This is a very well written evaluation of the application of macrophotogrammetry (MPG) to benthic ecological studies in marine systems, and in particular, coral reefs. I found the approach and methodology using photogrammetry to be very sound. I also think the topic if of great interest and appropriate for this journal. My main comments are around the focus and depth of analysis as I feel the manuscript is trying to do many things but as a result, does not go very deeply into any of them. These and a few other comments are given specifically below.

Major comments
There were two main objectives, 1) to test different systems in order to see which was best, 2) to apply the technique to monitoring in a coral reef setting. I had expectations based on this of some greater detail in the comparisons and monitoring than were delivered. The end result was more in the category of “proof of concept”.

This was probably most true for aim 1. I found the table giving details of the resulting models to be very useful, but the analysis stops short of really comparing the quality of the models in any quantitative manner. For instance, rugosity was captured, but not compared between different systems. Full 3D comparisons could be done to evaluate how different they were. Then these objective measures of model quality can be related to important drivers, like depth of field, image resolution, that sort of thing, to give a bit more guidance to the reader on tradeoffs when deciding on what system to use. It would be useful to have these details, along with costs in terms of time and money, in comparable units, in one table so its more clear why you would choose one system over another. My takeaway from the paper was that all the systems work but DoF was a critical element. But I don’t think there’s any actual information on DoF ultimately delivered by the different systems so its hard to see where this conclusion comes from. Note, I don’t disagree with that conclusion, and in fact it makes perfect sense given my experience with photogrammetry and the challenges of long focal length, macro lenses, its just not used in any quantitative manner to support these conclusions.

As for aim 2, I had a similar reaction…that the focus was just showing what you could do, rather than evaluating just how much more useful than other approaches this was. So a cynic (which I am not by the way, just playing devils advocate) would ask why 3D is even needed. Why can’t you just take high res top down photos and get the benthic cover data from those? To address this I’d say there are two elements. 1) Comparing the GSD and footprint size (area of imagery) you get from such an approach with that used here. So how much better is the resolution and bigger the footprint from MPG than just using a high res SLR photo (like the alpha Riv system). And importantly, how does that affect the ability to measure the things you want to measure and ID the things you want to ID for this application. Then 2) what additional value do you get from 3D. This case is probably already pretty well made by showing profiles but could add to this by highlighting the ability to get volumes of features, which may not always relate to their 2D planar area…which is a key reason to use 3D over 2D. I say all this as it is A LOT more trouble to get these full 3D models than take some top down photos.

Another really great addition here, given the importance of DoF would be to be able to compare the value of the lighting systems. Looks like all scenarios used lights but that will be an obvious question somebody looking to set up a MPG system would use…do I really need lights. So if you can make comparisons, all the better, but even discussing what no lights does to the aperture settings and thus DoF will be very useful.

Any idea why the RMSE was higher (so alignment accuracy lower) for the Sony systems? I would have thought it would be especially high for the AR system given its high res.

Can you include alignment percentage? I ask as I see marker detection was done after alignment. Doing this prior will typically improve alignment accuracy (RMSE) and the % or cameras that align.

Minor Comments
Were cameras white balanced at all?