Soil Organic Carbon Estimation and Transfer Framework in Agricultural Areas Based on Spatiotemporal Constraint Strategy Combined with Active and Passive Remote Sensing

Round 1

Reviewer 1 Report (Previous Reviewer 2)

Comments and Suggestions for Authors

The issues raised in the previous review have been largely addressed, but there are still a few minor concerns that remain.

1. The abstract, we usually used multi-spectrum (MS) data (passive remote sensing data) for SOC estimation and mapping. why using SAR data? In other words, the research background or the gap knowledge was not clear.

2. For results of the abstract, I want to see results like the estimation accuracy of SOC based on MS data alone, then the results based on the combination of MS data and active remote sensing data, as well as the results based on the combination of MS data and active remote sensing data and other covariates such as terrain factors. Then I can understand which data could achieve best prediction or after adding other data except MS data, the prediction accuracy improved or not.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report (New Reviewer)

Comments and Suggestions for Authors

This study discusses the potential of active and passive remote sensing for SOC estimation modeling in Manitoba, Canada. The SAR, MS, TB data are incorporated with five machine learning regression models. The research methodology is scientific, and the data analysis is rigorous, which is significant for understanding SOC estimation and transfer framework proposed in agricultural areas. However, the innovation is not very outstanding, especially this paper does not put forward a new method.

Specific Suggestions:

(1) This manuscript compared five advanced machine learning regression. However, it just introduced four machine learning models in section 3.1.

 (2) This manuscript collected 110 samples in 2012 and 252 samples in 2016. How to make sure the representative of the entire area with considering the impacts from soil texture, various crops.

(3) In your manuscript, I would suggest to use “R²”(coefficient of determination), where the goal is to evaluate the performance of different models.

(4) This manuscript analyzes the sensitivity of multi-source remote sensing features and SOC by Pearson’s correlation coefficient. The Pearson’s correlation coefficient is a good way to find the linear relationship between different factors. However, it is poor to explain the nonlinear relationships. Could you give other ways to describe the contributions between SOC and other data?

(5) This manuscript should give the solution about the training samples and estimation samples. It may impact the accuracy of result.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The article ‘Soil organic carbon estimation and transfer framework in agricultural areas based on spatiotemporal constraint strategy combined with active and passive remote sensing’ aims to evaluate the performance of five machine learning regression models for estimating soil organic carbon (SOC), using active and passive remote sensing data, including Synthetic Aperture Radar, Brightness Temperature, and multispectral data. The study employs two field observation datasets (SMAPVEX12 and SMAPVEX16-MB) for model training and validation, aiming to determine whether multi-source remote sensing data enhances SOC estimation accuracy, and whether the machine learning models are spatially transferable.  Additionally, the potential of incorporating SOC data to improve soil moisture retrieval is explored.

The topic is relevant, and the concept of estimating SOC using machine learning models is both interesting and aligns well with the scope of the journal. However, I consider that the following points must be addressed in order to adequately assess the results, discussion, and conclusion section. Notably, there is a clear disconnect between the methodology and the results. It is unclear which specific methods and data were applied to generate the findings presented in Figures 4–7.

1.       Introduction: The introduction lacks coherence and the English requires revision for clarity. Furthermore, the third objective related to soil moisture estimation is not connected to the primary objectives of this study, detracting from the article’s focus on SOC estimation. It would be more suitable for a separate article.

 

2.       Methodology: The methodology is insufficiently detailed to fully understand the SOC estimation process. Reliance on Figure 3 and the accompanying text in the methods section does not provide enough information about the functionality of the developed framework. Specific areas requiring improvement include:

·         The description of the ground-based data sampling strategy is incomplete. Key details such as sampling times, frequency, depth, and instruments used in the SMAPVEX12 and SMAPVEX16-MB datasets should be provided.

·         It is unclear whether quality control was performed on the data.

·         The reasoning behind the Sentinel-1A data period extending beyond the duration of the ground campaign data is not explained.

·         The explanation of machine learning algorithms is superficial. A detailed description of the equations, hyperparameters, and kernel functions is essential to interpret the results accurately, support the conclusions, and ensure reproducibility.

·         Information about the percentage split of test and validation samples should be included, along with confirmation of consistency across all models.

Given these issues, I recommend rejecting this article with an invitation to resubmit once these points are adequately addressed. Ensuring that the methodology, data quality, and coherence of objectives are improved will significantly enhance the article’s scientific impact.

Comments on the Quality of English Language

The article's coherence and language require revision to enhance clarity and readability.

Reviewer 2 Report

Comments and Suggestions for Authors

This paper aims at investigating the synergistic use of multi remote sensing data with different MLR methods for SOC estimation. As such, it should interest the readers of Remote Sensing. However, there are still some improvements need to be done.

 1. The results section in the abstract is too scattered. It should be more organized, highlighting the main findings, and supported with data.

2. Line 25-26, the manuscript was about SOC estimation, while here it shifts to stating that SOC can improve soil moisture prediction, which seems like a reversal of the main argument.

3. The introduction should be improved. Line 48, challenges remain, what challenges? It would be necessary to clarify which specific challenges are the focus, or which ones you can address. In other words, the gap knowledge of this manuscript was not clear. On this basis, the innovation of this paper needs to be strengthened.

4. The author mentioned the spatiotemporal transferability of SOC estimation. This indeed a challenge, while after reading the manuscript, I do not think the manuscript has addressed this issue. It seems to be just a simple model validation using different dataset.

5. Line 81, It would be better to verify soil moisture can improve SOC estimation,as SOC prediction is the main focus of the paper.

6. Line 82-88, delete this part.

7. How many samples for each sampling campaign?

8. Line 127, wrong equation, it should be SOM=SOC×1.724  

9. The SOPT and Sentinel data were bare soil or soil with crops? If using bare soil data, then the spectral index such as NDVI may not be very meaningful. Also, has the author considered using remote sensing data from different time periods?

10. Line 246-251, how to divide the dataset and what was the proportion of the training, testing sets? Please provide a clear explanation.

11. Could you list the best tuning parameters for each model?

12. Line 280, and most SAR features...check the sentence.

13. Line 355-357, please add an expatiation.  

14. 4.3 Spatial Transfer Accuracy Evaluation...I do not think this part was spatial transfer model.

15. 4.4 Improving SSM Modeling Performance by Introducing SOC Estimation Values. This part was not related to the topic of this manuscript.

Comments on the Quality of English Language

The language could be improved.