Using Tourist Diver Photos to Assess the Effects of Marine Heatwaves on Central Red Sea Coral Reefs

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The manuscript deals with the combination of AI tools with citizen science (massive imagery taken from non-scientist divers) with the overall aim of proposing, and applying, a method for contributing to coral reefs monitoring with reference to the effects of marine heatwaves. The idea, as interpreted by the title and the abstract, is interesting and could provide progress in the topic. Despite that, the present form of the manuscript is not yet ready to give the reader what is expected from the title and the abstract. The manuscript has, formally, the scientific structure but it does not explain the fundamental details that are needed for the reader to replicate and apply the method. I do not even understand if the manuscript is intended to focus on ecology results (then a deeper analysis and discussion of the impact for coral colonies could be expanded) or on the method (then a more detailed technical description of the approach is required to allow reproducing it). Moreover, reading is difficult because the results section is weighted down by continuous interruptions and references to statistical tests (that are useful but need to be presented in a different way): a general overview of the paper highlights that there are various datasets all individually analyzed heavily from a statistical point of view and there is, in the present form, not such a significant contribution to scientific progress. In my opinion, the work should focus on the method, adding technical details, showing main results (detailed with graphs and images but commented on the results section in aggregate form addressing) then addressing a deep scientific discussion on the method (potentially with ‘validation’ against a gold standard). The results should try to demonstrate/validate the reliability and the quality of the approach (against a traditional approach if possible) and with this purpose the manuscript requires a major improvement. I suggest Authors to clearly define the main goal of the manuscript (and declare it at the beginning) and then to rewrite the text on this aim; the novelty as well as the contribution to the scientific progress need also to be clearly put in evidence. Results section needs to summarize results and deeply discuss them (with reference to the main goal of the manuscript).

Some additional comments are also listed below:

- line 46: here the Author is basing the origin and the motivation on the principle that, the described method has been tested against a gold standard, and it is working; thus, it is now applied because the validation is given in [11]. Well, reference [11] is not yet published, and it is not eligible to be included in the reference list. For the same reason, it is not available to verify the method. Probably, the Author should wait for the publication/verification of the method and later go on with the application (following the scientific approach). Moreover, the title of [11] leads to imagining a similar content with reference to the present manuscript.

- section 2.1: It is not clear whether the images were commissioned to non-expert divers/tourists with some minimal instructions (where and when to take the images, shooting from the surface downwards or randomly, inserting the white balance system into the scene) or whether they were acquired by the Author.

- lines 88 and 89: the information of the camera (always the same two models) suggests that the images are taken from the same team. This is apparently in contrast with the title and with the idea of citizen science. Is this just an attempt to simulate photos taken by tourists? Please clarify.

- lines 88-89: please provide details about the in situ white balancing

- line 91: based on table 2, 7000 is the total number of processed images while 750 is the number of images used for AI training. Please correct or specify.

- caption of table 1: naming periods of time with Q# is redundant if these names are not included in the table. The Q# codes are used sometimes in the text and, if useful to do in this way, their definition should be reported in the text instead of in a caption.

- Figure 2: the caption is misleading: reading the first part one understands that the image is an example of annotation (thus the image is used for training and is manually annotated) but later (lines 133,134) one understands that the image is the result of classification….please clarify. Moreover, I do not understand the usefulness of reporting the labelset (with colors) in the bottom of the figure; should the colors be found in the + signs in order to highlight the correspondence between the annotation sign and the lableset?

- line 145: “I classified bleaching in white-balanced images only by subjectively comparing to CoralWatch cards”. How subjective is this approach? Give details because it sounds beyond the scientific method.

- lines 157-158: “it will therefore be important to assess the degree of congruency between this unstructured approach, and a gold standard one (sensu [16]), in future works.”. This verification should be given now (as mentioned in the general comment above), otherwise the manuscript is not assessable.

- caption of figure 3 (lines 202-203): “he blue line and red area chart correspond to the left and right 202 y-axes, respectively”. Eliminate the sentence and add colors on the axes (as in the plot of panel a).

- lines 206-208: the comparison between the in situ measured temperature and NOAA temperature should demonstrate the accuracy of NOAA SST so that later it could be used in the citizen-science based approach to discuss results of the image analysis. At the moment this is not clearly understood in the manuscript. If such temperature analysis is relevant then it should be deeper described and Figure S1a added in the manuscript (referring to supplementary material is non easy to manage in reading); otherwise if it is not adding significant information then it should be removed.

- lines 233-234: Were these calculations done for each epoch using the same net initially trained or was the training updated with new images at each epoch? Please explain.

- line 237: please remove the sentence “(though see caveats below.)”

- line 252: “from 25 to 6%”. Is the Author comparing the same things? Are these calculations based on equally reliable surveys? The reliability of the method should be discussed: how reliable is this approach is not discussed (how reliable is the resulting 6%?). It should be compared with a survey done in traditional mode at least in two different eras.

- Figure 4: the localization of dots in panels e and j is not corresponding, please verify. Panels d and I are apparently in contrast (particularly in Thuwal): if healthy coral cover is strongly reduced I would expect an increase of the coral bleaching. I probably didn't understand well but this is a sign of something poorly explained or unclear.

- Figure 5: in general the images are not very clear because they are very full, the double scales on the axes must be immediately recognizable (even without reading the caption). For example, use different colors for the axes depending on the colors used in the graph. I suggest less writing on the graphs. I point out different fonts for the sub-parts of a figure in the various figures.

- Table 4: I do not understand the standard deviation and the final mean standard deviation computation. Please comment.

- section 3.7 – line 391: modify as a new section (not subsection) named Conclusions.

- lines 410-412: a more comprehensive and critical discussion/conclusion about the citizen science approach in this case should be given. A sort of guidelines, checklist of operations to share instructions to non-scientist to profitably participate in sharing images could more successfully help spreading to non-scientists divers the approach and improving the final reliability of the monitoring.

- figure 7 is not adding value to conclusions. It could be coupled with figure 1.

Author Response

Thank you for taking the time to provide constructive comments on my article, all of which surely having improved it. Please see attached document. Briefly, I believe I have addressed all of the suggestions except for two. The first being the under review article which is, as of June 30th, still under review. I do hope it will be accepted soon. Secondly, I was unable to do a side-by-side comparison with another method because I was the only person conducting bleaching surveys at that time. In fact, that is the major selling point of the article; scientists were not expecting bleaching at those times, and so I was able to gather data to fill a key knowledge/data gap. 

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The paper is well written and interesting.  Even though collecting images from the general public to track coral bleaching is not unheard of, the use of simple snapshots of reefs and AI (CoralNet) are indeed a promising avenue to increase the capacity to continue monitoring coral reefs.

I only have a couple of very minor comments.

L217. Typo, "era" should be "error".

Figure 4 caption. Is "healthy coral cover" calculated based on the total annotation points classified as benthos or those classified as scleractinian coral. This distinction is clear in the caption for "coral bleaching", but not for "healthy coral cover".  Throughout the manuscript, it is logical and intuitive to think that "total coral cover" is percentage of coral per substratum, and "coral bleaching" is bleaching coral out of total coal cover, but I feel like healthy coral cover can be calculated either way...  It will be nice if the author could go through the manuscript and clarify them as I had to wonder multiple times while reading the manuscript.

Author Response

Please see attached document. 

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

I would like to thank the Author for going through my detailed review and for modifying the manuscript accordingly. I strongly appreciate the effort, and I guess that the final version has been significantly improved now. I still have some minor remarks, final but fundamental:

- Using the term ‘validation’ in this manuscript is not appropriate with reference to the approach, as shared by the Author himself in the response to my comments (validation requires a method to be tested and verified in comparison with another approach, being already recognized and more accurate; this is not possible in this case and it is not the goal of the present manuscript). Thus, I suggest to modify the word ‘validating’ in line 59 with ‘fostering’ (or something more appropriate than ‘validating’ which is not the case). Moreover, the word ‘validate’ has been added in line 253 and I ask to modify as well, in this case with ‘apply’ (or other alternatives, if considered more appropriate).

- old Figure 4 (now Figure 5): with reference to my previous comment to old Figure 4, I remark that I originally understood the legends (and I agree with the choice). I still do not understand why in sprig 2024 (figures d and i) I see Thuwal in red with a strong decrease in healthy coral cover and I do not have, at the same time, a red dot (increase) in coral bleaching. Why does figure i report Thuwal with a green dot?

- Finally, I strongly suggest to wait for the publication of reference [11]; this would help the success of both papers (the reader is lacking important knowledge when reading the present paper)

Author Response

Please see attached document. 

Author Response File: Author Response.pdf