Multi-Sensor Remote Sensing of Intertidal Flat Habitats for Migratory Shorebird Conservation

Round 1

Reviewer 1 Report

This paper presents a multi-sensor remote sensing technique of intertidal flat habitats for migratory shorebird conservation. In general, the paper is well written and the topic is interesting. However, some major issues need to be addressed before the paper can be found acceptable.

1.      Introduction part, the review on remote sensing methods to map the intertidal zone and to characterize sediment type by using multispectral and radar data, the cited reference papers are all old. I would like to see some new developments at least in the past five years.

2.      The novelty and main contributions of this work should be highlighted, which is not clear at this stage of presentation.

3.      I would suggest: 1) to move out the study area description part from the Introduction; to move out the data set introduction part from the Methods section; 3) combine them two into a new section.

4.      Please explain why the linear spectral mixture model was selected to identify the sediment composition of the intertidal pixels. Will the non-linear modal be more suitable in this case?

5.      Since the Sentinel SAR data was used, why did you also consider the Sentinel multispectral images, rather than the Landsat images as used now. Please clarify this.

6.      The technical details are missing of the considered methods, which should be also improved.

7.      Please discuss the possible errors might be produced during the spectral unmixing and classification results.

Author Response

Authors’ Response:

1. We have extensively rewritten the Introduction and the Discussion to include a number of additional papers, several of those within the past 5 years, including some from Remote Sensing.
2. We have reinforced the novelty and main contribution of this work in the Introduction and Conclusions. We have also included additional analysis of the observed shorebird habitat relationships (Section 3c Results: Habitat Characterization) to underscore the useful conservation management information provided by the habitat mapping approach described.

“Our results show that spectral unmixing analysis of medium resolution Landsat and Sentinel 1 SAR imagery provides an effective means of quantifying this continuous gradient between sand and mud sediment types on coastlines with extensive areas of intertidal habitat. The resulting mapping can also be segmented into qualitative habitat types if that is desired to inform conservation management. Google Earth Engine provides ready access to an immense catalog of imagery as well as the spectral unmixing tools to make this approach feasible for any coastline in the world, provided that field reference data is available for initial training and validation.”

3. We have moved the Study Area description into the Methods section.
4. Examination of the spectral feature space, as well as prior work on the spectral relationship between sediment size and spectral response suggested that a simpler linear model should be sufficient. A non-linear model might prove useful but was not examined. Language to this effect was included.

We employed Landsat OLI imagery rather than Sentinel 2 multispectral imagery as Landsat imagery was readily available prior to when we started planning/conducting our field work in 2015-2016.  With the ready availability of Sentinel 2 imagery, if we did the study now, we would incorporate it as well as other suitable image data sets. I have included language discussing the transferability of the approach to include other similar sensors such as Sentinel-2 in the Conclusions.

5. Have included additional information to the Methods section on the Unmixing Analysis.
6. In the Discussion, we discuss possible errors due to temporal mismatch, our method for determining dominant sediment type, linear vs nonlinear mixing model. We have also beefed up the discussion on complicating factors such as surface and interstitial water content.

Reviewer 2 Report

In this paper, authoers answered two research questions they posed: (1) how well can a combination of multispectral reflectance and SAR backscatter and spectral unmixing techniques as implemented in Google Earth Engine be used to discriminate between mud and sand intertidal types; and, (2) is the relationship between multispectral reflectance, SAR backscatter, and sediment type applicable across two very different geographic locations? The work content in the paper is very rich. But the expression of the whole paper makes me feel more like a work report than an academic paper. My simple suggestions are as follows: 1. Please add the content of the analysis and selection of the analysis methods in this article, 2. For the data part in the appendix, please add the content of data acquisition and analysis process; 3. The references of the full text are relatively old. Remote sensing is an excellent journal. Please read and cite papers published on Remote Sensing related to this article

 

Author Response

Author Response:

1. The Methods section goes through in a quite detailed fashion as to the methods. We have included additional information to the Methods section on the Unmixing Analysis to make that clearer. I have included a Methods Flowchart as an Appendix.
2. Unsure what is being requested here. To include specific information on the Scene IDs of selected imagery?
3. We have extensively rewritten the Introduction and the Discussion to include a number of additional papers, several of those within the past 5 years, including some from Remote Sensing.

 

Reviewer 3 Report

The topic of the manuscript is relevant for the journal. The issue addressed in the manuscript is significant to the concerned community since it explores how mud and sand may be distinguished using various geospatial data types, including multispectral data (Landsat 8) and SAR data (Sentinel-1). The work deserves to be published in its current form in Remote Sensing, since it demonstrates a high level of scientific consistency. The reviewer believes it would be beneficial for the authors to enhance the figures and use histograms instead of tables (in some cases). Other than that, the manuscript is well written and organized.

Author Response

Author Response:

1. Figures 6 & 7 were redone to improve the quality.
2. While we considered histograms, we determined that Tables were more suitable, especially for the accuracy assessment. Tables 4 & 5 could potentially be changed to histograms but it would be harder to directly compare numbers between Inside vs. Outside of the MER.

Reviewer 4 Report

REVIEW SUMMARY

The authors present a study in which they use Landsat 8 (Multispectral optical images) and Sentinel-1 (Radar images) to detect intertidal flat habitats in two study areas (Brasil and Chile).

While I think the manuscript has merits to be published in the remote sensing journal and can have some interest, I have several comments/suggestions and would like to ask for several clarification before I can recommend the publication. 

COMMENTS

[1] The authors motivate the study in line 53 with the importance of understanding the spatial and temporal dynamics of intertidal habitat availability, the research objectives itself do not represent this goal and the methods are not suitable to contribute to the understanding of the spatio-temporal dynamics. I would therefore recommend making the connection betwen the motivation, objectives, and methods clearer.

[2] It is a bit difficult to follow the methods and a figure showing the flowchart would be helpful. 

[3] Section “Intertidal Mask”: It is unclear to me which dataset is used in this step: The previously described one? 

[4] Section "Pixel Analysis": Please describe in more detail the mean pixel value is: Is it spatial or also temporal?

[5] Line 310: “the radar response differs”: This sentence is a bit unclear, I suggest to rephrase it, in particular the difference to what needs to be clear (I assume it’s the difference to the reflectance values). However, since the previous sentence it’s used in a different context it can be clarified.

[6] Please give more details about the transferability to other locations and sensors (e.g. Sentinel-2).

[7] The reference Jensen 2007 to describe the Sentinel-1 bands does not seem appropriate.

[8] Line 241: “were produced with …”, there seems to be something missing in the text, I assume “R statistical language”

[9] The visual quality of the images is in places not sufficient. I suggest increasing the resolution etc.

 [10] Please make proper headlines (with numbers) and put the images in appropriate places within the manuscript. It makes it easier to understand.

Author Response

Author Response:

1. We agree that we are not trying to map temporal change. We have revised to clarify the Objectives as such:

The need to characterize the various types of intertidal habitat, in particular the dominant sediment type as critical baseline information useful for monitoring and informing management decisions concerning the quality and availability of shorebird foraging and roosting habitat.  In this paper, two research questions were posed:  

(1) how well can a combination of multispectral reflectance and SAR backscatter and spectral unmixing techniques as implemented in Google Earth Engine be used to discriminate between mud and sand intertidal types; and, 

(2) is the relationship between multispectral reflectance, SAR backscatter, and sediment type applicable across two very different geographic locations?  

 

2. We have included a Methods Flowchart as an Appendix.
3. Clarified that the Intertidal mask is derived from the mosaicked Landsat imagery.

[4] Rewrote the Methods section describing the extraction/calculation of the mean pixel.

[5] Line 310: “the radar response differs”:

Removed/rephrased this line.

[6] Please give more details about the transferability to other locations and sensors (e.g. Sentinel-2).

I have included language discussing the transferability of the approach to include other similar sensors in the Conclusions.

[7] The reference Jensen 2007 to describe the Sentinel-1 bands does not seem appropriate.

I have removed that reference and included more appropriate references

[8] Line 241: “were produced with …”, there seems to be something missing in the text, I assume “R statistical language”

modified

[9] The visual quality of the images is in places not sufficient. I suggest increasing the resolution etc.

Figures 6 & 7 were redone to improve the quality.

 

 [10] Please make proper headlines (with numbers) and put the images in appropriate places within the manuscript. It makes it easier to understand.

We have moved the Figures and Tables to insert them into the text. Have numbered the sections.

 

Round 2

Reviewer 2 Report

The author did a lot of work but was too concise in his expression.

My 2nd suggestion in 1st round means, in Section 2E，2F，2G, auhtors can add content to explain your dataset, why your choose the method adopted in this paper, etc. 

Author Response

I have further revised the sections as suggested by Reviewer 2 to further clarify the methods. I have included an additional reference related to linear spectral unmixing analysis.

However without further guidance as to specific points of uncertainty, it is unclear to me as to what further might be needed .  

Reviewer 4 Report

The authors have revised the manuscript according to the reviewer's comments

Author Response

No changes requested.