Assessing the Long-Term Evolution of Abandoned Salinized Farmland via Temporal Remote Sensing Data
Round 1
Reviewer 1 Report
The authors provide a quite extensive literature review and define their scope clearly. The study area, their methodology and their approach is described in detail and very clearly. The same applies for the description of their data and results, while their conclusions are supported by their experimental results.
The publication is an innovative application of existing methodologies, metrics and algorithms that are customized to solve the problem at hand.
It would be interesting if they can add in the conclusions section how this method can be expanded and used in other areas that are facing the same problems.
Author Response
Reviewer 1
The authors provide a quite extensive literature review and define their scope clearly. The study area, their methodology and their approach is described in detail and very clearly. The same applies for the description of their data and results, while their conclusions are supported by their experimental results.
The publication is an innovative application of existing methodologies, metrics and algorithms that are customized to solve the problem at hand.
It would be interesting if they can add in the conclusions section how this method can be expanded and used in other areas that are facing the same problems.
Response:
We appreciate your positive comments on our work. Potential applications of the idea of this method in other areas are prospected in the conclusion, thank you. (please see lines 607-610)
Author Response File: 
Author Response.docx
Reviewer 2 Report
SUMMARY
The paper addresses the research area related to the assessment of abandoned salinized farmland long-term evolution in Hetao (Northern China).
It aims to propose a novel approach for the detection dynamics of abandoned salinized farmland using time series multispectral and thermal imagery.
It aims to assess the spatio-temporal patterns to assess the spatio-temporal patterns of the abandoned salinized farmland via remote sensing data backed by a robust and accurate method, and to analyse the influence factors that affect the long-term dynamics of the salinization process in Hetao.
BROAD COMMENTs
As a general comment, the manuscript is fluent and well structured.
The paper can be accepted after minor corrections.
MINOR COMMENTs
L204 Please, consider inserting more details about the satellite data (Acquisition time, Path&Row (for Landsat data), and Tile ID (for Sentinel-2 data). It would be useful for the replicability of the experiment.
L379 Please, consider inserting the number of validation points for each thematic class considered.
L382 Please, consider citing the figure related to the validation points distribution (Figure 1).
Author Response
Reviewer 2
SUMMARY
The paper addresses the research area related to the assessment of abandoned salinized farmland long-term evolution in Hetao (Northern China).
It aims to propose a novel approach for the detection dynamics of abandoned salinized farmland using time series multispectral and thermal imagery.
It aims to assess the spatio-temporal patterns to assess the spatio-temporal patterns of the abandoned salinized farmland via remote sensing data backed by a robust and accurate method, and to analyse the influence factors that affect the long-term dynamics of the salinization process in Hetao.
BROAD COMMENTs
As a general comment, the manuscript is fluent and well structured.
The paper can be accepted after minor corrections.
Response:
We appreciate your positive comments on our work. We have followed your suggestions to improve this manuscript.
MINOR COMMENTs
L204 Please, consider inserting more details about the satellite data (Acquisition time, Path&Row (for Landsat data), and Tile ID (for Sentinel-2 data). It would be useful for the replicability of the experiment.
Response:
Thank you for your suggestion, we clarified this in the supplementary material. (please see supplementary material and line 203)
L379 Please, consider inserting the number of validation points for each thematic class considered.
Response:
We have clarified this, thank you. (please see lines 380-382)
L382 Please, consider citing the figure related to the validation points distribution (Figure 1).
Response:
We have cited the related figure, thank you. (please see lines 386-387)
Author Response File: Author Response.docx
Reviewer 3 Report
The presented paper deals with a topical issue in the literature. The phenomenon of salinization affects agriculture worldwide. The research is conducted correctly. The chosen time interval is relevant (32 years), and the resolution used is sufficient to obtain conclusive results. It is to be appreciated the correlation of the soil samples with the results obtained by remote sensing. The figures and graphics are correctly presented and relevant. However, it is proposed to modify figure 12, in the sense of emphasizing the differences between the different moments of chronology.Author Response
Reviewer 3
The presented paper deals with a topical issue in the literature. The phenomenon of salinization affects agriculture worldwide. The research is conducted correctly. The chosen time interval is relevant (32 years), and the resolution used is sufficient to obtain conclusive results. It is to be appreciated the correlation of the soil samples with the results obtained by remote sensing. The figures and graphics are correctly presented and relevant. However, it is proposed to modify figure 12, in the sense of emphasizing the differences between the different moments of chronology.
Response:
We appreciate your positive comments on our work. We have improved Figure 12 as you suggested to better depict its evolution over time, thank you. (please see Fig. 12 in the revised manuscript)
Author Response File: Author Response.docx
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
In this paper the authors present an approach to detect the abandoned salinized farmland using the dynamic of multispectral and thermal images time-series. First, the paper presents the causes and consequences of ground salinization for agriculture in Hetao China, the study area. Second, the authors set out the satellite datasources Landsat 5 TM, Landsat 7 152 ETM+, and Landsat 8 OLI acquired from 1989 to 2019. In addition, the methodology deployed 133 soil samples of electrical conductivity (EC) collected from May 2018 to October 2019 and statistical crop area of Bayannur City that includes the planting area for spring wheat, corn, and sunflower.
Relying on NDVI data the authors apply the k-means algorithm to classify land cover into the categories wheat, maize and sunflower, body of water, forest, and non-vegetative land. To distinguish between the non-vegetative land with a similar temporal VI pattern the authors analyzed thermal images. Moreover, to detect the abandoned salinized farmland they applied the salinity index (SI), temporal VI data and thermal imagery.
The results show the time evolution of the estimated and statistical areas of maize and sunflower, crop land and wheat. The area of the wheat has decreased because it has been replaced by more profitable crops such as maize and sunflower.
The authors observed a significant exponential relationship between SI and measured soil EC values and desalination trend in Hetao in the last three decades.
In this work, the interpretation of the results based on remote sensing analysis are properly explained with respect to the changes and phenomena occurred in real world such as economic policies and climate variations. Nonetheless, the article lacks of details necessary to replicate the experimental protocol. For instance, it is missing information about the selection of the initial positions of the centroids for k-means algorithms and the value of k. In addition, the authors should provide quantitative information in metrics such as F1-score to describe the performance of the k-means clustering and shape-model-fitting method to classify the crop fields and non-vegetative land. Moreover, the paper lacks detailed information about the process to calculate the estimated areas from the remote sensing images.
The paper is really difficult to read and the usage of English is definitely to improve, from the grammar to the construction of the sentences to the errors of verbal forms and confusion between adjetives and nouns.
For all these reasons I am forced to propose a reject for this paper, suggesting the authors to perform on one side the presentation of details about the methodology and experimental results to replicate the experimental protocol, on the other side, to improve the use of English.
Additional remarks:
p6, line 188, please rewrite this sentence, it is unclear: the soil samples were mixed with five times of pure water and standstill.
p7, line 233, In this study, k = 20 was used initially to explore more potential land types. How the initial positions of the centroids affect the final clusters? How were selected these initial positions? what impact has the variation of these positions on the final results?
p7, line 235, what is annually mean NDVI?
p7, line 235, were regard as non-vegetative -> were regarded as non-vegetative.
p7, line 237, a shape-model-fitting method was introduced in the next section ->a shape-model-fitting method is introduced in the next section
p7, In Figure 2, explain why, in general, the standard error is larger for the body of water?
p7, How the standard error was calculated in Figure 2? How many repetitions?
p8, line 270, Different crop fields (such as wheat and corn & sunflower) and non-vegetative land (including sand dune area, bare soil area, residential area, and abandoned salinized farm-land) in Hetao were successfully detected by k-means clustering and shape-model-fitting method. -> Please provide quantitative information based on metrics such as F1-score, precision and recall, and AUC to support this claim.
p8, line 276. Define the term ET.
p11, line 318, how were calculated the estimated areas from the remote sensing images? Explain this process carefully.
How were taken into account the different spatial resolutions of temporal VI, thermal imagery, and SI data to calculate this area?
p11, line 326, the total 326 area of maize and sunflower was surpassed that of wheat in 2000 -> In 2000 the total area of maize and sunflower surpassed the area of wheat.
p12, line 327, correct symbols x in 2.0◊105 -> 2.0 x105 ha ha and all other numbers with scientific notation.
p12, line 343 The area for each saline class was calculated from the SI imagery using the established SI-EC relationship (Figure 9). Show the SI imagery and explain carefully the process to calculate the area of the saline class. What was the the area covered by one pixel in the image?
p12, line 349, The area with extremely saline was drop to a shallow -> The area with extreme salinity decreased to low level since 2000.
p13, line 361, saline-> salinity
p15, line 386, The tension relationship -> the tense relationship
p15, line 394, last thirty decades -> last three decades
p19, line 440, k-mean->k-means, keep the consistency of the notation across the article.
p19, line 439, the advantages of the k-mean cluster render it can be deployed on any dataset -> the advantages of the k-means clustering make it deployable on any dataset.
Reviewer 2 Report
The study shown is a good example of how to use remote sensing to salinity monitoring and assessment. In my opinion, the major contribution it is not the methodology, but the information obtained and its analysis for large areas and long periods. In addition, it is remarkable the combination of different images and classification procedures in the methodology.
The paper is suitable for its publication after some revisions. I include some suggestions and comments that in my opinion, could be useful to improve the manuscript.
Introduction
A review about the estate of arts and application of remote sensing to land abandonment identification should be include, especially those related to farmland and irrigation districts. Some examples:
Löw et al. 2018. Mapping Cropland Abandonment in the Aral Sea Basin with MODIS Time Series. Remote Sensing, 10, 159. https://doi.org/10.3390/rs10020159
Vidal-Macua et al. 2018. Environmental and socioeconomic factors of abandonment of rained and irrigated crops in northeast Spain. Applied Geography, 90: 155-174. https://doi.org/10.1016/j.apgeog.2017.12.005
Yin, et al. 2018. Mapping agricultural land abandonment from spatial and temporal segmentation of Landsat time series. Remote Sensing of Environment, 210: 12-24. https://doi.org/10.1016/j.rse.2018.02.050
Zaragozi et al. 2012. Modelling farmland abandonment: A study combining GIS and data mining techniques. Agriculture, Ecosystems and Environment, 155:124-132. https://doi.org/10.1016/j.agee.2012.03.019
Objectives
I suggest rewriting the objectives. In my opinion, the general aim of the study is to analyse the long-term evolution of salinization-desalinization process in the Heteo irrigation districts, its causes and impacts. Thus, the main objectives would be 1) to determine spatiotemporal patterns of the abandonment salinized farmland with remote sensing support and 2) to analyse the influence factors that affect the long-term dynamic of the salinization process. Remote sensing is the tool to obtain the information, data.
Study area
The description of study area must be completed. None data about representation of main crops, land uses and the irrigation (hectares or % surface, type of irrigation systems) in the irrigation district was showed in this section. This information is relevant to help readers to know better the study area and the causes of de-salinization. It should be already included in this section.
Line 243 should be moved to this section
Methods
The research design is appropriate and the methods adequate. There are two mainly actions. On the one hand, land-cover discrimination and crops identification to have information about its temporal area evolution. On the other hand, to identify from non–vegetative land the abandoned salinized farmland. It should be clear in the methodology exposure. I suggest including a methodological workflow figure.
Lines 145-146. It is not methods. It is objectives
Table 1- A separation (line or empty row) between satellite platforms would improve this table
Ground Data. More details about soil sample must be include. What is the sampling depth? How located observation sites are? GPS? What is the location accuracy?
Line191-192. What do you want to say with average? Do you have more than one sampling point per pixel of Sentinel?
Results, Discussions, Conclusion
The results, discussion and conclusions are very interesting, good explained and argued.
No data on the accuracy of the identification has been shown. How reliable is the land cover classification? And the abandonment salinized farmland identification?
Exponential relationship between SI (Sentinel2) and measured EC have a R2 =0.6. How does it affect the results? Is salinity estimation influenced by tis fact? Is better in Salt-free or slightly free saline areas (pixels) than in extremely saline areas? Some discussion about it could be included.
