for more clearly point that these are calculated from the curve at some typical dynamic step:
◦ "For each plot, QST indicators derived from curves (e.g. relative increase or decrease in soil mass, disaggregation speed, time to reach a threshold value of mass loss …) were compared to the results of the three tests of Le Bissonnais, targeting specific mechanisms of soil disaggregation."
• [X] L89, ‘in field conditions’, it is not clear to me. SLAKES mobile application should be use in a lab or a room.
◦ The strength of SLAKES application relies on the rapid acquisition of data on soil structure with little equipment needed (the measurement can be performed inside but also outside in absence of rain or wind) à we propose to remove “in field conditions”:
◦ “Recently, the SLAKES mobile application provided encouraging results as a tool for rapid data acquisition on soil structure. The test relies on image recognition to measure…”
• [X] L100 there is a repetition of the definition of the QST acronym?
◦ Yes, a mistake, we adapted:
◦ "In this work, we evaluated the performance of a new, simple test to evaluate soil structural stability, named QuantiSlakeTest (QST), a quantitative approach of the slake test."
• [X] Is Table 1 published on a data repository? If yes, add the DOI in the caption and the text.
◦ The data were available via the "gitlab" repository (https://frdvnw.gitlab.io/qst-openscience/). But not in a real "data repository". We added this table in the data repository linked to the "SlakingLab - users, data and codes for running QuantiSlakeTests" community on Zenodo. Reference and doi : Vanwindekens, Frédéric M., & Hardy, Brieuc F. (2022). QST - open data of the article Vanwindekens & Hardy (2022) - table 1 soil properties (1.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.7405113
◦ As suggested, we will add the doi link in the text, just after the ref to the table and in the caption of the table 1.
• [X] L230-240, the choice of the different indicators could appear arbitrary and in particular for the time (ii) or the threshold value (iii). I don’t think it is discussed. More generally, the use of the QST curves could be discussed in terms of statistics. You may use functional data analysis or curve modelling instead. Fajardo et al. use a Gompertz model and interpret the parameters of the curves. Maybe this point could be mentioned for future research. I think there is only one short sentence in the conclusion (l495).
◦ This is a very good point, Thank you for this comment. Actually, curve modelling has been tested (we used double exponential models) but the increase in soil mass occurring at the start of the test makes modelling complicated and model parameters correlated poorly with MWDs and soil properties. A perspective that we see is to measure the mass of soil leaving the basket additionally to the mass of soil in the basket. This second measurement would have the advantage to get rid of the interference of air bubbles leaving the basket and therefore limiting the number of factors controlling soil mass evolution under water. At the moment our device doesn’t allow such a measurement. We believe that such a development has the potential to improve greatly curve interpretation regarding the underlying mechanisms of soil disaggregation
• [X] Figure 4 and 5 I think we are missing some results on the MWD here to be able to compare.
◦ We think reviewer point figures: 4 and 6 (same kind of figures, with boxplot). We structured the results of our paper in three main parts: i) comparison QST <-> Le Bissonnais ii) indicators against soil properties (using QST & LeBissonnais) iii) then comparing the treatments of the LTE against soil structural stability, but with a focus on the QST indicators. The two first parts were there to evaluate the QST curves in comparison with recognized ones (Le Bissonnais), while for the comparison of contrasting farming practices, we used only the QST as the central approach of the paper. That's why it did not come into our mind to add boxplot with MWD indicators.
◦ We think also that the 2 graphs are already complete. We already made a "sub-selection" of most relevant QST indicators.
◦ Nevertheless, we found the proposition of the referee and we propose to add 3 graphs in supplementary materials. We will add a reference to these graphs in the text of this part.
• [X] L286 and Fig 3, yes there is a clear relationship but also there is a residual variability. I think this should be acknowledged in the text.
◦ In this part of the results, we focus the presentation of our data on correlation coefficients between variables from soil properties and from indicators. Our aim is not to produce a predictive approach with a model. We observe a standard deviation in the repeated test of a same plot, that could be explained by local soil conditions of the sampling sites (eg. slopes, presence of roots, or earthworms' galleries). We will acknowledge that in the new version of the text.
• [X] L344 some QST indicators. It is possible to be more specific?
◦ We will specify in the text:
◦ "Therefore, we think that that QST indicators Slope~max-60~ might be more specific to slaking than the fast-wetting test of Le Bissonnais."
• [X] L386 Add reference to figure 3 for clarity?
◦ Added:
◦ "It is important to underline that the close linear relationship found between W~max~-W~t0~ and the SOC:clay ratio (see Fig. 3)) has probably no general character and was obtained here because cropland soils of the current study were sampled under standardised conditions of seeding and cover (winter wheat)."
• [X] L410: something wrong with the reference.
◦ Yes, a mistake in the manuscript (omission of a LaTeX command). We adapted:
◦ "… allowed to maintain SOC content to the initial level (about 10 g kg−1; Buysse et al., 2013b). A smaller SOC storage for a similar C input means a higher rate of mineralisation. The formation of water-stable aggregates under the effect of microbial decomposition of root biomass is a known process (Dufey et al., 1986). In the present study, the more microbially active, labile"
• [X] L450 the use of a large soil volume facilitates the soil sample preparation compare to MWD.
◦ We agree that the soil sample preparation of the QST is easier compare to MWD, but we don't think that the sample size is the cause. We will add in the paper, in the beginning of this paragraph:
◦ "The soil sample preparation is far easier than approaches based on MWD as, after removed from the steel cylinder, the sample is dried “as is” and used “as is” in the QST."
◦ If soil sample preparation is shorter, we don’t see it as critical because clods crumbling by hand and soil sieving to 3-5 mm is neither complicated nor time consuming. To our point, the main advantage is the absence of bias compared to sieving that selects aggregates of a certain size, neglecting particles < 3mm that can represent a large fraction of a soil that is frequently tilled
• [X] L500 this sentence is right only for soils of central Belgium for the moment.
◦ We agree (and we are very interested to test the approach in other pedological context). We will adapt this paragraph of the conclusion as following:
◦ "For cropland soils of the silt loam region of central Belgium, we show that the early mass loss under water is mainly related to slaking, whereas after soil saturation with water, clay dispersion becomes the dominant process of soil disaggregation. "Citation: https://doi.org/10.5194/egusphere-2022-1092-AC1
-
AC1: 'Reply on RC1', Frédéric Vanwindekens, 19 Dec 2022
-
RC2: 'Comment on egusphere-2022-1092', Anonymous Referee #2, 18 Feb 2023
First of all, apologies for the delayed response.
This article presents a relevant, novel and worthwhile adaptation of an existing soil structural stability test, and applies it to different case studies, demonstrating its potential for routine evaluations within the context of soil (structural) quality assessments. It certainly deserves to be published. Nevertheless, in my opinion, the introduction needs to be revised, a number of issues have to be clarified (see details below) and additional indicators derived from the QST curves could be tested (without requiring additional measurements).
Introduction
The introduction is quite lengthy. There is a lot of interesting information, but it does not fully serve the purpose of the paper. Part of the difficulty stems from the fact that the authors were not able to decide whether the main focus of the paper is about a new tool or about understanding the effects of various practices on their novel indicators. To me, the paper is first and foremost about a new technique to quantify soil disintegration during immersion, which is then applied to a variety of soil samples from different cropping systems from loess soils in central Belgium. The focus of the introduction should therefore be on existing techniques and why another method like the QST is relevant (L78 and subsequent). L23-49 do not really seem relevant for introducing the QST.
It is also not obvious to me how relevant it is to discuss the hierarchical levels of soil structure, except that the samples should be sufficiently large to encompass all levels so as to be representative of the soil. The fact that the authors do not revert back to these concepts in their discussion clearly indicates that this information is of limited relevance for this paper.
In English, the usual term is ‘aggregate stability’, which in the case of Le Bissonnais is indeed what is being measured (3-5 mm aggregates). It may be worthwhile discussing the distinction the authors make between aggregate stability and structural stability (if any). I would argue that QuantiSlake = structural stability, whereas Le Bissonnais = aggregate stability …
Specific comments
L9 : this is not fully correct. see comments further down on the Le Bissonnais test.
L32 I suppose liming is also an important practice in these systems, that has contributed to changes in pH and base saturation. Liming does not fall under 'organic and mineral fertilizer'. It is a soil amendment.
L35-37 please add reference
L42 add reference for the ‘critical value of SOC content’
L47 ‘…. According to farmers ….’ and scientists !
Starting L46 (till L65), the text becomes somewhat confusion. The paragraph starts about conservation tillage, then switches to erosion, then switches to aggregate stability and the hierarchical structure of aggregates : these are a lot of different ideas for a single paragraph. In general, 1 paragraph = 1 idea.
L57 Ca++ … and Mg++ (as a rule, divalent cations are much more effective than monovalent cations)
L68 The mechanisms listed here are relevant for soil sealing / soil erosion studies. I would argue that in a broader context, mechanical breakdown also occurs as a result of mechanical stresses exerted during tillage operations or traffic.
L86-87 : the method of Koestel is not really about measuring aggregate stability; Likewise, the use of VIS-NIR does not allow to measure stability : it is based on correlations, and therefore will always require reference methods.
L90-93 : there is some confusion here in the text between ‘aggregate size distribution’ (or level of aggregation) and ‘aggregate stability’. Furthermore, the “profil cultural” (and other methods listed here) do not seek to measure aggregate stability.
L95 : not sure I understand why topography and climate would influence the choice of method; please clarify
L108-116 : can be deleted, as it will be repeated in materials and methods / discussion sections
L121 explain abbreviation : CRA-W
L123 : round off to the nearest mm; the decimal really doesn't provide useful information
L148 latin names are provided here but were not given in the previous section when the crops were first named; please correct
L153 “Complete random block with split plot design” : doesn’t sound quite right. ‘split-plot’ is not compatible with ‘completely randomized’ because by definition, at least one treatment (the ‘split’) is not fully randomized. Please clarify.
L165 sampling occurred in April 2019 but no fertilizer was added since 2016 ? Please check
L171 I’m not aware that chlorides can affect soil structure in soils dominated by permanent charge, but potassium (monovalent cation) definitely does!
L198 : did you replicate measurements for each plot ? I believe Le Bissonnais recommends 3-5 replicates per plot (and per test).
L200 : the mechanisms involved in the test are not presented correctly; see comment L211 (below)
L203 : the whole point of rewetting the soil using alcohol prior to shaking in water is also to minimize slaking (and swelling and dispersion)
L203 : No sieving at 50 µm during the immersion phase in ethanol ?
L205 : Start a new paragraph at ‘Two main indicators …’
L209 : how strongly correlated are the MWD and MA indicators ? Is it worth considering both ?
L211-212 : this way of presenting the tests is not strictly correct. rapid wetting involves slaking, dispersion, but also differential swelling! Slow wetting involves differential swelling and dispersion. The mechanical breakdown test seeks to minimize slaking, swelling and dispersion, but it is not obvious that this really mimics drop impact. From the discussion, it appears that the authors are aware of all this, so why present the three test in such a caricatural way ?
Table 1 : please specify the upper and lower limits for the different particle size fractions, as this can be different from one country to another
L225 not very clear. According to L220, the first measurement should be within less than 1 second after plunging the sample into the water. What does ‘right after buoyancy’ refer to ? How do you determine the time to buoyancy? Maybe this could be illustrated graphically.
L229: start new paragraph at ‘Several ….’
L229 If I understand well, this maximum mass is going to depend on the relative rate of disintegration vs. relative rate of wetting, as well as mass of the sample (corrected for buoyancy based on the volume of the solid phase). Doesn’t this also contain information?
L234-236 this has to be better explained. The sentence seems to indicate that it is the slope of the curve at t= 30 sec (with t0 taken as tmax), but this is not what Fig 2 shows. What Fig 2 shows is not the slope of the curve, but the mass loss over a certain time interval (so the slope between t0 and t30, and not the slope at t30).
L240 what about testing (Wmax-Wend) ?
L240 indicators are presented as ‘mass at the end’ and ‘area under the curve’. So units should be ‘g’ and ‘g.sec’. But in later graphs, Wend is presented as unitless and AUC is in 1/sec. So it seems that both indicators have been normalized. This must be explained more clearly.
L241 on line 219, it says the experiment is run for approximately 1000 sec. The AUC is evidently going the depend on the length of the experiment. So is a fixed duration used for all samples ?
The curves are normalized by Wmax, but not by Wend (use Wmax as upper limit and Wenfd as lower limit for normalization): isn’t that introducing some sort of bias ? Also, in the AUC, a large part of the value may come from Wend * duration (the bottom, rectangular area in Fig 2). So a large chunk of the AUC contains the same info as Wend. Normalizing the curves by Wmax and Wend would allow to have an AUC that is independent of Wend.
L243 check this sentence
Caption of Figure 1 : What does 'managing QST laboratory' mean ? Or do you mean 'for managing and graphically displaying QST laboratory data '?
L245 why only the roots remaining in the cage ? Isn’t this a biased estimate, as smaller root fragments may have fallen through the mess during sample breakdown ? Wouldn’t it be more appropriate to pass the entire soil again through a 2-mm mesh sieve to recover all roots?
L256 Although the choice of threshold is to some extent arbitrary, using a 10% probability threshold is somewhat unusual in statistics. Using a higher threshold can sometimes be justified based on the dataset characteristics, but what is the reason for this choice here ? Please justify.
L268 there is no such Fig. 2. Fig. 2 corresponds to the QST and Fig. 3 to another correlation. Perhaps you mean Table 2?
- 3.1 it is also worthwhile noting that correlations are usually stronger with MWD1 than MWD2 (which seems to make sense)
Table 2 : It seems that slope values are expressed in negative values, which was not immediately obvious to me after reading the materials and method section. (I was expecting a higher slope to be negatively correlated with SOC content; the positive correlation stems from the fact that negative values are used). This should be clarified.
L287 : the study of Johannes et al., 2017, did not investigate soil structural stability but soil structural quality (which are two different concepts)! Visual assessment methods don’t allow to assess stability.
In section 3.2, I’m missing information regarding the correlation among indicators from the same test. How are slopes, txx, AUC, … correlated in the QST test ? Is it worth considering all these indicators? Same for MWDs and MAs in Le Bissonnais tests : are they correlated ? This comment also relates to L301-304 : what was the basis for this selection of criteria? Correlation analysis would help justify the selection.
Fig. 4 Wmax – Wt0 are given as unitless, even though in materials and methods the authors talk about mass loss. This is confusing. The same problem arise for slopes, Wend, AUC. The confusion seems to stem from the fact that the indicators are not well explained in the M&M section (one speaks of ‘mass loss’ whereas it should be ‘relative mass loss’).
L336-337 : as mentioned above, it is not correct to attribute single mechanisms to each of the three tests of Le Bissonnais. In the FW-fast wetting test (MWD1), all mechanisms are involved (slaking, dispersion, diff. swelling) except mechanical breakdown. In the SW-slow wetting test (MWD2), dispersion and diff. swelling can be present. So the fact that the QST indicators of L335 are better correlated with FW than SW does not necessarily indicate that slaking is the main mechanisms, but merely that slaking significantly contributes to breakdown in the QST, making the correlation stronger with MWD1. This is a subtle but important difference.
L337-340 : the ‘problem’ with t50, t75, etc… is that these indicators inherently also include what happens during the early times. Wouldn’t it be much more relevant to consider the time it takes for the sample to loose 25% of its mass, then the time from 25 to 50%, then the time from 50 to 75, etc … ? In this way, the various indicators can be expected to be much more independent, and interpretation of processes would also be facilitated. If indeed dispersion and diff. swelling are the main mechanisms at later stages, then one would expect even stronger correlations between MWD2 and, say, the time needed to lose 75-90% of the mass than between MWD2 and t90.
L343 : indeed ! … and from differential swelling. If the authors are aware of this, then why present the Le Bissonnais tests in the wrong way in materials and methods ?
L345 I think the authors misunderstand the FW test of Le Bissonnais. Slaking plays an important role in this test, but it is not the only mechanism at play. It is the difference between the SW and FW tests that predominantly reflects the effect of slaking (though slaking may actually also facilitate dispersion, so the effects are not simply additive). One cannot say that the FW test alone reflects the sole effect of slaking.
L348 indeed, the continuous measurements of QST would allow to better discriminate between mechanism on a single sample, something that cannot be done with the le Bissonnais test. Nevertheless, as suggested above, using time intervals (t0-25, t25-50,t50-75, …) should allow to better discriminate between processes than t0-25, t0-50, t0-75, etc. Furthermore, the difference between MWD1 and MWD2 of Le Bissonnais should better reflect the sole effect of slaking than MWD1 alone, so testing correlations between QST results (early times) and (MWD2-MWD1) might be of interest.
The same reasoning would be applied to the slopes : slope max-60, 60-300, 300-600.
L385 : again, it is not correct to state that VESS assesses structural stability (it is a measures of quality)
L421 : this is rather speculative. It is not certain at all that this would be the case. FYM is very different from green manures or crop residues (much lower C/N ratio right from the start) and therefore will not interact with microbial life in the same way.
L424-425 : I think I sort of understand what the authors mean, but the sentence is awkward and must be clarified
L438-440: Again, I’m not aware that chloride can have such effects on soils with permanent charge. There is no mention of the effect of chloride on stability in the paper of Paradelo. It is related to the monovalent cation (K). This discussion must be revised.
L450 aggregate stability tests such as Le Bissonnais or drop impact tests (Imeson and Vis) where developed in view of relating the results to soil erodibility or even as a way of parameterizing soil detachment during rainfall in erosion models. In that respect, working with aggregates seems fairly relevant (particularly in case of cultivated tilled soils). As a means to assess soil quality, the larger sample size used in the QST seems relevant. Whether QST results could be used to assess soil erodibility remains to be proven, even though the good correlations with Le Bissonnais results are encouraging.
L460 : differential swelling ?
L461 see suggestions above on how to possibly improve the indicators
L463-465 : again, there is a bit of a mixup here between ‘stability’ and ‘quality’
L468 : indeed, some overlap is expected with the current indicators, but this could possibly be improved (see suggestions above)
L484 I may be wrong (those documents are not easily accessible), but I believe that the equation developed in Laon does not relate to detachment by drop impact per se (but rather to the aggregate stability tests of Henin and Monnier as a measure of the soil sensitivity to crusting). To be checked.
L494-497 : future developments are not to be mentioned in the conclusion
L503 : I do not believe this was really demonstrated in this paper (as a matter of fact, linear correlations – as demonstrated in this paper – are not compatible with thresholds) , and it should therefore not appear here.
See additional annotations in the attached pdf file
-
AC2: 'Reply on RC2', Frédéric Vanwindekens, 19 Apr 2023
First of all, apologies for the delayed response.
This article presents a relevant, novel and worthwhile adaptation of an existing soil structural stability test, and applies it to different case studies, demonstrating its potential for routine evaluations within the context of soil (structural) quality assessments. It certainly deserves to be published. Nevertheless, in my opinion, the introduction needs to be revised, a number of issues have to be clarified (see details below) and additional indicators derived from the QST curves could be tested (without requiring additional measurements).
Authors : Dear Reviewer,
First we would like to thank you about the time spent for a serious, in depth revision of our work. Most comments appear to be relevant and will help to strengthen the manuscript. [SEE the pdf and work files attached for a better formatting)
Introduction
- The introduction is quite lengthy. There is a lot of interesting information, but it does not fully serve the purpose of the paper. Part of the difficulty stems from the fact that the authors were not able to decide whether the main focus of the paper is about a new tool or about understanding the effects of various practices on their novel indicators. To me, the paper is first and foremost about a new technique to quantify soil disintegration during immersion, which is then applied to a variety of soil samples from different cropping systems from loess soils in central Belgium. The focus of the introduction should therefore be on existing techniques and why another method like the QST is relevant (L78 and subsequent). L23-49 do not really seem relevant for introducing the QST.
Authors : Thank you for this comment. We agree that the main originality of the work stands in the new method of aggregate/structural stability measurement by soil immersion under water. Nevetherless, it is also clear that a new method can hardly be evaluated properly without an appropriate case study that offers a range of conditions and outputs for the test. The whole manuscript, from introduction to discussion, is written accordingly, with a balance between 1) QST method description and evaluation and 2) application of the QST to silt loam Luvisols (having a fragile structure) subject to contrasting soil management practices. The introduction respects this duality, with one part presenting the soil and agricultural context of the loess belt of central Belgium and the other one focussing on soil aggregation and aggregate stability measurement.
We don’t want to lose this duality because it represents the guideline and the genesis of our work, and that limiting the text to the QST method only would be a substantial loss of information, important to evaluate the method itself. We believe that the agronomic outputs of the manuscript can be of interest for many readers of SOIL and will not prevent other readers to put the focus on the methodology.
- It is also not obvious to me how relevant it is to discuss the hierarchical levels of soil structure, except that the samples should be sufficiently large to encompass all levels so as to be representative of the soil. The fact that the authors do not revert back to these concepts in their discussion clearly indicates that this information is of limited relevance for this paper.
Authors : We understand the point raised by reviewer 2 but we believe that reminding some key elements of soil aggregation is not useless at all for such a paper. First, some critical advances of research on soil aggregation date from the eighties and even earlier. Therefore, several milestone papers are not easily accessible online. Second, the ultimate goal of the test is to provide a practical, cheap and effective solution for soil stability measurement for a wide community of end-users, including those with little expertise in soil analysis. We are convinced that many of the readers of this text and final users of the test will greatly benefit from such information, in order to provide them a clearer idea of what stands behing the concept of soil aggregation (actually the feedbacks that we have received on the paper by now support this point).
- In English, the usual term is ‘aggregate stability’, which in the case of Le Bissonnais is indeed what is being measured (3-5 mm aggregates). It may be worthwhile discussing the distinction the authors make between aggregate stability and structural stability (if any). I would argue that QuantiSlake = structural stability, whereas Le Bissonnais = aggregate stability …
Authors : We thank you for this important remark. Indeed we think that it is useful to stick to “soil structural stability” rather than to “aggregate stability” because we work on soil volumes exceeding the size of soil aggregates. This information will be added in the M & M in the revised version of the text. We propose the following complement of information l. 243 :
“ Whereas Le Bissonnais (1996) and many reference methods measure the stability of soil aggregates, the QST rather measures soil structural stability, as it works on 100 cm³ soil volumes rather than on soil aggregates. Therefore we will stick to “soil structural stability” when referring to QST measurements.”
We will also check carefully to avoid the use of “soil structural stability” for other reference methods measuring “soil aggregate stability”
Specific comments
- L9 : this is not fully correct. see comments further down on the Le Bissonnais test.
Authors : Agreed, thank you for this remark. Will be revised all over the text cf. Comment l.211-212.
- L32 I suppose liming is also an important practice in these systems, that has contributed to changes in pH and base saturation. Liming does not fall under 'organic and mineral fertilizer'. It is a soil amendment.
Authors : will be replaced by “organic and mineral fertilizers and amendments”
- L35-37 please add reference
Authors : reference : Goidts & Van Wesemael (2007)
- L42 add reference for the ‘critical value of SOC content’
Authors : reference : Meersmans et al (2011)
- L47 ‘…. According to farmers ….’ and scientists !
Authors : we propose to remove "According to local farmers” in the revised version so the statement is more general
- Starting L46 (till L65), the text becomes somewhat confusion. The paragraph starts about conservation tillage, then switches to erosion, then switches to aggregate stability and the hierarchical structure of aggregates : these are a lot of different ideas for a single paragraph. In general, 1 paragraph = 1 idea.
Authors : This is where the transition between 1) the soil and agricultural context and 2) soil aggregation and soil aggregate stability measurement occurs. The order of information “soil erosion” > “soil aggregate stability” > “soil aggregation theory” doesn’t sound inappropriate to us, as i) the issue of soil erosion is the driver of our research; ii) Soil aggregate stability measurement is the most common proxy to assess soil erodibility and iii) the theory of soil aggregation is key to understand the properties of soil aggregates and mechanisms of soil aggregation and disaggregation.
In the revised version, we propose to make the transition smoother and to reorganize the paragraphs (l. 53):
“..., soil aggregate stability is often used as an indicator of soil erodibility (Barthès & Roose, 2002).
The process of soil aggregation is critical to understand the factors controlling soil aggregate stability. The theory of aggregate hierarchy of Hadas (1987)...”
- L57 Ca++ … and Mg++ (as a rule, divalent cations are much more effective than monovalent cations)
Authors : the sentence will be completed accordingly
- L68 The mechanisms listed here are relevant for soil sealing / soil erosion studies. I would argue that in a broader context, mechanical breakdown also occurs as a result of mechanical stresses exerted during tillage operations or traffic.
Authors : Thank you for this. “The resistance of soil to mechanical breakdown also improves resistance to soil compaction due to traffic on the field” will be added l. 71.
- L86-87 : the method of Koestel is not really about measuring aggregate stability; Likewise, the use of VIS-NIR does not allow to measure stability : it is based on correlations, and therefore will always require reference methods.
Authors : We agree with this remark, we propose to split the information, with spectroscopic techniques providing information on soil structure mentioned in a separated paragraph:
“Recently, the potential of some non-destructive methods on the evaluation of soil structure and aggregation has been revealed, such as...”
Accordingly, the information about the SLAKES application (l. 88-90) will be moved up and attached to the previous paragraph
- L90-93 : there is some confusion here in the text between ‘aggregate size distribution’ (or level of aggregation) and ‘aggregate stability’. Furthermore, the “profil cultural” (and other methods listed here) do not seek to measure aggregate stability.
Authors : We propose to remove from l.91 to 93 about the field methods for the evaluation of soil structure, for the sake of clarity.
- L95 : not sure I understand why topography and climate would influence the choice of method; please clarify
Authors : The idea is that topography and climate are key controls of erosion risk. The amount, frequency and intensity of precipitations and length and steepness of the slope will control erosion risks as much as soil intrinsic erodibility. Therefore, in a flat area in the temperate zone, soil aggregate stability in wet conditions may provide useful information on the risks of soil sealing and crusting or compaction whereas in other contexts the mechanical strength of soil, or a rainfall simulator approach may bring more relevant information.
- L108-116 : can be deleted, as it will be repeated in materials and methods / discussion sections
Authors : Actually we believe that l. 108-112 are absolutely necessary as they present the general approach of the work, in line with the objectives presented l. 104 – 108. We propose to remove the text from l. 113 to 116 (indeed somehow redundant with information within M&M)
- L121 explain abbreviation : CRA-W
Authors : Abbreviation is defined at first use (l. 109). For information, CRA-W refers to “Centre wallon de Recherches Agronomiques”, with letters in a confusing order due for historical reasons …
- L123 : round off to the nearest mm; the decimal really doesn't provide useful information
Authors : We fully agree, this will be done (3 significant digits)
- L148 latin names are provided here but were not given in the previous section when the crops were first named; please correct
Authors : we will check carefully that the Latin name is given for each plant species at first use in the revised version.
- L153 “Complete random block with split plot design” : doesn’t sound quite right. ‘split-plot’ is not compatible with ‘completely randomized’ because by definition, at least one treatment (the ‘split’) is not fully randomized. Please clarify.
Authors :
Some sources indicate that the split-splot design is compatible with randomized complete blocks : Dagnelie P. (2012), Principes d’experimentation. Planification des expériences et analyse de leurs résultats. See. Chap 7.
After verification, the design of the tillage trial is more correctly described as a Latin Square with one studied factor (the four tillage treatments) and one controlled factor (the 4 blocks) without repetition (sensu stricto). We therefore propose the following formulation:
“The trial includes four tillage treatments repeated four times, following a Latin square design with the blocks aligned in a row.”
- L165 sampling occurred in April 2019 but no fertilizer was added since 2016 ? Please check
Authors : P-K fertilisers are applied once per rotation, before the sugar beet, which is the starter of the three-year rotation. The last application before sampling occurred after the 2016 harvest of barley in summer 2016. The next application occurred after the 2019 harvest, so after sampling.
- L171 I’m not aware that chlorides can affect soil structure in soils dominated by permanent charge, but potassium (monovalent cation) definitely does!
Authors : we propose the following adaptation: “on the potential effect of contrasting levels of KCl application on soil structural stability (cf Paradelo et al 2016)”
- L198 : did you replicate measurements for each plot ? I believe Le Bissonnais recommends 3-5 replicates per plot (and per test).
Authors : Indeed 3 to 5 replicates are recommended by Le Bissonnais et al. (1996) but here we didn’t make any replication within one single plot because the soil sampling area was very limited (1 m²) and because we relied on field repetitions (true repetitions decrease the relevance of pseudo-replication). Moreover, since the Le Bissonnais method works on 5-10 g of aggregates 3-5 mm in diameter, the result is already an average value for many small aggregates.
- L200 : the mechanisms involved in the test are not presented correctly; see comment L211 (below)
Authors : We agree that across the document we sometimes make the following shortcut : fast-wetting = slaking. This statement must indeed be qualified and we will revisit the text accordingly. Nevertheless, in the methodology, the initial formulation seems correct (fast wetting of dry soil aggregates in water aims to test first and foremost their resistance to slaking, even if clay dispersion and differential swelling may also play a disaggregating role after 10 min under water).
We propose the following formulation: l. 200 “The first test consists in fast-wetting soil aggregates in water, exacerbating the effect of slaking”
About description of the third test, see next comment
- L203 : the whole point of rewetting the soil using alcohol prior to shaking in water is also to minimize slaking (and swelling and dispersion)
Authors : Indeed. We propose to complete the sentence l. 203 “...to test their mechanical strength while minimizing slaking, differential swelling and dispersion”
- L203 : No sieving at 50 µm during the immersion phase in ethanol ?
Authors : we followed rigorously the norm ISO FDIS 10930:2011. After each treatment, remaining aggregates were transferred to the 50 µm sieve. The < 50 µm fraction is not recuperated for any of the three test (it is calculated by difference between initial weight and fractions remaining on the sieves)
- L205 : Start a new paragraph at ‘Two main indicators …’
Authors : We agree and will adapt the manuscript in this way.
- L209 : how strongly correlated are the MWD and MA indicators ? Is it worth considering both ?
Authors : Both indicators are recommended by the norm. They are generally positively correlated for test 1 and 3 but not much for test 2. MA 3 is generally more (negatively) correlated to QST indicators than MWD3 (see Table 2). We propose to add the correlation matrix in supporting information. We propose to remove MAs from Table 2 since we make little use of it. Please fin the correlation matrix here below
- L211-212 : this way of presenting the tests is not strictly correct. rapid wetting involves slaking, dispersion, but also differential swelling! Slow wetting involves differential swelling and dispersion. The mechanical breakdown test seeks to minimize slaking, swelling and dispersion, but it is not obvious that this really mimics drop impact. From the discussion, it appears that the authors are aware of all this, so why present the three test in such a caricatural way ?
Authors : As stated earlier, we agree with this comment and we are aware of this. The text will be revised accordingly (see comment l. 200). The link between the three tests and the mechanisms of soil disaggregation will be qualified in the introduction and we will stick to the names of the three tests, “fast wetting”, “slow wetting” and “mechanical breakdown” in the other sections.
- Table 1 : please specify the upper and lower limits for the different particle size fractions, as this can be different from one country to another
Authors : We used the following limits based on NF ISO 11464 :
- Sand (coarse, > 200μm)
- Sand (fine, 50 μm - 200 μm)
- Silt (coarse, 20 μm - 50 μm)
- Silt (fine, 2 μm - 20 μm)
- Clay (< 2 μm)
- As fine/coarse sand and silt data are not used in the manuscript we will simplify the table by removing them. In the revised version of the manuscript will only stand total sand, silt and clay:
- Sand (50 – 2000 μm)
- Silt (2 μm - 50 μm)
- Clay (< 2 μm)
- L225 not very clear. According to L220, the first measurement should be within less than 1 second after plunging the sample into the water. What does ‘right after buoyancy’ refer to ? How do you determine the time to buoyancy? Maybe this could be illustrated graphically.
Authors : Indeed an illustration speaks from itself. Actually the graphs as shown on figure 2 and figure 5 only show soil mass under water after Archimedes upward buoyant force (once soil sample is completely immersed) --> if you keep soil mass data before and during immersion, you see the big drop from Archimedes buoyancy.
We will produce a didactic graph in the new proposal. Here a rapid “raw” graph. We use simple mathematical approaches based on derivatives (‘primary’, ‘secondary’) for finding the first minima. This point id considered as the “T0” and points before it are deleted (soil sample in the air, or partly in the air). The final graph that we will propose will complete the fig. 2 of the submitted version (graph with indicators).
- L229: start new paragraph at ‘Several ….’
Authors : We agree and will adapt the manuscript in this way.
- L229 If I understand well, this maximum mass is going to depend on the relative rate of disintegration vs. relative rate of wetting, as well as mass of the sample (corrected for buoyancy based on the volume of the solid phase). Doesn’t this also contain information?
Authors : As explained in our previous comment, we deleted information before and during immersion (the buoyancy). Soil mass is then normalized according to the maximum value reached (WMax=1) after the considered T0 (first minimum), so that soil mass is expressed as a relative mass.
If you now look at figure 2, you see how soil mass generally behave for the soils of this study: soil mass increases due to wetting (due to water filling in porosity) and reaches a maximum (WMax = 1) before decreasing (once mass loss due to disaggregation becomes dominant compared to mass gain by wetting).
Going back to Archimedes buoyancy: indeed, mass drop due to immersion certainly contains extra information that we didn’t exploit in the present manuscript.
To clarify, we propose the following reformulation :
l.225 - … : “soil mass drops due to Archimedes upward buoyant force (data not shown). The first value of soil mass under water is defined as the time 0 (t0) of the QST test (soil mass before and during immersion is removed from the graph and not shown here). The graph on Figure 2 illustrates how soil mass behaved under water for the soils of the present study. In the initial phase, soil mass generally increases due to water filling porosity. After a few seconds or minuts, the soil mass reaches a maximum (Wmax) before decreasing, once mass loss due to disaggregation becomes dominant compared to mass gain by wetting. Soil mass was normalised according to Wmax, so that mass value vary between 0 and 1. …"
l. 231 --> 241: “mass loss” will be changed to “relative mass loss”
Authors : We don’t fully understand the question. I will try to clarify. All our graphs and indicators are based on a relative weight. As we used the Wmax as reference (denominator), the relative weight at “Max” is 1 (see also answer for l240).
- L234-236 this has to be better explained. The sentence seems to indicate that it is the slope of the curve at t= 30 sec (with t0 taken as tmax), but this is not what Fig 2 shows. What Fig 2 shows is not the slope of the curve, but the mass loss over a certain time interval (so the slope between t0 and t30, and not the slope at t30).
Authors : We agree with the comment. But, we computed in the new manuscript new indicators based on the proposal done in comment linked to L337-340. We computed 2 new kinds of indicators: local slope (slope between max and max+30sec, slope between max+30 and max+60,..) and the “delta - t” (ex delta-t50-75, time between 50% loss and 75% weight loss). After analyses, we proposed to switch to the new version of “local slopes”, which are relevant and show interesting results. When presenting these new slopes, we will be as clear as possible to avoid misunderstanding.
- L240 what about testing (Wmax-Wend) ?
Authors : As we consider the relative soil mass (Wmax = 1), Wmax-Wend won’t provide more information than Wend itself.
- L240 indicators are presented as ‘mass at the end’ and ‘area under the curve’. So units should be ‘g’ and ‘g.sec’. But in later graphs, Wend is presented as unitless and AUC is in 1/sec. So it seems that both indicators have been normalized. This must be explained more clearly.
Authors : As explained above (see l. 228-229 of the initial version of the text), soil mass is a “relative soil mass”, without unit. The ‘y-axis’ has no unit (relative soil mass, normalized according to Wmax [g/g] = [-]) --> Wend and all other “weight” indicators (Wmax, Wt0,…) have no unit.
As a consequence the unit of slopes are only [1 / sec] = [sec-1] and AUC [1 x sec] = [sec].
It was expressed in L228-229, but we will present that more clearly (see earlier proposition of reformulation).
- L241 on line 219, it says the experiment is run for approximately 1000 sec. The AUC is evidently going the depend on the length of the experiment. So is a fixed duration used for all samples ?
Authors: Thank you for this comment. for the sake of comparability, we fixed the time to 900s for each sample, because a few of them didn't last until 1000 s. This 900 s timestep is our reference time for AUC calculation. For QST that had a duration time < 900s because of reaching a steady state before that, the curves were artificially extended to 900 for a comparable AUC between all experiences.
- The curves are normalized by Wmax, but not by Wend (use Wmax as upper limit and Wend as lower limit for normalization): isn’t that introducing some sort of bias ? Also, in the AUC, a large part of the value may come from Wend * duration (the bottom, rectangular area in Fig 2). So a large chunk of the AUC contains the same info as Wend. Normalizing the curves by Wmax and Wend would allow to have an AUC that is independent of Wend.
Authors:
As suggested, the redundancy analysis of QST indicators revealed a strong (r=0.98) positive correlation between AUC and Wend. We made a try to split the AUC into two (the rectangular area delimited by Wend and the area between the curve and Wend. Nevertheless, the curve-dependent fraction of the AUC correlates poorly to MWDs and soil properties. As the interpretation of this indicator is not clear to us, we propose not to use it.
We agree that the final time of measurement is very important for the calculation of the AUC because the longer the time considered, the more the rectangular area controls the AUC... At the moment we didn’t look to optimize the time considered for AUC calculation with any objective criterion (such as, e.g., sensitivity to soil management practices...).
Calculation of the AUC between Wmax and Wend would provide a completely different information from the current AUC. Current AUC decreases with total relative mass loss by disaggregation; the proposed AUC (calculated between Wmax and Wend) would generally increase with mass loss but would also depend on the kinetics of mass stabilization. It would also be somehow redundant with Wend.
The contribution of the rectangular area delimited by Wend is important for samples with a good structural stability (elevated Wend) but decreases with soil structural instability (low Wend).
- L243 check this sentence
Authors: Sorry for this mistake, sentence will be removed.
- Caption of Figure 1 : What does 'managing QST laboratory' mean ? Or do you mean 'for managing and graphically displaying QST laboratory data '?
Authors: The `slaker` application is used also during the QST, for
- adding sample identification and useful metadata
- Defining experience parameters (max time, …)
- starting the recording of data,
- Checking the good processing of the data collection
- Stopping the experience (computer / data part)
- We agree the sentence is not clear, we propose to adapt : “application for parametrizing and driving the experience”
- L245 why only the roots remaining in the cage ? Isn’t this a biased estimate, as smaller root fragments may have fallen through the mess during sample breakdown ? Wouldn’t it be more appropriate to pass the entire soil again through a 2-mm mesh sieve to recover all roots?
Authors: The idea raised during the experience, seeing roots in the basket. We think you are right and, in further work, we will adapt the protocol for a more rigorous quantification of root content. We certainly underevaluated the root biomass, in the 2 tested treatments of the Tillage LTE. We propose to move the root biomass figure to supporting information.
- L256 Although the choice of threshold is to some extent arbitrary, using a 10% probability threshold is somewhat unusual in statistics. Using a higher threshold can sometimes be justified based on the dataset characteristics, but what is the reason for this choice here ? Please justify.
Authors: We will adapt the tables and the text considering only the significatively differences between treatments at a 5% probability threshold, more usual in statistical analyses
- L268 there is no such Fig. 2. Fig. 2 corresponds to the QST and Fig. 3 to another correlation. Perhaps you mean Table 2?
Authors: Yes, we should have written Table 2. We will correct the wrong reference in the revised version, the mistake occurs 8 times in the text
- 3.1 it is also worthwhile noting that correlations are usually stronger with MWD1 than MWD2 (which seems to make sense)
Authors: This is generally true, except for t75 to t95 which also makes sense, as the end of the curve is not expected to relate to slaking but rather to differential swelling and physico-chemical dispersion.
- Table 2 : It seems that slope values are expressed in negative values, which was not immediately obvious to me after reading the materials and method section. (I was expecting a higher slope to be negatively correlated with SOC content; the positive correlation stems from the fact that negative values are used). This should be clarified.
Authors: Slopes (max-30, …) and also new proposed “local slopes” (slope-30-60, …) are all evaluated in the decreasing phase of the QST (with Wmax as the starting point), so with negative values.
We will be more explicit int the materials and methods section: we propose to add information at l. 236: “...taking tmax as the starting point (...). The steepest slopes have therefore the most negative values.”
Higher values of SOC are associated with slope closer to 0. This is illustrated in the following graph given as an example
- L287 : the study of Johannes et al., 2017, did not investigate soil structural stability but soil structural quality (which are two different concepts)! Visual assessment methods don’t allow to assess stability.
Authors : We agree with the precision, we will correct the paragraph in that way:
stability changed to quality
- In section 3.2, I’m missing information regarding the correlation among indicators from the same test. How are slopes, txx, AUC, … correlated in the QST test ? Is it worth considering all these indicators? Same for MWDs and MAs in Le Bissonnais tests : are they correlated ? This comment also relates to L301-304 : what was the basis for this selection of criteria? Correlation analysis would help justify the selection.
Authors : Thank you for this comment. Indeed we need a clearer decision rule to select indicators from curves to compare the treatments of the trials. After calculation of the new indicators suggested by Reviewer 2 (delta slopes, delta t to be more specific to a certain part of the curve than initial slopes and t), we calculated the correlation matrix for QST indicators (below) and propose to add it as a supporting information.
From this matrix we observe that several indicators are highly redundant (r > 0.9):
- Wend, AUC Slope 300 & Slope 600
- Slopes 30 & Slope 60
- Tmax, t25 & t50
- T50 & t75, t75 & t90, t90 & t 95
- Delta t & t...
We propose to select four indicators to compare the different treatments of the three trials based on the following criteria:
- Avoid highly redundant indicators (r < 0.7), and if arbitration is necessary, choose the conceptually simplest one
- Having one indicator of each type/part of the curve :
- Start of the curve
- Local slopes (e.g. slope 30-60) for the early to intermediate mass loss
- Delta t (e.g. t50-t75) for the later mass loss
- One “global” indicator
- Select the most discriminant indicators between treatments.
According to this decision rules, we propose to keep:
- Tmax
- Slope 30-60 (for the tillage & PK trial) and slope 60-300 for the SOM trial
- t50-t75
- Wend
A more exhaustive selection but with already some arbitration (delta slopes & delta t) will be kept for comparison to MWDs & soil properties in Table 2.
We will adapt the result section accordingly.
- Fig. 4 Wmax – Wt0 are given as unitless, even though in materials and methods the authors talk about mass loss. This is confusing. The same problem arise for slopes, Wend, AUC. The confusion seems to stem from the fact that the indicators are not well explained in the M&M section (one speaks of ‘mass loss’ whereas it should be ‘relative mass loss’).
Authors : Indeed, as clarified earlier, all mass in our graphs and indicators are “relative” to the maximum mass (Wmax) reached after immersion. We will clarify in MM section and be more explicit when naming the indicators linked to max.
- So our indicator Wmax – Wt0 is indeed unitless
- More details in the answer of comment here above (l.225, l.240)
- L336-337 : as mentioned above, it is not correct to attribute single mechanisms to each of the three tests of Le Bissonnais. In the FW-fast wetting test (MWD1), all mechanisms are involved (slaking, dispersion, diff. swelling) except mechanical breakdown. In the SW-slow wetting test (MWD2), dispersion and diff. swelling can be present. So the fact that the QST indicators of L335 are better correlated with FW than SW does not necessarily indicate that slaking is the main mechanisms, but merely that slaking significantly contributes to breakdown in the QST, making the correlation stronger with MWD1. This is a subtle but important difference.
Authors : Thank you for this point. We propose to reformulate accordingly l. 336 : “... , which indicates that slaking significantly contributes to the initial stage of the QST.”
- L337-340 : the ‘problem’ with t50, t75, etc… is that these indicators inherently also include what happens during the early times. Wouldn’t it be much more relevant to consider the time it takes for the sample to loose 25% of its mass, then the time from 25 to 50%, then the time from 50 to 75, etc … ? In this way, the various indicators can be expected to be much more independent, and interpretation of processes would also be facilitated. If indeed dispersion and diff. swelling are the main mechanisms at later stages, then one would expect even stronger correlations between MWD2 and, say, the time needed to lose 75-90% of the mass than between MWD2 and t90.
Authors : We welcome this remark and the idea of new indicators more independent from each other. We integrated these new indicators in our code and analysed the results. It provided new insights, and we prefer to use them rather than former Slopes & t’s that were removed, to avoid highly redundant indicators (see earlier comment on redundancy analysis). We have to rewrite the result section accordingly in the revised version. (see also our answer to comment L234-236).
- L343 : indeed ! … and from differential swelling. If the authors are aware of this, then why present the Le Bissonnais tests in the wrong way in materials and methods ?
Authors : Agreed, see above the proposed changes
- L345 I think the authors misunderstand the FW test of Le Bissonnais. Slaking plays an important role in this test, but it is not the only mechanism at play. It is the difference between the SW and FW tests that predominantly reflects the effect of slaking (though slaking may actually also facilitate dispersion, so the effects are not simply additive). One cannot say that the FW test alone reflects the sole effect of slaking.
Authors : We take note of these important remarks. We propose the following reformulation l. 343 – 349 : “... from the fast and slow wetting tests of Le Bissonnais. It is also worth to mention that time of wetting of the soils of our study was relatively short (< 2 min), as indicated by the release of air bubbles from soil. We therefore advocate that indicators from the initial stage of the curve provide information much more specific to slaking than the fast wetting test of Le Bissonnais, lasting 10 minutes, which largely exceeds the time during which slaking occurs.”
- L348 indeed, the continuous measurements of QST would allow to better discriminate between mechanism on a single sample, something that cannot be done with the le Bissonnais test. Nevertheless, as suggested above, using time intervals (t0-25, t25-50,t50-75, …) should allow to better discriminate between processes than t0-25, t0-50, t0-75, etc. Furthermore, the difference between MWD1 and MWD2 of Le Bissonnais should better reflect the sole effect of slaking than MWD1 alone, so testing correlations between QST results (early times) and (MWD2-MWD1) might be of interest.
Authors : Thank you very much for these remarks providing new insights and perspectives of improvement. As answered here above, we computed now all the new proposed indicators and analysed them and we will integrate them in the text. We also explored the difference between MWD1 and MWD2. This will be commented in the new version of the manuscript.
- The same reasoning would be applied to the slopes : slope max-60, 60-300, 300-600.
Authors : We fully agree, deltas for slopes were calculated and will be analysed, too
- L385 : again, it is not correct to state that VESS assesses structural stability (it is a measures of quality)
Authors : “stability” changed to “quality”
- L421 : this is rather speculative. It is not certain at all that this would be the case. FYM is very different from green manures or crop residues (much lower C/N ratio right from the start) and therefore will not interact with microbial life in the same way.
Authors : We agree to remove this speculative statement about the timing of FYM application.
- L424-425 : I think I sort of understand what the authors mean, but the sentence is awkward and must be clarified
Authors : Proposition of reformulation l. 423: “...QST indicator (Fig. 6). However, indicators from the late part of QST curves (slope max 300 and slope max 600) and global indicators (Wend, AUC) tend to discriminate better between tillage treatments.”
- L438-440: Again, I’m not aware that chloride can have such effects on soils with permanent charge. There is no mention of the effect of chloride on stability in the paper of Paradelo. It is related to the monovalent cation (K). This discussion must be revised.
Authors : Proposition of revision l. 437 ff : “... the working assumption that KCl application might decrease soil structural stability (Paradelo et al. 2016) was not verified. This might be due to a relatively short-lived destructuring effect of KCl, since the last application occurred in the summer of 2016, almost three years before soil sampling. The beneficial effect of K fertilization on crop production and restitution of organic matter to soil might also have counteracted a potentially negative short-term effect.”
- L450 aggregate stability tests such as Le Bissonnais or drop impact tests (Imeson and Vis) where developed in view of relating the results to soil erodibility or even as a way of parameterizing soil detachment during rainfall in erosion models. In that respect, working with aggregates seems fairly relevant (particularly in case of cultivated tilled soils). As a means to assess soil quality, the larger sample size used in the QST seems relevant. Whether QST results could be used to assess soil erodibility remains to be proven, even though the good correlations with Le Bissonnais results are encouraging.
Authors : Again, we agree with the view of the reviewer. The link between QST and soil erodibility need new experiments/data. We propose to add the following complement of information l. 452:
“...inferior or superior equivalent diameter). Nevertheless, the relevance of QST curves to assess soil erodibility needs to be verified.”
- L460 : differential swelling ?
Authors : proposition of adaptation: “... three main mechanisms … (slaking, physico-chemical dispersion and differential swelling)”
- L461 see suggestions above on how to possibly improve the indicators
Authors : We computed already and we will adapt the manuscript accordingly
- L463-465 : again, there is a bit of a mixup here between ‘stability’ and ‘quality’
Authors : “stability” changed to “quality” in the revised version
- L468 : indeed, some overlap is expected with the current indicators, but this could possibly be improved (see suggestions above)
Authors : Same answer as above, we agree and we will adapt this statement in the revised version
- L484 I may be wrong (those documents are not easily accessible), but I believe that the equation developed in Laon does not relate to detachment by drop impact per se (but rather to the aggregate stability tests of Henin and Monnier as a measure of the soil sensitivity to crusting). To be checked.
Authors : Unfortunately, this soil sealing/crusting index is calculated routinely in the services of soil analysis for farmers in Belgium and we know the equation that is applied but we were unable to access the original article, so we were unable to check this point. We therefore propose the following modification l. 482:
“ However, soil resistance to sealing and crusting is routinely estimated by pedotransfer functions relating using pH in water, SOC content and clay content as input variables (Remy & Marin-Laflèche 1974), which appears complementary with the information offered by the QST.”
- L494-497 : future developments are not to be mentioned in the conclusion
Authors : Accepted, we will remove these lines in the revised version
- L503 : I do not believe this was really demonstrated in this paper (as a matter of fact, linear correlations – as demonstrated in this paper – are not compatible with thresholds) , and it should therefore not appear here.
Authors : we propose to remove the following information from the text l. 503-504:“...with the threshold value of 0.1 being a reasonable target for SOM management at field and farm scale...”
- See additional annotations in the attached pdf file
Authors : We thank you very much for this, we will take the annotations into account while preparing our revised manuscript for the next submission.
Citation: https://doi.org/10.5194/egusphere-2022-1092-RC2
Frédéric M. Vanwindekens and Brieuc F. Hardy
Data sets
Full git repository - QuantiSlakeTest - qst-openscience Vanwindekens F. M. and Hardy B. F. https://gitlab.com/FrdVnW/qst-openscience
Model code and software
R-package - slaker - Analysing the data of QuantiSlakeTest approach. R-package and Web Application Frédéric M. Vanwindekens and Roisin C. https://gitlab.com/FrdVnW/slaker
Notebook with codes, figures and tables - qst-openscience Vanwindekens F. M. and Hardy B. F. https://frdvnw.gitlab.io/qst-openscience/
Video supplement
A visualisation of the QuantiSlakeTest, comparing two contrasted samples Frédéric M. Vanwindekens https://youtu.be/G9UweThvHYI
Frédéric M. Vanwindekens and Brieuc F. Hardy
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 533 | 220 | 20 | 773 | 10 | 11 |
- HTML: 533
- PDF: 220
- XML: 20
- Total: 773
- BibTeX: 10
- EndNote: 11
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1