From Remote Sensing to Species Distribution Modelling: An Integrated Workflow to Monitor Spreading Species in Key Grassland Habitats

Round 1

Reviewer 1 Report

Many thanks to the authors for resubmitting their work. The manuscript reads well and it will make a nice contribution to remote sensing mdpi. I have a minor revision though. For some reason, all the figures look blur, please revise. No need to send me back the revised version of the manuscript.

Author Response

Many thanks for your time.

Best regards,

The Authors

Reviewer 2 Report

The authors use modelling techniques based on RS image classifications to examine ecological factors relating to the spread of the invasive Brachypodium genuense. I appreciate the make up of this paper, as this is an issue that has been examined in some ways before, but their approach is novel, innovative and in my opinion, correct.  I feel like I may be reading an advanced version of this (maybe second round review?) as it is very well written!

The abstract adequately justifies the importance of the study, the methods, and shows the broader implications of the study.

The introduction goes straight into European grasslands. To broaden even further, could include two lines about global importance of grasslands.

Line 24/362: cover of Brachypodium genuense (do not use scientific names as possessive)

Line 29: missing an s after '

Line 59: SDMs have

Line 72: You cite to references supporting B. genuense as a definite negative invasive - but your study is also finding additional support - so perhaps its invasive behaviour has been shown to cause (so it is less definitive)

I find the workflow to be an ultimate broader aim of the study. I would make Lines 69-83 a single paragraph and says finally we aim to provide a new workflow - as I can see using this workflow outside the European context as well.

Figure 1 and 2 are beautiful but a bit low res in my version - maybe different in the final?

108-can 2.2 and 2.3 be combined? As there are only two lines about the species and it is confusing as you mention but do not define the phenological phases here but do so about four lines later in another section

Line 126: products (add an s)

Line 144: delete the before presence/absence

Excellent methods that are reproducible - very clearly written with good reference to the literature.

Figure 3 is impressive - I would outline the photo in black just to make it stand out from the map as it almost looks part of the model. I would leave it in though...

Table 1 - I would provide a more specific table heading

3.3 In my experience, going to 1 decimal place is more suitable and less over-specific for these models; if you choose to leave it at 2, make sure all figures have 2 decimals even if ending in 0

Line 309 - combine the single sentence with the next paragraph - add The add the before ensemble

Discussion: rather than starting with the first objective, an overall statement - perhaps about the success of the ensemble model - would be more appropriate.

Line 343 - you mention another B genus (Bromopsis) so in like 345 you need to spell out Brachypodium again

Line 354 - remove furthermore. Can you give examples of other species that would benefit from this approach? You suggest that others suggest your resolution would not be suitable; you found that it was...what other applications are there?

Line 381 - it would be good to know more about the livestock and land use in the methods as this feels to come out of nowhere

Conclusions - very choppy - could be one single paragraph

Line 411 - communities/ habitat management

I agree with this last sentence - it would be good to give more examples throughout the discussion.

 

Author Response

Dear reviewer, thanks for your precious comments. We revised all the text following your suggestions.

Best regards,

The Authors

 

 

The authors use modelling techniques based on RS image classifications to examine ecological factors relating to the spread of the invasive Brachypodium genuense. I appreciate the make up of this paper, as this is an issue that has been examined in some ways before, but their approach is novel, innovative and in my opinion, correct.  I feel like I may be reading an advanced version of this (maybe second round review?) as it is very well written!

The abstract adequately justifies the importance of the study, the methods, and shows the broader implications of the study.

A: Many thanks.

The introduction goes straight into European grasslands. To broaden even further, could include two lines about global importance of grasslands.

A: Thanks, we changed the first three sentences, we modified the first reference.

Line 24/362: cover of Brachypodium genuense (do not use scientific names as possessive)

A: Thanks, we have revised as suggested.

Line 29: missing an s after '

A: Corrected, thanks.

Line 59: SDMs have

A: Corrected, thanks.

Line 72: You cite to references supporting B. genuense as a definite negative invasive - but your study is also finding additional support - so perhaps its invasive behaviour has been shown to cause (so it is less definitive)

A: Sorry, we did not well understand your comment. The aims have not focused on the effects of B. genuense on plant communities. Instead, based on this knowledge, we have focused on the distribution, and the associated topographic drivers, to map the target species for conservation purposes.

I find the workflow to be an ultimate broader aim of the study. I would make Lines 69-83 a single paragraph and says finally we aim to provide a new workflow - as I can see using this workflow outside the European context as well.

A: Thanks, we revised as suggested.

Figure 1 and 2 are beautiful but a bit low res in my version - maybe different in the final?

A: Thanks, we will provide a pdf with high resolution images.  

108-can 2.2 and 2.3 be combined? As there are only two lines about the species and it is confusing as you mention but do not define the phenological phases here but do so about four lines later in another section

A: As suggested we combined the two subsections.

Line 126: products (add an s)

A: Corrected. Thanks.

Line 144: delete the before presence/absence

A: Corrected. Thanks.

Excellent methods that are reproducible - very clearly written with good reference to the literature.

A: Many thanks.

Figure 3 is impressive - I would outline the photo in black just to make it stand out from the map as it almost looks part of the model. I would leave it in though...

A: Thanks, we modified figure 3 as suggested.

Table 1 - I would provide a more specific table heading

A: We provide more specific headings, thanks.

3.3 In my experience, going to 1 decimal place is more suitable and less over-specific for these models; if you choose to leave it at 2, make sure all figures have 2 decimals even if ending in 0

A: Corrected. Thanks.

Line 309 - combine the single sentence with the next paragraph - add The add the before ensemble

A: Corrected. Thanks.

Discussion: rather than starting with the first objective, an overall statement - perhaps about the success of the ensemble model - would be more appropriate.

A: Thanks, we completely revised the first paragraph of the discussion, inserting an overall statement of the main results.

Line 343 - you mention another B genus (Bromopsis) so in like 345 you need to spell out Brachypodium again

A: Corrected. Thanks.

Line 354 - remove furthermore. Can you give examples of other species that would benefit from this approach? You suggest that others suggest your resolution would not be suitable; you found that it was...what other applications are there?

A: Thanks, we add a sentence about application of the S2 images.

Line 381 - it would be good to know more about the livestock and land use in the methods as this feels to come out of nowhere

A: We have mentioned it in MM (study area section) “Pastoral activities strongly affect the structure and composition of grassland habitats over time. During the last century, the evolution of land use from sheep breeding into cattle and equine pasturage has favoured the spreading of competitive and not palatable species”. Lines ≈110.

Conclusions - very choppy - could be one single paragraph

A: We revised as suggested. We divided it in two main paragraphs.

Line 411 - communities/ habitat management

A: Corrected. Thanks.

I agree with this last sentence - it would be good to give more examples throughout the discussion.

A: We add a sentence with references of some examples.

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

Dear authors, I have read with interest your article.

From my point of view, it is a topical research.
The techniques presented in this paper have applicability on grassland's habitats and the applied models are successful.
In other words, it is a well-structured work, it is well documented with bibliography and most importantly, the method approached in this study has applicability to large areas of grasslands.

Author Response

Thanks for your time.

Best regards,

the authors 

Reviewer 2 Report

This study used random forest algorithm to classify the B. genuense grassland, but there are several concerns should be taken into consideration.    

1. The introduction should be improved. For instance, L50-58 should be regarded as one paragraph. The authors should review how to use remote sensing to classify grassland. Why do the authors use random forest rather than other algorithms? What is niche model and is there any other niche model? The authors should review the references about their main topics.   
2. The 40 points was collected as the field data. It’s not too much and is it possible to plot them on the map to make sure the spatial pattern of them is homogeneous.
3. Why does the value of sensitivity equal to one in table 1?
4. The analysis of topographic niche model looks like topographic spatial analysis. I’m not sure why the authors use “niche” as I can’t find the ecological niche analysis.
5. The resolution of figures are quite low and vague.

Overall, this study use random forest to classify the B. genuense grassland and use spatial analysis to understand the relationship between spatial distribution of B. genuense and topographic indices. However, the methods and conclusion are not very scientific soundness and creative.  

Author Response

Dear Reviewer,

Thank you for the precious help to improve the manuscript, we submit the revised version following all your suggestions.

Best regards,
The authors

 

 

This study used random forest algorithm to classify the B. genuense grassland, but there are several concerns should be taken into consideration.    

The introduction should be improved.

Thanks, we improved the introduction.

For instance, L50-58 should be regarded as one paragraph.

We merged these lines in one paragraph.

The authors should review how to use remote sensing to classify grassland. Why do the authors use random forest rather than other algorithms?

We ran the RF model again and compared the results with two other classification algorithms (SVM and CART). We also varied the hyperparameters to observe the performances of all models. RF was the best in terms of overall performance and reflected the real situation very well. We have revised the work according to these new results.

What is niche model and is there any other niche model? The authors should review the references about their main topics.   

Thanks, we revised the sentences using species distribution model.

1. The 40 points was collected as the field data. It’s not too much and is it possible to plot them on the map to make sure the spatial pattern of them is homogeneous.

We added the collected points (presence/absence) in the Figure 1.

2. Why does the value of sensitivity equal to one in table 1?

The sensitivity value = 1 was referred to the performance of the previous model. We rerun the RF model and other classification models (SVM and CART) under the advice of the reviewers and compared the accuracies while also performing cross-validation.

3. The analysis of topographic niche model looks like topographic spatial analysis. I’m not sure why the authors use “niche” as I can’t find the ecological niche analysis.

Thanks, we revised the sentences, choosing topographic species distribution model.

4. The resolution of figures are quite low and vague.

Corrected. In this version we provide the images at 400 dpi.

Overall, this study use random forest to classify the B. genuense grassland and use spatial analysis to understand the relationship between spatial distribution of B. genuense and topographic indices. However, the methods and conclusion are not very scientific soundness and creative.  

Reviewer 3 Report

The manuscript with the title “Integrating remote sensing and ecological modelling to assess the potential impact of Brachypodium genuense on grasslands habitat conservation” presents a study that investigates the distribution of Brachypodium genuense by using a binary Random Forest (RF) classification of a RS image and field sampled presence/absence points. At first glance, the paper takes into a worthy topic. However, the structure of the manuscript is not strong and is not well organized. The document fails to demonstrate the unique contribution compare to what’s already written in literature. I am going to suggest rejection but encouraging the authors to revise the manuscript and resubmit.

 

General revisions.

 

The title does not reflect the content of the manuscript. A title like “A new approach to monitor important grassland habitats in view of the future...” could be more appropriate.”

 

Please define Nk2 habitats or provide a reference?

 

First, the introduction needs to provide general background and context of the study; I found a lack of connection between remote sensing data and species distribution model to support the hypothesis that a combination of both improves species distribution models (lines 59-72). Maybe if the authors provide some references for lines 70-71 could help.

 

Lines 80-83 look more suitable for the methods section.

Is target (objective?) 3  cont

 

 

The methods section is too large; at some points, it is hard to follow. Some parts of the methods section look like part of the introduction or part of the discussion section. Please revise and stay focus on the methods followed. I think a general flowchart would be greatly beneficial.

 

The number of images used should be described within the methods section, not in supplementary data.

 

It is not clear how the absence data were collected. Please elaborate.

 

 

Line 142. Remove the symbol “÷.”

 

 

Line 267-268. What do you mean by “expert-based process”?

 

Section 2.9.

This is the core of the paper. Why only investigate topographic drives?

This section needs to be carefully revised because the manuscript's title suggests “integration of RS and ecological modelling,” but according to what was written, there is no integration, only extrapolation.

 

All topographic variables are highly correlated; how the authors deal with multicollinearity?

 

Please enhance the quality of all Figures.

Author Response

Dear Reviewer,

Thank you for the precious help to improve the manuscript, we submit the revised version following all your suggestions.

Best regards,
The authors

 

 

The manuscript with the title “Integrating remote sensing and ecological modelling to assess the potential impact of Brachypodium genuense on grasslands habitat conservation” presents a study that investigates the distribution of Brachypodium genuense by using a binary Random Forest (RF) classification of a RS image and field sampled presence/absence points. At first glance, the paper takes into a worthy topic. However, the structure of the manuscript is not strong and is not well organized. The document fails to demonstrate the unique contribution compare to what’s already written in literature. I am going to suggest rejection but encouraging the authors to revise the manuscript and resubmit.

General revisions.

The title does not reflect the content of the manuscript. A title like “A new approach to monitor important grassland habitats in view of the future...” could be more appropriate.”

We have modified the title; we prefer to use workflow instead of approach.  

Please define Nk2 habitats or provide a reference?

Corrected. It was just a typo, we meant N2k (Natura 2000). We already have added the references previously. 

First, the introduction needs to provide general background and context of the study; I found a lack of connection between remote sensing data and species distribution model to support the hypothesis that a combination of both improves species distribution models (lines 59-72). Maybe if the authors provide some references for lines 70-71 could help.

Thanks, we have completely revised the introduction, we have added new references.

Lines 80-83 look more suitable for the methods section. Is target (objective?) 3  cont

Corrected. We shift these sentence in the Material and Methods section. Thanks

The methods section is too large; at some points, it is hard to follow. Some parts of the methods section look like part of the introduction or part of the discussion section. Please revise and stay focus on the methods followed. I think a general flowchart would be greatly beneficial.

We have revised the MM section and added a new figure (flowchart) to make it easier to read. Thanks.

The number of images used should be described within the methods section, not in supplementary data.

Corrected. We added the images used in the MM section and removed the table from the Appendix.

It is not clear how the absence data were collected. Please elaborate.

Corrected. We rephrase the sentence.

Line 142. Remove the symbol “÷.”

Corrected.

Line 267-268. What do you mean by “expert-based process”?

Sorry, it was a misleading sentence. We prefer to remove it. Thanks.

Section 2.9.

This is the core of the paper. Why only investigate topographic drives?

We focused only on topographic drivers for their reliability and for their fine resolution (10 m per pixel). We avoided bioclimatic variables due to the coarse resolution available for the study area, lastly until now there are no other variables available such as soil pH or soil moisture.

This section needs to be carefully revised because the manuscript's title suggests “integration of RS and ecological modelling,” but according to what was written, there is no integration, only extrapolation.

We have changed “integration of RS and ecological modelling,” in the whole MS. Thanks.

All topographic variables are highly correlated; how the authors deal with multicollinearity?

Thanks, we forgot to specify it in MM section. Autocorrelation among variables was tested using Pearson correlation coefficient, excluding variables with Pearson R > 0.75, we have added this sentence in MM. 

Please enhance the quality of all Figures.

Corrected. In this version we provide the images at 400 dpi.

Round 2

Reviewer 3 Report

Many thanks to the authors for taking the time for addressing my comments and revisions. The manuscript has substantially improved. However, there are still issues that need to be addressed.

 

 

Here are my comments:

 

Please enhance the quality of Figure 1 because it is hard (impossible) to read what is inside the boxes.

Also, I believe figure 1 would make perfect sense at the end of the methods section instead of the beginning.

 

According to the comments above, please modify lines 114-118

 

From my previous comment, “The methods section is too large; at some points, it is hard to follow. Some parts of the methods section look like part of the introduction or part of the discussion section. Please revise and stay focus on the methods followed.” It was partially addressed.

There is still information that can be summarized/omitted.

               Lines 144-154. Can be summarized

               Lines 157-164 “S2 launched in June…tree coverings[51].” can be omitted.

               Lines 175-179 can be summarized

               Lines 181-191 can be omitted

               Lines 246-259. For the remote sensing community, I think is not necessary to describe the mentioned machine learning classifiers. It is enough to provide the references for each one of them.

               Lines 282-289 “The aim.. results [72,75,76] can be omitted.

 

In the results section, at least include a table with the accuracy values.

 

 

For some reason, the Figures still have low quality and look blur.

 

Lines 415-417 Is this part of the manuscript??

Author Response

We thank you for your valuable advice. They have greatly improved the manuscript. 

Best regards,

The Authors

 

 

Here are my comments:

Please enhance the quality of Figure 1 because it is hard (impossible) to read what is inside the boxes.

We have increased the character size and as before we have provided the image at 400 dpi.

Also, I believe figure 1 would make perfect sense at the end of the methods section instead of the beginning.

Thanks. We moved Figure 1 to the end of the MM section.  

According to the comments above, please modify lines 114-118

Thanks. We have modified the figure number.

From my previous comment, “The methods section is too large; at some points, it is hard to follow. Some parts of the methods section look like part of the introduction or part of the discussion section. Please revise and stay focus on the methods followed.” It was partially addressed.

There is still information that can be summarized/omitted.

Lines 144-154. Can be summarized

Thanks. We rephrased the sentences.

Lines 157-164 “S2 launched in June…tree coverings [51].” can be omitted.

Thanks. We erased the sentences and providing only the reference for the Sentinel-2 mission.

Lines 175-179 can be summarized

Thanks. We rephrased the sentences.

Lines 181-191 can be omitted

Thanks. We erased the sentences.

Lines 246-259. For the remote sensing community, I think is not necessary to describe the mentioned machine learning classifiers. It is enough to provide the references for each one of them.

Thanks. We erased the sentences providing the references for the machine learning classifiers.

Lines 282-289 “The aim.. results [72,75,76] can be omitted.

Thanks. We erased the sentences.

In the results section, at least include a table with the accuracy values.

Thanks. We added a table with the cross-validation results for all tested models.

For some reason, the Figures still have low quality and look blur.

During the review process, we have provided 400 dpi images. Probably in your version the images have lost quality.

Lines 415-417 Is this part of the manuscript??

Corrected. This is a typo. This sentence was part of the template provided by the RS journal. Thank you for reporting this error.