the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 20 May 2026
Estimating attainable soil organic carbon and farm-level limiting factors across Australia’s grain-growing regions
Abstract. Soil Organic Carbon (SOC) is a determining factor of soil health and agricultural crop productivity. The SOC level has generally declined since European settlement in Australia, primarily due to the clearing of native vegetation for agricultural purposes. To enhance soil health and crop yield, it is necessary to determine the attainable SOC potential and site-specific soil constraints that limit SOC build-up. To achieve this, we applied a Boundary Line Analysis (BLA) model to a dataset consisting of 1,782 soil sample sites, collected from 72 farms across the grain-growing regions of Australia. Laboratory measured soil properties, including clay content, electrical conductivity (EC), cation exchange capacity (CEC), pH, exchangeable sodium percentage (ESP), and silt+clay from both the topsoil (0–30 cm) and subsoil (30–60 cm) were used as predictors in two separate BLA model to assess attainable SOC levels in the topsoil (0–30 cm), representing topsoil-driven and subsoil-driven constraints, respectively. Upper-envelope functions between SOC and each predictor variable were derived using a locally estimated scatterplot smoothing (LOESS) model in the BLA framework. A separate BLA was performed for different rainfall regions across Australia to reflect differences in attainable SOC. Digital soil maps for each soil property for a case study farm in the Wimmera, Victoria, were used to illustrate how the BLA can be used to identify limiting factors at the within-field level. Next, we applied the developed BLA models to that case study farm to determine the attainable SOC at the farm scale across six key soil properties. The results of the entire regional BLA model revealed that the attainable SOC in the topsoil varied from 0.98 % (mean minimum) to 1.39 % (mean maximum), in contrast to a mean actual (measured) SOC of 0.68 % (SD = 0.37; IQR = 0.45–0.82) across the region. This signifies a sequestration potential of 0.3–0.71 % in the uppermost soil layer, depending on rainfall zones and management practices. In addition, digital maps of the case study farm generated by applying BLA-derived models to spatial layers of soil predictors determine the attainable SOC, show its spatial pattern, and identify the most limiting factors across soil depths. The resulting map also illustrates the potential SOC gain after ameliorating the constraints. This BLA model is reproducible and can be readily applied across the 72 individual farms in Australia (if they have Digital Soil Maps) to quantify attainable SOC and identify site-specific soil constraints. Thus, providing a rapid and effective decision-support tool for farmers and farm managers to implement site-specific soil and agronomic management, thereby improving soil carbon levels.
- Preprint
(3196 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 02 Jul 2026)
- RC1: 'Comment on egusphere-2026-2466', Anonymous Referee #1, 04 Jun 2026 reply
-
RC2: 'Comment on egusphere-2026-2466', Anonymous Referee #2, 09 Jun 2026
reply
The paper describes a data analysis technique based on Boundary Line Analysis to identify which soil properties limit SOC accumulation in the topsoil and to what extent. The authors draw conclusions about which values of soil properties are ideal for SOC accumulation. They then present a methodology, tested on a specific farm in Australia, for site-specific determination of SOC sequestration potential and limiting soil properties, which can support the application of targeted carbon farming practices.
Overall, the research is interesting and shows great practical potential to aid decision-making in carbon farming and help land managers identify which targeted practices to apply for the biggest gain in SOC content. The work is well-rooted in previous publications, and the methodology is explained in sufficient detail. The methods developed in this paper have the potential to be applicable internationally, however, some aspects need clarification to allow broader applicability. The paper is also weak in terms of structure and writing style. The writing requires significant revision before the paper meets publication standards. I would like to urge the authors to tighten the logical structure, the correctness of their sentences, and their grammar throughout the paper. Furthermore, the results section could be significantly shortened. Much of the analysis consists of panel-by-panel descriptions of figures, which adds limited value, as readers can extract this information directly from the figures themselves. The discussion should not describe the figures in detail, only point out relevant features and reflect on their significance. The figures are frequently described without being used to draw meaningful conclusions. In addition, some tables and figures could be excluded without sacrificing any relevant scientific findings, and most figures require improvement in terms of presentation.
Comments concerning the content of the paper:
- It is stated that this method can be applied to the 72 farms involved in the study, provided that they have digital soil maps. However, broader applicability is not discussed. Can this BLA model be applied to other farms in Australia that were not used for the fitting itself? If not, what would be required? How readily could this be applied in other countries, and what is its international relevance?
- I suggest to discuss BLA further in the introduction, including its potential weaknesses. This point also arises in the discussion where BLA is compared to linear regression. It is only stated that BLA results in higher attainable SOC estimates (L650–655), which does not necessarily mean it is more accurate. The authors should justify why estimates from BLA should be more trusted than those from linear regression.
- In section 2.5 it is unclear what is meant by a 3D data cube. What is the third dimension, is it depth, or soil parameters? I assume that it is depth, with separate RF models trained for different soil properties, but this is not clearly stated. It would also be valuable to include some reflection on how the accuracy of the digital soil map, discussed via the validation metrics in section 3.3, potentially impacts the results and the validity of the final limiting factor and attainable SOC outputs.
- The results section should be more consise, as it currently contains overly detailed descriptions of the figures without sufficient reflection. One suggestion that I believe would improve the consiceness is merging sections 3.1 and 3.2, as the observed trends are often very similar and the same statements are repeated with slightly different but largely irrelevant numbers. This would allow common relationships to be stated once, with differences discussed and reflected upon. Also, idividual subplot values should not be described exhaustively, only interesting features, outliers, and their interpretation should be highlighted.
- For instance, why is the shape the same for clay% and silt+clay%?
- In addition, parts of the results section would be better placed in the discussion (for instance, L545–550).
- Figure 10 is the main result and should receive the most attention, rather than being overshadowed by the lengthier discussion of earlier figures.
- Lines 365–375 devote considerable time to comparing variability in measured SOC values and attainable values. I understand that some comparison is useful as it shows that the attainable values are consistently higher than measured SOC. However, it is unclear why the variability itself is important and how variability in the true values and the attainable values is directly comparable. The interpretation and significance of this result should be clarified.
- Why is the SOC map not included in Figure 8? It would also be informative to include the sampling locations, if this information is not restricted.
- L55–60: The specific values from Ma et al. (2023) do not appear relevant here. This study uses SOC content (%) and therefore the stock values does not provide comparable context. It would be more useful to include values expressed as SOC content from literature. Otherwise I suggest removing these values altogether.
Comments concerning grammar, structure, and presentation:
General comments:
- "Topsoil" is consistently followed by "(0–30 cm)" throughout the text, repeated approximately 15 times. As this is a standard depth range for topsoil, it is sufficient to state this once, including in the abstract. The same applies to subsoil depth ranges.
- I attempted to point out most grammatically incorrect sentences, but there are likely more. I would like to encourage the authors to proofread their work carefully, as in its current form the writing does not meet publication standards. I would also advise to remove filler phrases that increase word count without adding scientific content (e.g., L180: “and illustrate the varied climatic gradients that influence Australia's agroecosystems”).
Abstract:
- "BLA model" → "BLA models"
- The following two sentences largely repeat the same information, and the connective "next" is not appropriate in context. These should be rewritten:"Digital soil maps for each soil property for a case study farm in the Wimmera, Victoria, were used to illustrate how the BLA can be used to identify limiting factors at the within-field level. Next, we applied the developed BLA models to that case study farm to determine the attainable SOC at the farm scale across six key soil properties."
Introduction:
- L105: "This process involves fitting a line to the data cloud edge, which is often the upper part (Hajjarpoor et al., 2018)." — This sentence could be tightened, as it is informal and imprecise.
- L105: "This 'boundary line' indicates the maximum yield, also often referred to as yield potential, under a condition when a parameter of interest lies on the x-axis." — Referring to an x-axis without a figure is quite confusing to the reader. I recommend a clearer explanation.
Materials and Methods:
- L180: Please provide a reference for the dataset, instead of a link in the main text.
- L220: "attainable SOC" → "maximum attainable SOC"
- Section 2.41 discusses only outlier removal, please chang the title accordingly
- L235: The brackets in the equation should be enlarged. The i in Qi should be a subscript.
- L245: "Nevertheless, this approach takes a lot of effort and time-consuming, particularly when dealing with big datasets." — Please fix the grammar. Please clarify what you mean by "looking at the data cloud” in the previous sentence.
- "Outlier = TRUE if Dₘ > Qᵢ; otherwise, Outlier = FALSE" should be removed, as it is not a mathematical expression and does not aid understanding. Just including "Dₘ > Qᵢ" in the text would be better.
- The brackets in Equation 2 should be enlarged to fully cover the fraction.
- L370: "smallest BLA among" should be changed to "lowest attainable SOC estimate among."
Results:
- L320: "study farms" → "study farm"
- L425: "physiologically" is not the right term here.
- L400: "results of the boundary line" → "resulting boundary line" or "results of the BLA"
- L495: "firm" → "farm"
- L500–505: "The accuracy of predicting SOC was consistently moderately good across all layers (CCC = 0.64–0.74, NNSE = 0.68–0.74), and almost no bias." — Please fix the grammar.
- L620: "However, the SOC gains if these subsoil limiting factors are removed were not illustrated (similar to Figure 10c for topsoil) because soil amelioration is very hard, and mostly impossible at depth." — Please fix the grammar.
Figures and Tables:
- Explaining the BLA abbreviation beneath each figure seems redundant.
- Figure 1: Why not just highlight the specific farm used in the case study in white? It is already clear from the yellow color which one is the Wimmera VIC SA. If the location of the specific farm is classified, then I suggest just coloring these points red as well. This representation is confusing and suggests that all of these white points were used as case study farms.
- Table 1: Consider removing the Classification ID column altogether.
- Table 2: The formatting needs some attention. Word spacing in the first column should be corrected. Maybe replace "Soil landscape grid" with "terrain derivatives." Resolution information should be moved to a separate column. A data source for proximal sensing is also missing.
- Figure 3: This flowchart would benefit from significant revision. One dashed box contains a single element and is therefore redundant. "Develop six BLA models" does not connect to the resulting topsoil and subsoil BLA models. The arrows pointing to "Six attainable SOC maps (subsoil)," "Six attainable SOC maps (topsoil)," and "Multivariate BLA model" do not have a starting point. The flowchart should be restructured to show only the key logical steps in the correct order and with correct relationships. The flowchart could also be better utilised in the main text.
- Figures 4 and 6 are low resolution and should be improved.
- Figures 5 and 7: The top value on the y-axis occasionally overlaps with the upper plot boundary, I would suggest placing it outside the plot. The legend should use descriptive labels ("Attainable SOC," "Measured SOC") rather than programmatic variable names.
- Table 3: A clearer description of how these values were obtained would be useful. Specifically, whether they represent the mean of the entire dataset or a subset. If it’s for the whole dataset, then please explain the rationale behind averaging attainable SOC across all subregions.
- Figure 8: Label placement and font size are inconsistent and should be standardised.
- Figure 9: Labelling is inconsistent and in some cases repeated (e.g., (b) appears both above and below the plot). The top row of plots are not aligned. The variable notation "SOCatt" is not explained and does not appear in the text, so please change the titles to something more descriptive.
- Figure 10: I suggest removing subplot (c), as the SOC gain is the more informative metric. The "SOCatt " notation should also be revised here.
- Tables 6, 7, 8: These tables could be removed altogether. The maps combined with relevant area percentages stated in the text would be sufficient.
Bibliography:
- All references should include a DOI where available.
Citation: https://doi.org/10.5194/egusphere-2026-2466-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 178 | 121 | 16 | 315 | 10 | 9 |
- HTML: 178
- PDF: 121
- XML: 16
- Total: 315
- BibTeX: 10
- EndNote: 9
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
In this manuscript, the authors used a boundary line approach to define attainable soil organic carbon (SOC) content based on several explanatory variables that could limit SOC storage. They developed boundary line models in southern Australia and applied them at the farm level to illustrate their use. The authors then produced an interesting map of limiting factors for the farm. The methodology is sound, and the manuscript deserves to be published with a few adaptations. Below are my comments and recommendations for improving the manuscript.
The progression is very slow in the Materials and Methods and Results sections. The authors are encouraged to adopt a sharper writing style by presenting the methodology synthetically at the beginning of the Materials and Methods section then detailing the different steps, making better use of Fig. 3, and avoiding repetitions. This would also solve the problem of information in the Materials and Methods section being mixed (for example, BLA and DSM are briefly presented in Section 2.1 Study site and dataset). The description of the subsoil model results is especially lengthy (BLA figures, attainable topsoil SOC, and limiting factor map). Adopting a broader perspective would be more efficient, by highlighting the differences compared to the topsoil models. Additionally, many sentences in the Results section read more like discussions.
The discussion section could be expanded:
Because of the LOESS method used to adjust boundary lines, no equations are shared with the readers. How do you plan to disseminate the results?
Please, see also the following minor comments:
In the introduction, some of the phrasings are too abrupt: “boosts soil water holding capacity”, “stop soil loss”.
l. 54: Why do you add “although SOC dynamics rely on carbon inputs and losses from the system”.
l. 60: Remove “are highly varied”.
l. 61-62: Please rephrase or make a better link between site-specific factors and soil-forming factors.
l. 86-87: Please rephrase “determines the SOC for a given property”.
l. 106-107: Please rephrase “under a condition when a parameter”.
l. 117-125: I don't think it is useful for the understanding of the manuscript.
l. 139: Avoid using abbreviations ESP and EC here.
l. 140-141: Rather group this with your objective (ii).
l. 144: Add if SOC is a concentration or a stock.
Fig. 1: Maybe you need to add the source of the image and of the agroecological zones in the figure caption. Add the meaning of the red and white symbols in the legend or figure caption. You can also consider removing this figure because the different regions are not used then, and only showing the outline of the entire grain growing zone in Figure 2 (then it will be quicker to say that you have stratified your dataset with the rainfall classification).
l. 160-162: Rephrase to better fit a Material and Methods section (not a Discussion section).
l. 161: “and the same commercial laboratory was used”, add that it is for soil analysis.
Title 2.2 could be simplified.
Caption of Fig. 2: Remove or move “Sites span six rainfall regimes, demonstrating the wide climatic range represented in the dataset” to section 2.2.
Table 1: Move the column “Rainfall region” to the left. Column “Classification ID” is not useful.
l. 212: Rephrase “Subsoil is about processes that control plant growth and so impact on biomass production, which relates to biomass inputs into the topsoil”.
l. 220: End with “:”.
l. 228-230: Theoretically, your kernel density approach should be enough to remove outliers. Why not here? (It is just curiosity, no modification required)
l. 240: Not useful.
l. 247: and is time-consuming.
l. 256: "but retained in the data cloud”. Remove, “data cloud” not used then.
l. 253-263: Could you please explain this more directly?
l. 302-303: “The above-mentioned steps are significant as they involve identifying the most limiting factors influencing the SOC accumulation and stabilisation for the case study farm”. Remove?
Fig. 3 is crowded. You could reduce the information to a minimum and split it between a development and test part. Refer to Fig. 3 in the text.
l. 364: Better explain how you computed a boxplot from the BLA models (actually, I don't understand how “the measured SOC demonstrated narrower variability than attainable SOC in all soil properties” in l. 367-368).
l. 405-407: And also due to correlations between topsoil and subsoil properties.
l. 413-414: “The summer 350-650 mm region showed soil containing 40-50 % clay produced a moderate level of attainable SOC (1.25 %)”. Two verbs, please rephrase.
l. 484-485: “From Figure 8, distinct spatial variability is evident throughout the case study farm, emphasising regions with differing soil properties”. Please rephrase.
l. 493: “Silt+clay represents the summation of clay and silt content”. Can be removed.
l. 495: Replace “firm” by “farm”.
l. 496: Vertosol and soil description can be moved to section 2.1.
l. 501: You are moving from the present to the past tense.
l. 505: Rephrase “Textural qualities”.
Caption of Fig. 5: Add definitions for CCC, NNSE and RMSE to understand the table without reading the text.
l. 523-526: Can be removed.
l. 530: Refer to Fig. 9.
Fig. 10: Homogenize “Min SOCatt” and “SOCatt” in the figure and caption (I guess it is the same).
l. 614-617: Please shorten.
Fig. 12. Title of panel a, above the map: is it topsoil or subsoil?
l. 655-656: “Jien et al. (2025) also demonstrated that the BLA estimated SOC sequestration potential was 2.1 times higher than that of quantile regression, reinforcing the suitability of the BLA model in determining SOC”. I would not give so much importance to this 2.1 ratio because this ratio highly depends on the arbitrary choices Jien et al. (2025) made for both methods (i.e., which quantile is used). You would find another ratio with your dataset. The main advantage of your approach is rather the ability to adjust nonlinear functions.
l. 660-664: “In contrast, the attainable topsoil SOC derived from the subsoil-based BLA model was calculated to be 1.01 % (mean minimum) -1.39% (mean maximum), indicating a further sequestration potential of 0.33-0.71%. These findings highlight…”. It is not a “further sequestration potential”, it is another way to calculate it, isn't it? Please, clarify.
l. 689-688: “Across all rainfall regions, attainable SOC is found to be higher in the topsoil compared to the model built on the subsurface soil in all soil properties. This is expected as topsoil receives direct organic matter such as plant residue, litterfall, organic compost and microbial activity”. You did not use subsoil for this reason, I don't understand why you give such an explanation.
Quadratic curve: when this term is employed alone, we cannot know whether the parabola opens upward or downward.
Homogenize the use of the terms SOC storage, sequestration, accumulation and buildup throughout the manuscript.
Consider producing a single figure with panels for the different limiting factor maps.