Describing Polyps Behavior of a Deep-Sea Gorgonian, Placogorgia sp., Using a Deep-Learning Approach

Round 1

Reviewer 1 Report

The authors used a deep-learning approach to quantify polyp activity, a proxy for active feeding, in a deep-sea octocoral species in an automated way. Technological advances in the last few years have been allowing the collection of an ever-increasing amount of data in the deep sea. As a result, some datasets are now too large to be analyzed manually, requiring the use of semi-automated methods. This manuscript will thus be of great interest to the science community, especially since it is focusing on the poorly understood feeding biology of deep-sea corals.

However, this manuscript requires major revisions before it can be published. A lot of important information is missing from the Introduction and Discussion sections. Specifically, the authors do not mention any of the published studies on the feeding behavior of deep-sea corals (including gorgonians), omitting a large amount of relevant literature. Similarly, the Discussion section is missing information on the implications of this study for the ecology and conservation of this deep-sea coral species. The Methods and Results sections are often too vague with important information missing, preventing a proper understanding and replication of this study. Moreover, additional analyses could be performed to fully exploit this rich dataset.

On the form, most figures are too small to read and units are often missing. Overall, this manuscript needs to be carefully proof-read to fix the multiple spelling errors, and inconsistent use of tenses throughout the manuscript (sometime present, other times past tense). I corrected some errors/typos but not all, as there were too many.

In summary, I believe that this manuscript could represent a good contribution to the field if the concerns I raised are addressed.

Please find my detailed comments below:

 

Abstract

 

Line 19 – Please define what you mean by deep sea. Usually anything below 200 m depth is considered deep sea.

 

Line 21 – Why do we care about polyp activity? I think it should be mentioned that it is commonly used as a proxy for feeding activity in corals.

Some colonies? How many and where were these colonies (depth, location)?

 

Line 23 – Some people may not be familiar with CTD and ADCP. It would be good to either define these terms or not mention them and say that you looked at the effect of environmental condition and currents on polyp activity.

 

Line 26 – Should be “annotating” instead of “annotated” I think.

 

 

 

Introduction

 

Line 38 – Reference needed.

 

Line 42 – Reference 10 does not seem the most relevant here. There are many studies, including on recovery after heat waves in the Mediterranean, oil spills in the Gulf of Mexico or bottom trawling that could be cited here.

Lines 48 and 49 – References are needed here. Also, not all gorgonians feed on the same type of food. Some species prefer zooplankton while others favor fresh phytodetritus. Corals’ diet can also vary seasonally depending on food availability. I think mentioning this, with supporting references, would give the context needed to discuss the feeding behavior of Placogorgia later in the manuscript.

 

Line 60 – As mentioned previously, the concept of deep sea should be defined.

 

Line 64 – Please define acronyms (here ROV, ROTV).

 

Line 66 – What do you mean by “time-consuming associations”?

 

Lines 76 to 87 – To me, it would make more sense to move this paragraph after the one on coral monitoring (lines 88 to 97). You could talk about coral feeding rhythms and monitoring methods (often video and time-lapse cameras), and then explain that these methods produce large datasets, hence the need for automated methods for analysis.

 

Lines 88 to 97 – This paragraph is incomplete and misleading. It implies that there is no information available on deep-sea gorgonian feeding activity, which is not the case. Similar studies looking at polyp activity in deep-sea corals (some of them using deep-learning approaches) have been published: for instance, Zuazo et al. 2020, Johanson et al. 2017 and Girard et al. 2022 for Paragorgia arborea. Although not a gorgonian, similar studies have focused on D. pertusum: Osterloff et al. 2019 and Buhl-Mortensen et al. 2015. These studies and their main findings should be mentioned.

 

Lines 98 to 101 – This paragraph is very broad and does not add much information to the Introduction. This would be a good opportunity to mention the threats deep-sea corals are facing in more details (especially bottom trawling) and the importance of Marine Protected Areas to prevent impact. You could also mention your study site and the ecological importance of Placogorgia. This would really emphasize the importance of your study.

 

Line 102 - This sentence is grammatically incorrect. I would suggest writing: “Therefore, the aims of this study are to (1) describe etc…”

 

Line 104 – I think you could add environmental conditions and food availability.

 

 

Materials and Methods

 

Lines 110 and 111 – Could you give more details on the MPA? For instance, is fishing completely prohibited or just restricted?

 

Lines 119 to 126 – This should not be in the Methods section and would go much better in the Introduction (see my previous comments).

 

Line 134 – You haven’t mentioned this analysis in the abstract or Introdution. I feel like there should be at least a sentence referring to it. Moreover, the title of this section is a bit misleading. It implies that you characterized the physical habitat, while you are actually looking at colony orientation, size distribution and associated fauna.

 

Lines 143 to 148 – How were the 3D models built? Based on ROV video transects? Which software program was used?

 

Line 150 – Which image dataset are you referring to? How were these images collected? You never mentioned that before.

 

Line 151 – Which visualization software?

 

Lines 155 and 156 – Could you please indicate which type of statistical analysis you used to look for relationships between datasets?

 

Line 159 – Very accurate in terms of positioning?

 

Line 161 – I am assuming you mean that no ropes or buoyancy elements are attached to the lander? Please clarify.

 

Line 167 – “deployment” instead of “deploy”.

 

Line 172 – Were these strobe lights?

 

Lines 174 and 175 – What is the depth range covered by the ADCP? Did you use the entire depth range (averaging all bins)? Or only a limited range. This information is important as currents near the sea floor may be different from currents higher up in the water column. This question is also valid for acoustic backscatter.

 

Line 199 – Which software program did you use for manual annotations?

 

Line 244 – By “activation and deactivation” do you mean polyp extension and retraction. These terms are not common in the literature so this is a bit confusing.

 

Line 265 – Why look at seasonality? Your study only lasted 23 days, this is too short to detect seasonality. If you are interested in periodicity, why not use classical time series analyses such as Fourier periodograms on detrended data? This would be much more informative and likely easier to perform that the analysis you chose. You could use periodograms to look at periodicity in active feeding (colonies with extended polyps) and the different environmental time series and then look for relationships using cross correlation (for instance).

 

Results

 

Line 275 – As I pointed out earlier, I find this title misleading.

 

Lines 278 to 281 – I do not think this should be in the results section (either Methods or Discussion). Right now, it seems that you did this analysis, which is not the case. The result section should only be presenting your results.

 

Lines 281 to 285 – These are not results, this paragraph should be in the Introduction or Discussion.

 

Lines 286 to 296 – Please give some information on colony size here. I know this data is presented in Table 1 but the average size and/or range should be indicated in the text.

 

Line 307 – What do you mean by very diverse? This statement is subjective, is mobile fauna here more diverse than in other coral ecosystems? Could you give numbers (species richness, diversity index such as Shannon)?

 

Line 323 – “deployment” instead of “deployed”.

 

Lines 324 to 326 – Please indicate the average current speed as well as the range.

 

Line 371 – What do you mean by long? Please give numbers? How long is that?

 

Line 377 – What grid size values did you use?

 

Line 380 – Again too vague. Please give a proportion of images instead of using the word “most”.

 

Lines 382 to 387 – These are not really results. This paragraph would go better in the Discussion.

 

Line 427 – “enormous” is a subjective term that does not really mean anything. Please be specific (for instance you could say it is x times larger than other weights).

 

Lines 444 and 445 - -How about CTD data? Temperature can definitely affect coral feeding activity.

 

Line 451 – “specific point in time”? When is that? Please specify.

 

Line 451 to 453 – This should be in the Discussion section.

 

Discussion

Line 459 – This is the first time you use the term octocoral. Why not use gorgonian to be consistent throughout the manuscript?

 

Line 464 – Reference needed.

 

Lines 469 and 470 – “The polyp is basically etc.”: I would remove that sentence. Not needed here and too informal.

 

Line 488 – Reference needed.

 

Lines 511 to 514 – The link between environmental conditions (currents, temperature), food availability and coral feeding is missing here. You have collected these data so why not have a deeper look into this. You have enough data to do time series analyses and identify periodicities and correlations between your time series. Several studies show that changes in current speed and direction linked to tides can influence coral feeding activity. There are many studies looking at the feeding behavior of shallow- and deep-water corals in relation to environmental conditions that you could cite.

 

Line 535 – This is the first time you are mentioning cross-correlation analyses. Have you done these analyses?

 

Overall, there isn’t much of a discussion on the ecological implications of this study, or any comparison with the current state of knowledge on deep-sea coral feeding behavior and cyclic variations in the behavior of other deep-sea species (for instance an influence of seasons and tides have been evidenced in multiple deep-sea ecosystems). There is also no discussion on conservation implications and how this newly acquired knowledge could contribute to MPA management.

 

Lines 549 and 550 – You did not really show that, and did not discuss that relationship (see my previous comments).

 

Figures and tables

 

Figure 1 – The map (a) is small and difficult to read. Could you please indicate the signification of the green and yellow squares in the legend? The title of the green square is not readable on the map. You could also mention in the legend that the arrow indicates the lander deployment site.

 

Figure 2 – Do you have a broad estimate of the surface area visible in the field of view?

In the legend, by “situation” do you mean “positioning”?

 

Figure 3 – This figure is really difficult to understand for someone not familiar with deep-learning algorithms. Could you add a description of the different components of this figure in the legend? Or the text?

 

Figure 4c – The legend is too small and cannot be read.

 

Figure 5 – This figure should be larger. Moreover, all units are missing.

These time series display clear periodicities. As I mentioned earlier I think that periodograms would be really valuable.

 

Figure 7 – The font on this figure is too small. Do color gradients represent the percentage of open colonies? If so this should be specified.

It is pretty difficult to visualize differences between sections/colonies with this graph. Surely there is a better way to represent these data (for instance show the average percentage per day for each section/colony, this would also allow you to statistically test differences).

 

Figure 8 – This is too small, I cannot read anything. I do not understand the legend either, what do you mean by “time of the <90% of open polyp”?

 

Figure 9 – You could add statistical tests to complement this figure. You could easily test for differences in the proportions of open vs closed polyps using Chi-square or Fisher’s exact test.

 

Figure 11 – Units are missing. The y-axis titles should be “Percentage of open polyps” not “polips” (including spelling error).

 

Table 3 – Three classes? I only see two presented in this table.

 

Table 4 – Units are missing. Are these hours? Minutes per day?

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments are attached.

Comments for author File: Comments.pdf

Author Response

Please see the attachment

Author Response File: Author Response.docx

Reviewer 3 Report

The topic of this paper is on the deep-sea gorgonian feeding activities and other behaviors using machine learning method. In a word, it is well written. But some important details are missing, thus, I recommend to accept it after the following issues are well addressed:

1.       Some abbreviations should be accompanied with their expansions or definitions when firstly appearing in the paper such as CTD, ADCP…  Please check.

2.       What is the relation between Gorgonians and polyps? Sometimes polyps feeding behavior are studied, sometimes the feeding behavior of this species of gorgonian is mentioned in the paper. What are the difference of two feeding behaviors? Please clarify. Also, in conclusions, the polyps behaviors are never mentioned which does not distract the focus of the title.

3.       Actually, underwater sensing approach is the most important part in this paper but there is only one paragraph (line 76-86) describing it in the introduction. It is very necessary to add more relevant literatures to introduce more various sensing approaching such as acoustic based, magnetic based, etc.

4.       In the papers, no equations or formulas are found. It is necessary to add some mathematical formulations for the theory or model such as 2.4.3, 2.5, etc.

5.       In Figure 6, the results of segmentation using the proposed method actually do not match the original image well. Please clarify.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Thank you for addressing my concerns.

I am mostly satisfied with the changes made. However, I noticed that my comments on Figures have not been addressed yet. In particular, units are still missing and fonts too small. I am assuming these will be addressed at a later stage?

Author Response

Thank you very much for the exhaustive review of this manuscript. 
We believe that the changes applied according to your comments have improved the quality of it.
The comments about Figures have been taken into account in this new version, sending separate figures for greater clarity with metric units included and larger font sizes.

Reviewer 3 Report

Thanks for the authors' efforts. My comments are well addressed thus I recommend an accept decision.

Author Response

Thank you very much for the exhaustive review of this manuscript. 
We believe that the changes applied according to your comments have improved the quality of it.