Investigating Detection of Floating Plastic Litter from Space Using Sentinel-2 Imagery
Round 1
Reviewer 1 Report
Investigating Detection of Floating Plastic Litter from Space Using Sentinel-2
I thank the authors to work on this interesting project. However, the English needs to be improved in this manuscript. I have some major concerns about the manuscript, and they are listed, as follow:
Title
- I would like to suggest the Authors to think about expanding the title after Sentinel-2 (e.g., satellite imagery, imagery)
Introduction
- in order to strengthen the Introduction, I suggest that you specify exactly what has been done in Lines 93-94 - „several research studies have used various remote sensing methods“
- reference [81] mentions the University of the Aegan, but I suppose that the authors of the aforementioned research work at that University
- I suggest that the last paragraph of this Section starts with: „The objective in this study..“ and Lines 109-110 should be moved to the previous paragraph with description of what has been done
Materials and Methods
- starting from Line 118 and further in manuscript, for satellite imagery, I think that it is enough to say the spatial resolution of 10m (redundant is 10m x 10m)
- In my personal opinion, one of the Figures (1 or 2) is redundant
- Line 159, can You tell us (one or two sentence) something more about the spectroradiometer
- Table 1: Band08A is missing, since there are 12 bands in the Table 1, and S2 has 13 bands (also, B9 -> B09)
Results
- Lines 241, 242, 243: it is not necessary to underline Band B06/07.. B06/B07 is enough
- Lines 246 – 248: this text description should be replaced with the legend in the picture since the description is very hard to read and understand the meaning of lines/objects, also x-lab Title covers the Wavelength values in Figure 7
- Lines 252, 255-256: now the Authors mention B8/ Blue band, etc, the expressions about the spectral bands need to be consistent throughout the whole manuscript
- Line 269-270: I would not agree with that assumption, since 30 m2 are not so similar to 100m2, it is the best what You can get using S2 imagery
- Lines 274-282: Description about the sensitivity analysis and explanation about the determination of the discriminative value should be moved to the end of the Section Material and Methods
- Lines 289-290: is the sentence „In Figure 10 (right) the scatterplot features the values of plastics and water using all nine indices.“ consistent with the caption of the Figure 10 (right)?
- I did not quite understand the connection between (if there is any) Table 2 – Plastic pixels detected and Figure 12. False positives are circled in blue, and the target in dashed yellow – but there are different sizes of the pixels/clusters.. So does circled area represent plastic pixel (1 yellow and 6 blue circles vs 7 plastic pixel detected) because the yellow areas within the circle vary in size
Discussion
- Line 312: littler?
- the whole Discussion section need to be strengthen. Lines 316-319 indicate that in this study the Plastic index is developed, and there is not a word about the Reversed Normalized Difference Vegetation Index (RNDVI), which the authors mention in the Abstract (Line 23) and in Material and Methods (Line 200-202). RNDVI was not analyzed in the Disscusion section, and even in Results the whole focus was on the PI index
Conclusions
- again, in the Abstract the authors mention that they will propose two new indices, and only PI index is mentioned
- Line 348: this research was made using multispectral imagery, so I think that the part „can only be used during the day“ is really not necessary
- Line 359-360: please explain this sentence. Will the indices be compared to SAR imagery or in which way they can be adjusted?
- Main concern of the Conclusion part is using a references. The Conclusion section is not the place to present new facts (should be in the body of the manuscript), so conclusions don't usually have references unless you come up with a 'punchy' quote from someone special as a final word.
Author Response
Please see attachment
Author Response File: Author Response.pdf
Reviewer 2 Report
Please see the attached document
Comments for author File: Comments.pdf
Author Response
Please see attached
Author Response File: Author Response.pdf
Reviewer 3 Report
The manuscript shows a detection method of floating plastic litter using Sentinel-2 data. The introduction shows well the problem and the goals of this study. But the methodology section needs more effort to show how the Sentinel-2 data could achieve the main goal of detection plastic.
Major considerations:
In terms of satellite data uses, there are a lot of indexes (hundreds) that could be tested in this study, even could be interesting to search for another specific one studying the relation among bands for this purpose. Test the NDVI in this area could be criticized because Normalized difference vegetation index has been developed to agronomy/forestry studies. The amount of false positive means that the index could be improved.
What is the minimum size of the plastic surface detected by the selected index?
To compare different satellite data source could improve the manuscript. Especially, when the authors conclude that "The methodology developed in this manuscript can be adapted for use with other satellite images and with other imaging devices, similar to other research that has been done on identifying plastics in the water [73–79,82–85]". To conclude this, previously must be in discussion and tested in methodology, but is not recommended to conclude this one if it hasn't been studied in your research.
Minor considerations:
The color in figure 9 could be changed to show better the results, for example, painting the plastic that was detected in another color (red for example). The same for figure 11.
Why use L1C level and not L2A? Moreover, one advantage of using Sentinel-2 is the frequency of sensing data. Is possible to test the results in more that one day to know if the detection changes?
The sentence in conclusion: "The findings of this study are significant, since, as indicated by various researchers, the South-East Mediterranean faces a significant problem with plastic debris." is not adequate because that the significance of the problem not makes the results significant.
Author Response
Please see attached
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
The authors have improved the manuscript from v01 to v02. However, in order to publish this manuscript in this journal, some parts of the manuscript still need to be improved:
- Table 1 - after B08, Band 31 is added?
- Figure 7 is still unclear, and legend for aforementioned Figure is a necessity, along with correction of the x-lab
- Line 248 and 250: word Band is not needed, and also Line 249: 20 m spatial resolution is enough
Although most of the comments / suggestions were adopted and corrected, some parts need to be harmonized in order to present a fluent workflow of the research.
Author Response
Thank you very much for your comments and contribution. They were very helpful. As requested, all your changes were incorporated into the manuscript.
Tthe following changes were made:
Table 1 - after B08, Band 31 is added? - CORRECTED
Figure 7 is still unclear, and legend for aforementioned Figure is a necessity, along with correction of the x-lab - CORRECTED
Line 248 and 250: word Band is not needed, and also Line 249: 20 m spatial resolution is enough - CORRECTED
Reviewer 3 Report
I was waiting for more discussion in response to my comments. However, there were changes in the manuscript that improves and correct it. In my opinion is enough to be published, because is novel in this area, but the methods and discussion could be notably improved in the next research.
Author Response
Thank you very much for your comments and contribution. They were very helpful.
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 research utilizes the Sentinel-2 (S2) L1C imagery for identifying plastic litter on the sea surface. The study evaluates various spectral indices in order to detect plastic litter. At the end, the authors propose two new indices for detecting plastic litter using satellite imagery.
The major concerns are the novelty and the contribution to remote sensing field. Used target has a dimensions of 3m x 10m and pixel spacing of S2 imagery is 10m x 10m. In conclusions, the authors mention that the „methodology in this paper can be used with other satellite images..“, my question would be why has it not been confirmed with e.g., Planet imagery (Planetscope, RapidEye) with a spatial resolution of approx. 3m and 5m? Aforementioned sensor also acquire imagery in NIR bands.
Abbreviations needs to be adjusted througout the research. Whilst in the Section 1 the authors mention UAVs, further from Section 2 the phrase 'drones' is used. Table 1 should be included in Section 2, as well as Equations (1) – (9).
Section 3 (Results) lacks of additional quantitative measures. The authors mention that „The aerial images from the drone were compared to the images from the Sentinel-2 satellite.“, did it reffer only for a visual examination? I suggest atmospheric correction to Bottom-of-Atmosphere (BOA) reflectance product (Level-2A) using e.g. Sen2Cor processor.
Overall, the research design is appropriate, but in order to be published in this journal, some additional research has to be made (e.g., comparison of spectral data gathered durig research with other spaceborne sensors, additional verification of newly proposed indices).
Author Response
Please find attached our responses to your comments in the attachment. We thank you for your assistance.
Author Response File: Author Response.pdf
Reviewer 2 Report
Please see the attached PDF file.
Comments for author File: Comments.pdf
Author Response
Please find attached our responses to your comments in the attachment. We thank you for your assistance.
Author Response File: Author Response.pdf
Reviewer 3 Report
Overall the experiment and measurements are sound. However, there are some serious gaps in the description, figures, and analysis, and perhaps some ambiguity in interpretation of some of the measurements.
Major:
- Many of the figures show images in black/white or false color to indicate the intensity of reflected radiation or some "index", but it is necessary to include color bars with values and units in order to correctly interpret the images. These are figures 5, 6, 7, 10, 11, and 12.
- Line 184: The FLIR derived temperature is based on some assumed emissivity. What was used for seawater? Is the emissivity of plastic the same as for sea water? If not, then the "apparent" temperatures will be different even if the true temperatures are identical. Also, for Figure 7, the seawater image shows a gradient in temperature, as well as a sharp edge. Are these real or do they have to do with emissivity changes? Why is the value taken from the center of the image? Finally, on line 249 "the temperature of the plastic bottles in water is higher than that of the water alone" cannot be supported unless the emissivities are known.
- Caption for Figure 8 makes no sense. What do the different colors for the curves represent? Do the B1, B2, etc. denote sentinel-2 channels? Do the shaded regions represent the bandwidths of each channel? Later, it appears from Table 1 that the letters do correspond to spectral bands of Sentinel-2. This should be explained in the caption or referred to Table 1. In Table 1, "Resolution" should be better defined. Is this a spatial resolution, and if so, is it square or circular and how is it quantified and related to pixel size?
- Line 220: All of the terms in eqs 1-9 require definition in terms of the bands B1-B8 that are shown in Figure 8 and in Table 1. That way, relevant wavelengths can be understood.
- Line 236: It is stated that the optimal index for identifying plastic was the PI (eq 8) but it is not described how "optimal" was determined quantitatively. Did PI have the largest percent difference between the value of the target compared to that of the surrounding seawater? Or is PI less noisy than the others? Or did it contain fewer false detections? Similarly on line 270, the words "most effective" are used, but the effectiveness needs to be quantified in order to declare PI the most effective index.
- Line 248: "Plastic bottles were used to mimic a large plastic target that would be evident on a satellite image." How common are these large targets compared to smaller ones? How much smaller could the target be and still be detected by the satellite instrument? What fraction of the total mass of plastic is below the detection size threshold? On line 255 the paper states that the intent was to mimic a plastic garbage cluster, but this needs to be described better. Why do clusters form? Where do they form most often? What is their size distribution, and is the mean about area of these clusters 30 m^2 just like the simulated target?
- It is not stated very clearly, but because this method relies on reflected solar radiation, it seems that it may only work during the day and in cloud-free conditions. This is not a major drawback, but one that still should be clarified in the paper, such that the potential for identifying targets could be limited.
- Line 265: The paper says that bands 6,7, and 8 can be used to identify plastic, but the PI as defined by eq 8 appears to need only two bands, NIR and Red. However as stated above, there is no clear 1:1 correspondence between the equations and Table 1, so this reader is somewhat confused. In addition, on lines 268-269 it states that bands 4 and 8 are used for the PI, but there is no justification for why band 4 is needed, because as stated above, bands 6-8 show the largest difference in reflectance between water and plastic. One suspects that this may be a normalization technique, but it is not discussed in the paper.
Minor:
- Line 145: should be more quantitative by providing FWHM specifications
- Lines 175, 178: "Little" should be "Litter"
- Figure 5 caption, which wavelength is shown, 660 or 850?
- Figure 9: It would be useful to indicate with an arrow or a marker where we should expect to see the target.
- Caption for Figure 10: Please explain what the numbers in parentheses represent.
- Figure 12: What are all of the other small targets in this figure? Or is that all just noise?
- Line 256: grammar
- Line 264: More precisely, "energy" should be "solar radiation"
- Line 278: "...as well as the high value in the infrared wavelengths." High value of what quantity? Is this meant to say "higher reflectance of solar radiation at NIR wavelengths" ?
Author Response
Please find attached our responses to your comments in the attachment. We thank you for your assistance.
Reviewer 4 Report
The manuscript shows an application of Sentinel 2 in the detection of plastic litter. The introduction is well writing and shows the problem and the goals of this paper simple and easy to read. However, the methodology needs to be improved. First, the goal is to use Sentinel data to detect plastic litter but it is necessary to explain the advantages and disadvantages of this data source.
This article does not show a proper methodology section only shows an experiment to test a method that is not explained. This experiment could be criticized because it was created to be in the best conditions to be detected by Sentinel 2. But what happens in the other conditions? For example, scattered plastic garbage (low density)? Obviously, something that has enough dimensions could be detected by Sentinel 2, overall if this thing has a different reflection. But a bottle, for example, not have enough dimension to be detectable by 10x10 (in the best cases) resolution of Sentinel. To be detectable is needed a lot of bottles together and floating. So, after reading this manuscript, I have enormous doubts about if really the Sentinel 2 data could be used to achieve the goal proposed, and the results and discussion are not enough to get objective conclusions about the effectiveness of identifying litter plastic.
In these terms, the effort that needs this paper to improve it is more than a major revision requires, so my decision is to reject it. But I encourage authors to develop and improve a methodology that really shows that the Sentinel 2 data is useful to identify litter plastic in the sea, and resubmitted.
Author Response
Please find attached our responses to your comments in the attachment. We thank you for your assistance.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
The authors provide minor answers and updates from v1 to v2, however some major limitations still exist regarding the manuscript.
A lot of answers to the reviewer's indicate that some informations are beyond the scope of this study or will be investigated in the future research. Still, paper without strong Discussion (without single reference) cannot be accepted for publishing in this journal.
The Discussion section lacks of explanation if your findings agree with what other authors have shown in existing literature (as described in Introduction). Perhaps similar investigations have not used Sentinel-2 imagery, but still similar sensor was used.
Figure 8 and Figure 13 from Koumi et al. (2019) (URL: https://www.watertank.se/wp-content/uploads/2019/06/cyprus-sjwp-2019.pdf) are similar to the Figure 2 and Figure 7 in Your manuscript. When you use a figure in your paper that has been adapted or copied directly from another source, you need to reference the original source.
In the 2b) answer to the Reviewer 2, the authors have responded that "In Figure 10, statistical analysis was performed for each index to identify the range at which the plastic litter target was visible."
I do not see any statistical analysis in Figure 10, and if the analysis has additionally been done, I suggest to show it in Table with basic descriptive statistics. I also suppose that blue circles in Figure 11 should round up the yellow targets, which is not the case in the present form.
Reviewer 2 Report
Please, see attached document
Comments for author File: Comments.pdf
Reviewer 3 Report
Please see attached pdf file
Comments for author File: Comments.pdf