This is an interesting and well written method paper. Although a method paper, some observations are thought-provoking. The ecological aspects may deserve a dedicated paper rather than ‘hiding’ the data in a technical note.

Important is if the more complex method is better, and worth the extra costs and effort. In the end the results with 3 sensors are similar those from the 2 sensor design, so would the conclusion be that 1 is enough? Line 164 states that the differences were not statistically significant. The advantages of the 3 sensor system is at the moment not well explained. This is my main issue with this manuscript.

The introduction is complete, but the focus could be a bit shifted. It should start with the method rather then the carbonate sands in reefs, as the method is more generally applicable than in reef sands. The strength of the eddy covariance method is that it can measure exchange fluxes with all benthic ecosystems, with few exceptions, and from those fluxes infer metabolic activities. As the paper describes an improved method, the current issues should be better introduced. At the moment designs with 1 and 2 sensors are used. The reasoning for using a 3th sensor should be better explained. My understanding is that the variations in flow direction make it difficult to link the local flow direction and speed to the local oxygen concentration, as the flow and concentrations are measured a few cm apart. The corrections are now made by continuously time-shifting the oxygen signal by the measured flow speed and direction, and thus synchronize both. This is apparently not perfect. Exactly how 3 sensors can improve things should be described in detail, best with sketches.

The habitat in which the tests are done seems not the most challenging, as the flow direction is rather constant. Not much wave action occurs at 10 m depth.

The method deserves better explanation. Fig. 1 is good in the sense that is complete, but one needs a microscope. Crucial parts need to be magnified: the yellow cylinder and red triangle in 1b (in my high quality printout an amorphous colored little blob), the yellow ring in 1c is invisible, 1d can be omitted as it has no information. Is the pink stuff in 1a an error or something real?

A mathematical proof is needed for the crucial assumption that the average of the 3 sensors is the concentration in the geometric center of the sensors. This is crucial for the paper and should be provided. From Fig 8 one can see that a lot of detailed post processing is needed to get to the average!

Comparing Fig 2 and 4, there seems to be no relation between ecosystem activity and hydraulics. There seems a relation between the integral of the irradiance (photons/day) and both productivity at day and the respiration at night. Would be good to do some calculations on this. Would be indeed logical and understandable if the metabolic activity, including respiration at night, is controlled by the light input.

The cited numbers in L 211 and 212 are 1000x off. Actually Koopmans measured similar (to this study) metabolic rates in an equally illuminated, also oligotrophic, ecosystem. That the two ecosystems, although totally different (seagrass canopies versus reef MBP), are similarly productive at the same light level seems meaningful. Both benthic photosynthesis and respiration are controlled by light input.

The oxygen exchange rates measured in this study are well in the reported ranges. These ranges are very wide. L 195 states 'matching or exceeding', but rates are truly about average, sometimes below, sometime above average (see Fig 6).

It is interesting that small eddies can transport oxygen in a different direction as big eddies. These go thus in different directions. Can a physical interpretation be given for this phenomenon, and an indication of their sizes? The big eddies persist for several hours, is that correctly understood? This phenomenon, if true, requires a larger discussion, as it may have a large impact on the interpretation of such data.

Some minor points:

The ecosystem is net autotrophic, and since no organic deposits are build up, the produced organics must be exported. Any suggestions how?

‘elevation’ in legend Fig. 8 must be defined and explained.

Overall this is a very interesting paper and one I enjoyed reading very much. It tests whether adding additional O₂ sensors to an eddy-covariance instrument improves aliasing in data due to the separation of velocity and O₂ sampling locations.  Given the increasing use of eddy covariance measurements to estimate benthic primary production/respiration at scale technological improvements are timely and welcome. I am very supportive of this paper however there are a number of elements that if considered in revision could improve the focus and clarity of the manuscript.

My main comment addresses the multiple elements to this paper. It seems to bounce around between a confirmation of the fact that shallow water permeable carbonate sands are hot spots of benthic primary production and organic matter processing and testing  whether additional O₂ sensors improves the precision of flux measurement. Given the paper is submitted as a technical note it could be improved (and shortened) by focusing on the increase in performance of adding additional sensors. The sections of the paper discussing the high production/respiration rates of carbonate sands confirms previous studies (see Fig 6) and in my opinion distracts from the method which is generalizable to many systems.  If using eddy-covariance techniques I would want to know exactly what gains could be made by adding sensors as the additionally increases costs and/or may reduce the ability to spatial replicate units. These trade-offs are import – is it more important to increase the precision at one location or potentially increase the number of locations at which flux measurements are made to assess spatial variability? So the questions I would like answered explicitly are what improvements are made my adding sensors in terms of precision and do these improvements vary with hydrodynamic setting (eg. uniform steady flow vs more wave dominated flows), do these improvements really matter in system with high natural variability in fluxes and what other conditions/settings need testing to confirm the value of sensor additions.  Revising the manuscript (mainly editing the Introduction/Discussion) with this comment in mind I think would result in a much more assessable paper with a tighter focus. In short make the technical note more about the method than the system in which it was tested.

Specific Comments

1. Ln 160 Please provide more detail on what data was used in t-test comparing the 3OS and 2OS. The DF indicates 7 data points – were these the average of the 15 min blocks across the four sample dates? I am assuming that both the 3OEC and 2OES systems were synched so perhaps a better test may have been a paired t-test where you ask whether the difference between the data is <> than 0.
2. Line 165 Two sentence paragraph that does not make sense on its own
3. Fig 3 – Add the p values for the regression statistics and clarify what data is being average for each of the visible data points. If the data represents averages of 15 min intervals then surely there is a variation in mean current velocity between intervals that should be plotted as an error term? I would also like to see what if any difference results from the generated PI curves from the 2OEC and 3OEC systems – are the fits better (r²) are the fitted parameters known with greater precision and does this matter?
4. Fig 5 Are there any corrections applied to data in the (a)? That is, is the variation observed between individual sensors a function of sensor performance vs data that has not been corrected for R, S, T & W.
5. Line 180 The order is a little illogical – when looking at Fig 5 c&d I see a T correction was included in the 3OEC data processing – yet in the Introduction it was emphasised that the advantage of using 3 sensors was to avoid needing to do this. It is not until the Discussion that this is discrepancy is explained. I would suggest that Fig 8 & text goes into the Results to explain/justify this correction. Alternatively improve the legend to Fig 5 and point to the Discussion for explanation.
6. As mentioned above the Discussion might be better focused on the comparisons between sensor configurations. To aid this the authors could consider adding a summary table that summarises the increased in precision and compares this the natural variability and provide some ‘cost-benefit’ analysis for investigators.  Are there conditions where a 1 or 2 sensor system may give similar results to a 3 sensor system and where should researchers favour a 3 sensor system?