Rapid Large-Scale Wetland Inventory Update Using Multi-Source Remote Sensing

Round 1

Reviewer 1 Report (Previous Reviewer 3)

The last manuscript with the indicated revisions and the authors'  point-by-point responses responds to the doubts and gaps observed in the first version of the article.

Author Response

Thank you for taking the time to review our paper. We appreciate your efforts

Reviewer 2 Report (New Reviewer)

1. Table 5 compares the confusion matrices for performance of baseline approach and the proposed method. What are the specific steps of the baseline approach? Where is it different from the proposed method? Is the training sample data different? Is there a difference in the methodology above?

2. The description of the method in the text is mostly step-by-step diagrams and lacks the listing of specific formulas, such as how to calculate DSWE? only a reference is given.

3. Figure 10 vertical coordinate unit is not uniform

4. Manuscript pdf many highlighted parts, maybe submitted without completion?

5. There are many input parameters (both active and passive remote sensing images) involved in random forest training, which are the main factors affecting the results? Which are secondary? It is recommended to have quantitative analysis and discussion.

6. Is the overall 2% improvement in accuracy of the method over the baseline method already a significant improvement? Are comparisons with the accuracy of other relevant already established wetland classification products considered?

7. In the results section, there are fewer graphs for the classification results, and it is recommended to add graphs for the results of different types of wetland classification.

8. In the discussion section, has the change characterization of wetlands with multiple time series been considered?

Author Response

Thank you for taking the time to review our paper, we appreciate your time and efforts. Please find attached our edits and revisions in response to your comments.

Author Response File: Author Response.docx

Reviewer 3 Report (New Reviewer)

Dear authors,

Thank you for your exploration of how we can better understand updating wetland inventories using remote sensing data. Here are some suggestions that should improve the quality of your manuscript.

Author Response

Thank you for taking the time to review our paper, we appreciate your time and efforts. Please find attached our edits and revisions in response to your comments.

Author Response File: Author Response.docx

Reviewer 4 Report (New Reviewer)

Major comments

Abstract 

1. Summary of the contribution of this work to literature of wetland mapping needed 

Introduction 

1. Line 48-49, transition from wetland’s role in global carbon cycle to the state of Minnesota’s lack of wetland inventory is abrupt. Please clarify.

2. A definition of what “wetland inventory” is needed in the introduction. In Line 60-70, you described various RS techniques for mapping different aspects of wetlands. But what you are really talking about is wetland classification and extent mapping. My question is, is wetland classification the same as wetland inventory? Because in forestry, forest inventory is different from forest extent mapping. In the context of wetland, I suggest the author’s explicitly define what they mean by wetland inventory, and how it is related or differs from wetland classification. 

3. Line 86, the authors mentioned “field data” collection in challenging and costly, what type of information is typically collected in the field for wetland inventory, how it is challenging, and how RS techniques help is not really well elaborated.

4. Line 85-111, it’s a lengthy classification method review. However,how all of these different methods are relevant for wetland inventory is not well synthesized. I suggest shortening this paragraph and only discuss the relevant ones to the wetland inventory mapping.

5. Line 119, similar to the comment for abstract, why the study is relevant to a large body of literature needs to be identified here. 

Data

Section 3.1 

1. Figure 1a does not seem to add much information, suggest to remove

2. Clear definition of wetland needed, preferably in the introduction section 

3. The description of the National wetland inventory thematic map is lengthy, and much of the information does not seem to be relevant to this study. I suggest clearly identifying how the authors use this existing project for their wetland mapping, instead of describing lots of information about this existing product’s accuracy and production process. 

Method 

1. Line399 - 400, do you mean stratified random sampling sampling? 

2. For section 4.1.1, the Dynamic surface water extent (DSWE) first got introduced here, it seems like its quality wasn’t discussed prior?

3. One major question I have about section 4.1.1 is, the authors didn’t clearly explain how is it suitable to train on aggregated indices at 100m-solution and make predictions of SWF at 10m pixels. I understand the rationale is to use the proportion of wet pixels, but how are these proportions trained using RF to make predictions at 10m?

4. Section 4.3.1, for this section on validation, only generic information on accuracy score calculation is described, but most importantly how the classified pixels were deemed correctly classified is missing.  Are the map producers validating the map? Or did you have an independent expert validate the produced map? What is considered a corrected classified pixel? At what resolution, 30m or 100m? Or did you validate at object level? How do you validate wetland change? These are more important information related to this wetland mapping study, then the generic accuracy score calculations.

Results

1. What is the spatial resolution of your final product?

2. Can you provide an example illustrating an area where the new method performed better than the previous method to demonstrate novelty?

3. Why are sentinel-1 SAR indices added? What are the values added to your mapping? Did the previous method include SAR-derived metrics as well? 

Discussion 

1. Section 6.3, though the section is labeled “scalability”, the authors are really only talking about scalability on GEE, rather than the method as a whole and transferability to other wetland environments. I think talking briefly about GEE’s computational limits is sufficient, but what would be really helpful to other researchers is to identify the potentials and challenges of applying this work to different wetlands?

Minor comments

1. Line 12-15 grammar 

2. Line 28-29 grammar

3. Line 42, modify “characteristic heterogeneity”

4. Line 253-254 should be “were generated”

Please check the manuscript for grammatical mistakes and clarity of word choice. I made a few suggestions in the minor comment sections, but I suggest doing another round of thorough check for other mistakes.

Author Response

Thank you for taking the time to review our paper, we appreciate your time and efforts. Please find attached our edits and revisions in response to your comments.

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

This is a very well designed study and a well-written manuscript. I only had one comment in the attached version. 

To me, the particular merits of this manuscript are that the reasons for the study are clearly stated, the study area is well described, the methods are detailed enough that the study could be duplicated, the results are analyzed properly, the discussion and conclusions logically follow the results.

I think this manuscript could be published as is.

Thanks for doing this work and best of luck to you all going forward. 

Comments for author File: Comments.pdf

Reviewer 2 Report

Overall this paper is fine and just requires some tidying up in places. Please ensure that all abbreviations are spelled out in full and avoid using abbreviations where possible as it is confusing for the reader. The introduction, methods and results are ok, they are pretty usual. The discussion needs some more work, it does not say very much about the interpretation of the results, for either the method or rapid monitoring of wetland dynamics. The figures need some work in places, please note that Fig 5 is incorrect and must be done again. The references are ok. I have made some comments on a scanned pdf of the paper which you should download. Please make these corrections as marked. 

Comments for author File: Comments.pdf

Reviewer 3 Report

Comments for author File: Comments.pdf

Reviewer 4 Report

This manuscript provides a rapid method to mapping wetland dynamic by using wetland maps and combine Sentinel-2 and Sentinel-1 data to detect changes. LandTrendr algorithm was used to analyzed surface water and vegetation changes during high water time. The method was practical and reliable for the wetland management since the updater of Minnesota wetland map stopped. Generally, the manuscript was well organized.

But the abstract result and discussion should not only include conclusion about also provide some information how the wetlands in Minnesota changed during 2016-2022?

In Line 23, please provide the mean of abbreviation “F1”.

Reviewer 5 Report

the paper presents a method for the mapping of wetlands with a case study from Minnesota USa. the paper is interesting in particular in connection with the monitoring of biodiversity. the text is relatively clear and appropriate but I find some paragraphs that are noty perfectly clear of accurate. the description of the study site can be improved. Please find the details of the comments in the attached pdf

Comments for author File: Comments.pdf