This article describes the construction and use of fish-specific autonomous reef monitoring structures (FARMS). These FARMS were deployed in a variety of habitats and conditions to quantify the cryptobenthic fish communities in a standardized manner. FARMS successfully collected a variety of different cryptobenthic fishes in the habitats in which they were deployed.

Overall, I thought the logic and justification behind the development of these FARMS was good. In addition, I thought the design was technically sound, and the construction details were excellent with the video in particular ensuring readers could easily reproduce these units. I also wanted to state that I have no issue with the use of these artificial sampling structures as a standardized sampling technique to describe cryptobenthic fish communities, especially in areas where enclosed sampling with fish anesthetics is not feasible. Indeed, I thought the manuscript did a good job of highlighting that enclosed fish sampling with anesthetics is still the ‘gold standard’ for quantifying cryptobenthic fish communities on coral reefs, and that FARMS are not aimed at replacing such sampling. Rather, it appears FARMS have been pitched as a complementary tool that may facilitate the consideration of cryptobenthic fishes in research and monitoring programs more widely. An endeavour I believe is certainly valuable given the importance of cryptobenthic fishes in marine ecosystems. However, I do have some major comments regarding the composition of the current manuscript and the sampling design that need some careful consideration.

I note the editor’s suggestion on the original submission was to revise the manuscript into a ‘Practical Tools’ format that centered around the tool itself. I certainly agree with this suggestion, and it appears the manuscript has been revised substantially from the original version. However, the current version could still be improved considerably to help center it more on the tool itself. I found myself confused at several points in the manuscript because there did not appear to be any clear purpose to all of the data collected from FARMS sampling, and I found myself wondering how the data currently presented in the manuscript actually gave support to FARMS being a potentially valuable tool. However, I do have several suggestions which I believe would help improve the manuscript in this respect. I hope this rough ‘road map’ that I have outlined below will be useful in revising the manuscript to ensure it is focused on the tool itself.

Firstly, I suggest removing the ‘global scale’ analysis (i.e., results shown in Figure 3) entirely from the paper. My reasoning for this suggestion is that this ‘global scale’ analysis is confounded by a variety of factors (including deployment time, FARMS design, habitat, depth variation, spatial sampling designs) that undermine any results from the comparison and, therefore, does little to justify the use of FARMS beyond showing that they can collect fish. Even from a statistical point of view it is hard to justify this analysis considering that these confounding factors mean that independence is violated in several ways, and these various sources of dependence are not accounted for in the statistical models. In this respect, I think any use of this data to try and describe ecological patterns should be abandoned because FARMS do not appear to have been deployed in any sort of standardized manner across or within geographic locations. For example, the deployment for FARMS across the shelf-gradient in Saudi Arabia is confounded by depth which varies from 1-25 m across the different shelf positions. Therefore, the data is no use for describing either cross-shelf or cross-depth patterns in this case. I must say, I was somewhat perplexed why FARMS were not deployed in a standardized manner given that one of the main justifications for their use is standardized sampling.

Secondly, following the removal of the global analysis there needs to be some analysis in the main text, with a clear purpose that is focused on the tool itself and would help readers understand the limitations/shortcomings of this method. In this respect, I would urge the authors to take inspiration from the recent paper by Wolfe and Mumby in Methods in Ecology and Evolution on artificial RUbble samplers (which I note is already cited in the current manuscript). In their paper Wolfe and Mumby simply show a) how the abundance and community composition of organisms in their rubble samplers varies through time, and b) similarities/differences in the abundance, size and community composition of organisms collected in rubble samplers versus natural rubble. Unfortunately, the data in the current manuscript is not amenable to a temporal comparison because the length of deployments was only varied across different habitats, not within the same sites/habitats. However, the data from the Panama could be used to directly compare the fish communities from FARMS to the natural reef community in the main text. Indeed, the nMDS in Figure 5 suggests the community composition of cryptobenthic fishes on FARMS is largely nested within the natural community which is an important result. Therefore, I would move this comparison and associated results into the main text. However, when doing this, I would also think carefully about a) standardizing the abundance and species richness by sampling area (i.e. individuals m2) in some way to ensure a more direct comparison between the FARMS and enclosed fish anesthetic samples, and b) appropriately caveat the fact that the FARMS deployed in Panama were not the exact same design as those described in the current manuscript.

Thirdly, I would also include the data, analysis and results from Texas in the main text as this comparison between FARMS and OSUs is interesting and again focuses on the tool itself. In addition, it is a demonstration of how FARMS compare to another method for quantifying cryptobenthic fishes. This is important because it helps readers see how robust FARMS are as a sampling strategy.

Finally, it would seem a shame to not utilize the substantial amount of valuable data leftover in some way. However, given all of the potential confounding factors it is unclear how useful this data is beyond showing that FARMS can catch fishes. My suggestion would be to re-do the supplemental material with the express purpose of demonstrating how versatile FARMS are as a sampling unit. I would simply show the various different habitats (ranging from Estuarine to Deep Sea) in which FARMS have been deployed, and the types of fishes collected in each scenario. This should highlight to the reader that this is a versatile tool which can be applied in a variety of situations.

I believe this blueprint for the revision should help focus the manuscript on the tool itself as the main comparisons will be on how fish communities from FARMS compare to a) natural communities, and b) fish communities collected using another method. In addition, the supplemental material would highlight that this is a versatile tool that can be deployed in a variety of conditions. Obviously, this major restructuring of the paper would mean that the discussion would have to be revised accordingly. I would also suggest that when revising the manuscript care is taken with final editing. There were a number of spelling and grammatical mistakes throughout, and the details of each reference do not appear to have been checked/corrected (for example no article numbers are provided for any of the articles published in Scientific Reports that are in the reference list).

Overall, I thought this was an interesting concept and I hope my suggestions will be of use to the authors in revising the manuscript.