Multi-Source Data Integration for Sustainable Management Zone Delineation in Precision Agriculture

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Precision agriculture (PA) is generally deemed to be a contemporary advanced approach to managing agricultural production that integrates various technologies to enhance yields  and safeguarding natural resources at the same time. Currently, jointly examining EM38-MK2, basic soil properties (BSP), and satellite-derived indices is rarely found in the previous related literature and the information on the influence of classification techniques (e.g., kriging, fuzzy means, or percentile-based methods) on spatial reliability is very limited. The aims of this study were to evaluate various methods for delineating agricultural management zones using data from EM38-MK2 measurements, BSP, and satellite-derived indices (VI/BSI), with a focus on assessing their spatiotemporal stability over a five-year period (2018–2022) and determine the optimal number of zones and the minimal yet effective set of input parameters for practical application in precision agriculture. After carefully reading manuscript, the following comments and suggestions are recommended to consider to improve the manuscript.

• Abstract: The section has covered the background, aims, research methods, key results and significance of this research, no further revision is required.
• Lines 30-107: More literature published in recent three years (2023-2025) are essential for understanding the research progress and development, please supplement them in literature review process as they were not found at all in these lines.
• The Section 2 Materials and Methods were logically organized, including study area, data collection methods.
• Figure: The texts in the boxes are too small, please enlarge them to make them more recognizable.
• Section 3 Results: Significance analyses were carried out for many results in this section (e.g., Tables 1-7), please revise them.
• Figures 10 and 14: The figure legend is missing, making these colors confusing for readers, please revise it. Moreover, subtitles are required here to distinguish these figures.

Author Response

Response to Reviewer 1

We would like to express our sincere gratitude to Reviewer 1 for their thorough reading of our manuscript and for providing constructive and insightful comments. We appreciate the recognition of the manuscript’s structure, relevance, and the novelty of combining EM38-MK2, basic soil properties, and satellite-derived indices for management zone delineation. The suggestions provided were highly valuable and have contributed to improving the clarity and scientific rigor of the manuscript.

Below, we address each of the reviewer’s comments point by point. Revisions made in response to these comments are clearly marked in the manuscript using Track Changes.

Comment 1

Lines 30–107: More literature published in recent three years (2023–2025) are essential for understanding the research progress and development, please supplement them in literature review process as they were not found at all in these lines.

Response:
We thank the reviewer for this important observation. In response, we have revised the Introduction section (Lines 40–42, 60-63, 103-108) to include several recent studies from the period 2023–2025. These additions emphasize recent advances in multi-source data integration, data fusion techniques, and the application of Sentinel-2 and proximal sensors in precision agriculture. By integrating references from Kaya et al. (2025), Rodrigues et al. (2024), and Chaali et al. (2024), we aimed to strengthen the contextual background and demonstrate the continued relevance of the topic. The revised sentences have been marked using Track Changes.

Comment 2

Figure: The texts in the boxes are too small, please enlarge them to make them more recognizable.

Response:
We appreciate this useful observation. Figure 1 has been revised to improve readability by increasing the font size of all text elements within the boxes and arrows. The updated figure ensures that all workflow components are clearly visible in both digital and printed formats. The revised version has been inserted into the manuscript and marked accordingly.

Comment 3

Section 3 Results: Significance analyses were carried out for many results in this section (e.g., Tables 1–7), please revise them.

Response:
We appreciate the reviewer’s suggestion regarding the interpretation of results presented in Section 3. In response, we have expanded the explanatory text accompanying Tables 1–7 to improve clarity and highlight relevant trends and differences in zonation performance. Specifically, brief yet informative remarks were added after Tables 1–7 (Lines 390-394, 400-402, 442-445, 467-469, 494-496, 506-508 and 547-549). These additions summarize the implications of class number, input configuration, and index type on zonation stability, while avoiding excessive length.

To maintain the overall manuscript within 30 pages, we opted for concise comments that draw attention to the most significant patterns, rather than including extended discussion for each individual case. We hope this targeted approach enhances readability while fulfilling the intent of the reviewer’s suggestion.

Comment 4

Figures 10 and 14: The figure legend is missing, making these colors confusing for readers, please revise it. Moreover, subtitles are required here to distinguish these figures.

Response:
We thank the reviewer for this valuable observation. 
Figure 10 presents correlation matrices for 43 raster layers across four spatial resolutions (10 m, 20 m, 50 m, and 100 m). Including full raster names directly in each figure would result in excessive visual complexity and unreadably small fonts, while also significantly expanding the figure size and manuscript length.
To resolve this, we have:
Revised the figure captions, and uploaded a separate Supplementary Excel file (Supplementary Table S1.xlsx) that contains the full list of raster names for all four grids, organized accordingly.
For Figure 14, the challenge was even greater due to the inclusion of 131 raster layers, covering all zoning outputs derived from EM38-MK2, basic soil properties (BSP), Sentinel-2 VI and BSI indices, and their weighted combinations. To ensure that the visual layout remained legible and interpretable, only abbreviated raster labels were displayed along the axes.
To fully address this concern and enable detailed interpretation, we have added Supplementary Table S2, which provides the complete list of all 131 raster layers used in the 100 m correlation matrix shown in Figure 14. We believe that this approach offers the necessary level of detail without compromising figure clarity or exceeding the manuscript length guidelines.

We again thank Reviewer 1 for the valuable suggestions, which have helped us significantly enhance the quality and readability of our manuscript.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The article largely meets the formal and substantive requirements of Sustainability magazine: The structure follows the typical layout of Introduction, Materials and Methods, Results, Discussion, Conclusion. Key words have been correctly stated. All data sources and methodologies have been accurately documented and illustrated with numerous graphs and tables. 

The objectives of the study are clearly defined, the authors aim to:

Evaluate methods for determining management zones (MZs) in precision agriculture using data from the EM38-MK2 sensor, soil chemical properties (BSP) and vegetation indices (VI/BSI) from Sentinel-2.

Analyzed the spatio-temporal stability of these zones and determined the optimal number of classes (zones) for practical applications.

Although the term “hypothesis” is not used explicitly, the implicit hypothesis is that the integration of data from different sources allows for more stable and practical determination of management zones compared to single data sources.

The presentation of the results is very detailed, supported by numerous charts, tables and comparative analyses. Statistical indicators (including Kappa, Jaccard, Dice coefficient) were used, allowing a transparent assessment of the stability of the designated zones:

EM38 and BSP results had the highest stability (Kappa > 0.9).

Sentinel-2 showed high correlation with soil data (r > 0.85), confirming its suitability as an alternative under conditions of limited access to field data.

Single-class and multi-class analyses (2-8 classes) allow to compare the granularity and interpretability of zones under different conditions.

Synthetic presentation of results in tables and figures significantly increases the clarity of the analysis.

The argumentation is consistent and well supported by empirical data. The paper demonstrates a critical approach to the limitations of particular methods (e.g., kriging interpolation under low sampling density conditions). The authors compare classification methods (fuzzy k-means, WOA, percentile division) qualitatively and quantitatively.
The authors' final conclusions are consistent with the presented results, e.g. that 3-4 classes of management zones offer the best compromise between detail and spatial stability.

Strengths of the article include the comprehensive integration of data from three different sources. The statistical analysis used by the authors is rich and well established .
The paper provides recommendations for both data-rich farms and those with limited access to field data.

However, the text lacks a full summary of the study's limitations, and it would have been worthwhile to more openly point out possible errors related to, for example, the effect of weather conditions on Sentinel-2, or errors in BSP interpolation.

Some of the graphs could have been more concise - at times the overabundance of graphs can weigh down the reader.

The article assumes a high level of familiarity with geostatistical and classification methods (the authors demonstrate a very good knowledge of them) so a brief introduction or reference for the less experienced reader might be worth considering.

The article presents original, well-documented research and contributes to the development of precision agriculture through the integration of multi-source data. It meets the requirements of Sustainability magazine in both structure and substance.

Author Response

We would like to thank Reviewer 2 for their  evaluation and thoughtful feedback on our manuscript. We highly appreciate the recognition of our methodological approach, detailed analyses, and the practical contributions of our work. The reviewer’s constructive suggestions have been carefully considered and addressed point by point, as outlined below. All changes in the manuscript are marked using Track Changes.

Comment 1
The text lacks a full summary of the study's limitations, and it would have been worthwhile to more openly point out possible errors related to, for example, the effect of weather conditions on Sentinel-2, or errors in BSP interpolation.

Response:
We fully agree with the reviewer’s observation. A new subsection titled 4.7. Limitations of the Study has been added near the end of the Discussion section. This part explicitly addresses the main sources of uncertainty, including atmospheric influences on Sentinel-2 data, potential errors in BSP interpolation due to sample density, and sensitivity to parameter choices in classification. We believe this addition improves the transparency and completeness of the study.

Comment 2
Some of the graphs could have been more concise – at times the overabundance of graphs can weigh down the reader.

Response:
We appreciate the reviewer’s observation regarding figure density and visual complexity. In response, several actions have been taken to improve the overall clarity and readability of graphical content:
Figure 1 has been updated by increasing the font size of all text elements within the boxes and arrows, ensuring full readability in both digital and print formats.
Due to the large number of raster layers involved (43 in Figures 10, and 131 in Figure 14), abbreviated raster labels were used for visual clarity. To support detailed interpretation, we have included Supplementary Table S1 (for Figure 10) and Supplementary Table S2 (for Figure 14), which contain corellation matrices with full raster names for each layer.
We have taken care to limit the manuscript to 30 pages, in line with journal expectations, while ensuring that all key results remain visible and interpretable. We believe this solution addresses the reviewer’s concern and achieves a good balance between scientific transparency and visual accessibility.

Comment 3

The article assumes a high level of familiarity with geostatistical and classification methods (the authors demonstrate a very good knowledge of them) so a brief introduction or reference for the less experienced reader might be worth considering.

Response:
We thank the reviewer for this helpful suggestion. To improve accessibility for readers less familiar with geostatistical and clustering methods, we have added a sentence at the end of Section 2.5 referring to key references ([50, 51, 55]) that provide concise overviews of fuzzy k-means clustering, kriging interpolation, and zonation agreement metrics. These references are already cited in the manuscript and offer an accessible entry point without expanding the methodology section.
As the manuscript currently stands at 30 pages following the incorporation of all reviewer comments, further elaboration on these methods in the main text would risk exceeding journal formatting guidelines. We therefore believe that this reference-based solution achieves the desired clarity while maintaining the required manuscript length.

Comment 4

The article presents original, well-documented research and contributes to the development of precision agriculture through the integration of multi-source data. It meets the requirements of Sustainability magazine in both structure and substance.

Response:
We sincerely thank the reviewer for this generous and encouraging evaluation. We are especially grateful for the recognition of our methodological approach, the integration of diverse data sources, and the potential practical applications of our findings. The positive feedback reinforces our confidence in the relevance and quality of the submitted work.

Once again, we would like to thank Reviewer 2 for their thoughtful, constructive, and well-articulated feedback. All comments have been carefully addressed, and corresponding revisions are clearly marked in the manuscript using Track Changes. 

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

Dear Authors,

The article raises an important and topical issue that fits perfectly into the current discussion among politicians, scientists and farmers. Precision agriculture is a hot topic in agricultural economics. I believe that the topic of data integration in implementing precision agriculture is particularly timely. This article clearly and simply describes the logic behind data collection, its sources, and methods for analyzing its value. The conclusions may not be particularly surprising, but that is a good thing. The results are understandable even to less experienced researchers and farmers.

Critical comments on the text:

1. The language (as I noted in the section on language).

2. I think it is also worth taking a closer look at the section describing the research methodology – I think the reader could have been better introduced to your research approach, for example by more precisely describing the "study area" – I am not entirely sure how to compare your research study area with areas I know, so I think farmers might also have difficulty imagining how to use the results (especially since they can also be useful for small plots).

3. Look at your figures, sometimes they are missing information, descriptions or illegible - I have difficulty understanding, it is difficult for me to identify samples for figures 6, 7, 12.

4. I am not a supporter of so many sub-items either, but in this article it does not bother me that much.

Please consider my comments as sound advice. I expect you to consider them and, if you consider them important, make corrections. I am not making my decision dependent on them, as I believe the text should be published.

Kind regrds,

Comments on the Quality of English Language

Dear Editor, Dear Authors,

I am not a native speaker (my native language isn't English), but I speak it fluently and usually have no difficulty reading and understanding written texts. Unfortunately, I believe the text under review needs proofreading. I recommend the authors reread the text and ensure the English is correct. If they are confident in the language, I accept.

Sorry for the inconvenience.

Author Response

Response to Reviewer 4

We would like to sincerely thank the reviewer for the encouraging and constructive comments. Your positive assessment of the topic’s relevance and clarity is very much appreciated. We especially value your practical suggestions, which have helped us identify several areas where the manuscript could be made clearer and more accessible—particularly for readers coming from applied or non-academic backgrounds.
In response, we have made several improvements throughout the manuscript. These include clarifying the methodological section (especially regarding the study area), revising figures to improve readability and labeling, and conducting another careful review of the language to address fluency and consistency.
All revisions are clearly marked in the revised version of the manuscript. Below we respond to each of your comments in detail.

Comment 1:
The language (as I noted in the section on language).
Response:
We appreciate the reviewer’s honest remark regarding the language quality. In response, we carefully revised the manuscript to improve clarity, sentence structure, and overall readability. We paid special attention to long and complex sentences, simplifying them where appropriate without altering the intended meaning. Particular focus was placed on the Introduction, Methodology, and Discussion sections, where nuanced concepts required more accessible phrasing. We trust that the revised version now offers a smoother reading experience for both technical and non-technical audiences.

Comment 2:
I think it is also worth taking a closer look at the section describing the research methodology – I think the reader could have been better introduced to your research approach, for example by more precisely describing the "study area" – I am not entirely sure how to compare your research study area with areas I know, so I think farmers might also have difficulty imagining how to use the results (especially since they can also be useful for small plots).
Response:
We appreciate the reviewer’s constructive comment. In response, we revised Section 2.1 to clarify both the rationale for selecting the study area and its relevance to practical applications. We now specify that the selected parcels were included due to the unique availability of synchronized datasets: soil chemical analyses and EM38-MK2 scanning were performed on the same day across all fields. This alignment allowed for direct comparison between data types and strengthened the reliability of our integration methods. We also clarified that, while the parcels range in size from 16 to 92 ha, the applied methods are scale-flexible and can be implemented on fields as small as 0.5 ha. However, because Sentinel-2 data has a spatial resolution of 10 meters, we recommend applying these methods to plots above that threshold to ensure meaningful spatial interpretation. We hope this addition improves the clarity of the methodology and reinforces the generalizability of our approach for both research and on-farm applications.

Comment 3:
Look at your figures, sometimes they are missing information, descriptions or illegible - I have difficulty understanding, it is difficult for me to identify samples for figures 6, 7, 12.
Response:
Thank you for this helpful remark. We fully agree that clearer figure descriptions improve readability, especially for multi-layer visualizations. Originally, some of this detail had been omitted during manuscript length reduction, as we were working to stay within the journal's suggested 30-page limit. We have now reintroduced this information to improve clarity without substantially affecting the overall length.
Figure 6 has now been clarified in both the main text and caption. We specify that the figure includes 25 raster layers: four from EM38-MK2, five from basic soil properties, one BSI index, and five temporal layers each for NDVI, SAVI, and LCI. These were all interpolated and normalized to ensure comparability.
Also, in the revised version, we have expanded the description of Figure 7 to clarify its structure and purpose. This clarification is now added to Section 2.4. 
Figure 12 was generated directly from the correlation matrices produced during the zonation analysis at 100 m resolution. As part of the automated workflow, the x-axis labels retained original raster names (in Serbian) from the data processing environment. While this limits immediate interpretability, we addressed the issue by adding a clarifying sentence in the main text that refers to Supplementary Table S2. This table provides the full list of all input layers used in the correlation analysis, along with their English names and metadata.
We hope this clarification and supporting reference resolve the reviewer’s concern.

Comment 4:
I am not a supporter of so many sub-items either, but in this article it does not bother me that much.
Response:
We appreciate your openness and are glad to hear that the use of multiple sub-sections did not hinder the readability of the manuscript in this case. Given the multidimensional nature of the analysis—including several data types, methods, and comparison frameworks—we found that a more structured layout helped maintain clarity and flow. Thank you for your understanding.

Once again, we thank the reviewer for the thoughtful feedback, which helped improve the clarity and overall quality of the manuscript.
Regarding the English language, we undertook a thorough revision to improve clarity, sentence structure, and overall readability. Although we did not use professional editing services, we thoroughly revised the language ourselves to ensure clarity and fluency.We hope the current version meets the reviewer’s expectations.
We are grateful for your constructive feedback and encouragement, and we appreciate your recommendation to proceed toward publication.

In line with the editorial team's request to better highlight the connection with sustainability, we have updated the manuscript title, abstract, and keywords. The revised title now explicitly includes “Sustainable Management Zone Delineation,” and the abstract has been expanded to emphasize how our proposed approach supports sustainable agriculture through efficient input use and site-specific management. We also added “sustainable agriculture” as a keyword to improve indexing and align with the journal's scope.

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

A more comprehensive introduction to the state of precision agriculture in Serbia is recommended, analyzing the determining factors with the highest number of references. 

• The methodology used in this article is solid, well-structured, and rigorous. The question that arises is:

• Regarding the selected plots, are there 11 plots shown on the map in Figure 2?

• In areas with dense vegetation, how are the possible limitations of NDVI related to data saturation corrected?

• In relation to the weightings of the weighted overlap analysis (WOA) method, how are they assigned?

• It is recommended to point out the technical and operational limitations overcome with this study.

Author Response

Response to Reviewer 3
We thank Reviewer 3 for their constructive feedback and insightful suggestions. We highly appreciate the recognition of the manuscript’s methodological rigor and structure. The reviewer’s comments have helped us improve both the clarity and contextual depth of the paper. Below we address each comment in detail. All revisions are clearly marked in the manuscript using Track Changes.

Comment 1
A more comprehensive introduction to the state of precision agriculture in Serbia is recommended, analyzing the determining factors with the highest number of references.

Response:
We appreciate this valuable suggestion. In response, we have expanded the Introduction section (Lines 43-57) to provide a more detailed overview of the current state of precision agriculture in Serbia. This includes discussion of key limiting factors such as farm size structure, adoption barriers, regional disparities, and infrastructural constraints.
Additionally, we have integrated three new references from recent national and regional studies that highlight empirical findings related to digital adoption, technical readiness, and economic challenges in both Western Serbia and the Autonomous Province of Vojvodina. These sources offer relevant insight into the structural limitations and adoption potential of PA technologies in Serbian agriculture.
The revised text also clarifies the relationship between broader European trends and Serbia’s gradual but uneven uptake of PA tools. We believe this improvement provides a clearer rationale for our research objectives and strengthens the contextual foundation of the study.

Comment 2
Regarding the selected plots, are there 11 plots shown on the map in Figure 2?

Response:
This was an unintentional inconsistency that we had overlooked, and we sincerely thank the reviewer for pointing it out. The correct number of plots included in the analysis is 10, as shown in Figure 2. The previously stated number "11" appeared in three instances in the manuscript and has now been corrected to reflect the accurate count. We appreciate the reviewer’s careful observation, which helped us resolve this issue and improve the clarity of the paper.

Comment 3
In areas with dense vegetation, how are the possible limitations of NDVI related to data saturation corrected?

Response:
We thank the reviewer for raising this important technical point. NDVI is indeed known to saturate under dense vegetation canopies, which can limit its sensitivity to biomass variation.
To address this issue, our zonation was not based solely on NDVI. We also used two complementary indices—SAVI and LCI—that reduce NDVI’s limitations. SAVI adjusts for soil background influence and performs better under moderate to dense vegetation cover, while LCI utilizes Sentinel-2’s red-edge bands and is less prone to saturation effects. This multi-index approach helped enhance the robustness of classification across varying canopy conditions.
A clarifying sentence has been added to Section 2.2.3 (Lines 231-237) to explain this methodological choice.

Comment 4:
In relation to the weightings of the weighted overlap analysis (WOA) method, how are they assigned?

Response:
We thank the reviewer for this important question. The weighting coefficients in the WOA method were derived from established agronomic principles, especially those related to nutrient availability and the role of electrical conductivity in soil structure characterization.
For BSP-only zonation, we used two weighting configurations: one reflecting typical fertilization ratios for nitrogen, phosphorus, and potassium (humus 0.65, P₂O₅ 0.23, K₂O 0.12), and another that also included pH with slightly adjusted weights (humus 0.63, P₂O₅ 0.21, K₂O 0.11, pH 0.05).
For BSP + EM38-MK2 combinations, we adopted several schemes that allocated a portion of the total weight to apparent electrical conductivity (ECa) layers—specifically C1 and C0.5—typically assigning 0.2 to each. The remaining weight was then distributed across the BSP parameters, while maintaining internal consistency with standard N:P:K ratios and, in some cases, including additional variables such as pH or total nitrogen. For example:
- One configuration used humus 0.56, P₂O₅ 0.18, K₂O 0.095, pH 0.0825, and C0.5 0.0825.
- Another involved C1 0.20, C0.5 0.20, humus 0.32, P₂O₅ 0.16, K₂O 0.06, and pH 0.06.
- More complex schemes balanced humus, P₂O₅, K₂O, pH, nitrogen, and both ECa depths equally (e.g., 0.15–0.18 for each parameter).
These configurations were selected to preserve the agronomic relevance of each soil parameter while enabling integration with geophysical data. A clarifying paragraph summarizing these schemes has been added to Section 2.4 of the manuscript.

Comment 5:
It is recommended to point out the technical and operational limitations overcome with this study.

Response:
We thank the reviewer for this thoughtful and constructive comment. In response, we have added a new subsection (Section 4.8) titled “Overcoming Operational and Methodological Constraints”. This part outlines how our study addresses key limitations often encountered in precision agriculture workflows—such as the need for high sampling density, integration of heterogeneous data sources, and the complexity of classification. We also emphasize that a zonation scheme using three to four management zones was found to offer an optimal balance between agronomic resolution and implementation simplicity (as noted in Section 4.3). These improvements enhance the broader applicability and operational scalability of the proposed methodology.

In response to the reviewer’s remark regarding English quality, we have carefully revised the manuscript to improve clarity, fluency, and readability of the English language. Particular attention was given to simplifying long or complex sentences, avoiding overly technical phrasing where not necessary, and enhancing the overall coherence of the narrative. These improvements were made throughout the Introduction, Methodology, and Discussion sections, with a focus on rephrasing long sentences into more natural, concise formulations suitable for an international scientific audience. All such changes are clearly marked in the manuscript using Track Changes, and line numbers are provided where relevant (31-32, 34-35, 39-41, 42-44, 65-66, 73-74, 167-170, 173-175, 177-179, 703-705, 731-732 and 772-774).
We believe these revisions significantly enhance the linguistic quality of the manuscript and ensure that the research is communicated more effectively to readers across disciplines.

Once again, we thank Reviewer 3 for their detailed and constructive comments. Their suggestions have significantly contributed to improving the clarity, methodological transparency, and practical relevance of our manuscript.

Author Response File: Author Response.pdf