This is a clearly written paper with nice figures describing nice analysis of an extraordinarily rare and hard to obtain dataset.  The manuscript should be published.  I do have a number of comments, questions and morsels for thought that I list below in the order in which I read.  The majority are (very) minor, amounting to text and grammar nits, but some are more substantive.  In particular I would like to see

*more supporting evidence behind the claim that mixing is weak (for the reasons given in the final comment below), *better figure 3 and 4, which currently mixes aspect ratios, has the reader going back and forth and does not allow direct comparisons of the most relevant quantities - specifically epsilon and the different instability indicators

*quantification of the ADCP vertical wavenumber response and hence justification of the numerical values of Ri presented (or alternately toning down the reference to specific values such as Ri=1/4 given the estimates are noisy and not fully resolved),

*justification for use of median versus mean

*and finally and perhaps most substantively, an explanation for why the turbulent heat fluxes just above the bottom are important to measure.  Ie, is that the water that will eventually meet the grounding line, or should the study have been done nearer the top of the mCDW watermass where the gradients and heat losses are much stronger?

Good luck.  I enjoyed reading the paper and hope that these comments are useful.

11: topography, turbulent or both not resolved?

26: awkward

35, 53: “this” is a weak reference.  Please reword; see Strunk and White if needed.

48: Melt rates two words?

52: This statement is actually not true: epsilon is the dissipation rate and further assumptions must be invoked to infer the mixing.  This needs to be corrected and expanded upon.

55: This would be a good place to distinguish what is different about this study from the other two.

56: which -> that.  Also, is this the only reason mixing is important to know for these situations?  

66-68: Please give order of magnitude of the clock offsets before correction and the precision of the alignment afterwards.

74: Please explain why you used median instead of mean?

95: Could indicate this is likely because of F=ma; ie the same force on the huge autos produces much smaller accelerations.

105: on which this study focuses.

105 general: is this the first paper that presents the details of shear microstructure from Autosub? Surprising if so but if true, you might consider showing a few spectra and additional details, possibly in an appendix, so that future work can cite this paper.

111: Shih et al is a very bad reference for this!  They find a Re_b-dependent Gamma.  Suggest just citing Osborn (1980).  There are also now a handful of observational references supporting the assertion that gamma = 0.2.

113: How close to the bottom of the ice is the shallowest CTD measurement shown? The very strong gradients at the very top of the cavity CTD casts (Fig 2 black) are interesting.  

123 and throughout: I believe units should be in roman, not italicized, font.

136: Suggest reformatting the equation.

140: Please make it very clear that Ri (under the ice at least)  is based on a single N2 profile whereas the shear is a function of location and time.  This is OK, but appropriate caveats as to its governing local instabilities without in-situ N2 should be given.

173: Generally, avoid “there is” in favor of more active language such as “flow is to the …” 

177: High compared to what?

177: runon sentence.

Figure 3, lines 2 and 4 of caption: runon sentences. Also, the dots are said to indicate the starting locations - but they are a continuous line.  I’d have thought there would just be two starting locations, one for center and one for east?  Please clarify.

Figure 4: Personally I think it would be better to keep the aspect ratio constant between Fig 3 and 4.  Also, sine you already plotted velocity in Figure 3, suggest including a panel of N2.  The aspect ratio is all the more a problem later when the authors are comparing epsilon to the different instability indicators - but the reader must go back and forth between figure 3 and 4.  Suggest standardizing the aspect ratio and including an epsilon panel in Figure 4.  Possibly even adding Ri contours to the epsilon panel or epsilon contours to the Ri panel since the authors are trying to demonstrate correspondence between the two quantities.

Also, the Ri panel is just a big sea of red.  Consider plotting something else to highlight the unstable regions such as Ri^{-1} or Fr = Uz/N.

182: Doesn’t negative PV mean unstable? The whole water column is unstable? Is it backwards in the southern hemisphere?  Some statements to clarify would be useful.

188: I don’t agree with this statement - the high dissipation does not appear to me to line up at all with for Ri.  Furthermore, given the ADCP’s finite vertical resolution and noise, some additional detail needs to be given on how seriously we are to take the numerical value of Ri.  I think that either some wavenumber spectra and transfer functions a la Polzin 2002 need to be included, or Ri used as a qualitative indicator.  

191: I disagree; elevated mixing is much broader than the regions of Ri < 1/4 - augmenting my previous point.

193: This statement is not justified.  Epsilon appears surface intensified as well.  And while it is bottom intensified, I do not think the statement that it is heightened over rough topography, shear or high currents (of which you generally must choose either high current or high shear, not both…) is supported.  And as before, I don’t think that high epsilon lines up with low Ri either.  Either way, if this statement is retained, more analysis needs to be shown - scatter plots, binned averages, etc.

197: runon sentence.  And seemingly unrelated sentences.  Ri governs shear instability, not symmetric instability… (I understand they are highly correlated here, but they are different, so clarification is needed).

202: What is a barotropical jump?

207: Please rewrite this passive and vague sentence.

204-210: Suggest moving this speculative bit to the discussion.

216: I think it would be nice to compare this to open ocean values at a similar depth and/or abyssal values, for context.  Otherwise “weakly stable” doesn’t have meaning.

218: Style guides such as Strunk and White suggest avoiding “Figure x shows…” in favor of “statement x is true (Figure y).”

223: Figure 6 and 5 -> Figures 5 and 6

236: redundant.  Suggest “Maximum values were” or “Values reached.”

238: Again, I’m afraid I don’t see this.  There are counter examples where epsilon is high over flat bottoms.  Please include plots that allow direct comparison such as plotting epsilon with Ri, current speed or bathymetric slope over plotted, or scatter plots or binned averages (e.g. epsilon(Ri) etc)  if you want to make this claim.

241: Please remind reader that it’s Ri computed from in-situ where and 

245: Again, please include transfer function and instrument response information if you wish to quantify the numerical value of Ri versus using it as a qualitative indication.  Note as well that these transfer functions and hence the mapping of true to measured Ri will be different for the Autosub and the LADCP.

257: Is it really necessary to use a package like this to compute a spatial gradient?  More fundamentally I do not see a relationship between RMS bathymetric slope and dissipation rate.

264 onwards: consider moving all of this comparison to past work to the discussion, so that the results section just has your results?

270: I’m confused here, sorry.  Weren’t the ALR measurements entirely in the warm inflow, since they were so deep?

272: runon sentence.

273: Due to what mechanism?

281: Please change “this” to “their” to avoid confusing with your study.

285: If you are going to state dissipation rates this low, I think you do need to demonstrate your minimum detectability threshold.  Earlier you said it was 1e-10.  So how then do you get a median lower than this.

Again, I think median should be avoided for all quantities unless there is a good reason.  Why not just use the mean?

333: The reason for these calculations is revealed here - suggest giving it earlier to make the reader understand why they are being told all of this.  More fundamentally, is that the only reason turbulence is important to measure under ice shelves?  Ie, as a possible mitigator of the advective heat flux by these warm flows?

I, at least as a non ice sheet person, would like to see a cartoon (words or actual graphic) showing a cross section of the hypothesized warm water flow to the grounding line.  The reason for this is that I don’t currently understand why the study focused so much on the near-bottom mixing.  I’d think that the heat loss out of the mCDW would be better quantified near its upper edge.  As the authors point out, the water near the bottom is very weakly stratified so the heat fluxes are expected to be small.  Aloft nearer the interface, the gradients would be stronger, but also the distance from the topography which is presumably generating most of the turbulence (my comments above about that not having been adequately demonstrated notwithstanding).  So, statements that mixing is weak such as on lines 356-258 should be tempered somewhat.  And I think the cartoon or written description of the flow giving readers the sense of which depths are thought the most likely to eventually contact the ice would help inform this discussion, at least for me.

This paper discusses microstructure observations made beneath Dotson Ice Shelf using an Autonomous Submersible Vehicle.  The data appear to lack the temporal and spatial coverage that would enable substantive conclusions to be drawn about the role of turbulent mixing in the larger-scale processes that operate beneath the ice shelf.  They are, nevertheless, intrinsically interesting, in that they represent some of the very few direct observations that we have from within a sub-ice-shelf cavity.  That remote part of the ocean plays a pivotal role in setting the mass balance of the Antarctic Ice Sheet and hence its impact on global sea level, so any observations are of value.  I would therefore recommend publication of the paper with only relatively minor changes.

Suggested changes:

Title:

It’s a minor point, but the current title does not reflect the content of the paper very well.  It promises observations of ocean currents.  While they are included there is very little discussion of them, and no more space is devoted to currents than to water properties.

Abstract:

Reflects the content of the paper and thus its main weakness, which is a lack of substantive conclusions.  I accept that it is hard to put such detailed observations into a broader context, especially when they are made in such a data-poor region.  However, I wonder if it might be possible to put an order of magnitude estimate on the cavity-wide mean vertical heat flux, given estimates of inflow/outflow temperatures, residence time, melt rate, etc.  That would put the numbers quoted in the abstract into a useful context.

Introduction:

The first paragraph talks about the DIS contribution to Amundsen Sea “mass loss”, suggesting that the term refers to shrinkage of the ice sheet.  However, the latter part of the paragraph partitions “mass loss” for the ice shelves between calving and melting.  In this instance the term does not refer to shrinkage of the ice shelves, but the contribution to the wastage side of the mass budget.  Those are different concepts, and the distinction should be clarified.

The last paragraph states that there have been only two previous published studies of mixing beneath ice shelves, but that overlooks studies based on borehole data.  The oversight is repeated in other parts of the manuscript.

Data and methods:

On line 127 there is a parenthetical note to authors that has not been addressed.

On line 140 the dimensionless parameter could more precisely be referred to as a “gradient Richardson number”.

On line 165 there is a mention of detiding LADCP data using CATS2008.  Elsewhere it is stated that tides are unimportant, and CATS cannot be trusted because of the poor bathymetry in the model.  One comment refers to (mainly) sub-ice data and the other to ice front data, but nevertheless the treatment seems inconsistent.  If bathymetry is poor beneath the ice, won’t that influence currents at the ice front?  If tides are weak enough to be ignored, why bother with detiding the LADCP data?

Results and discussion:

In the title of section 3.1 and elsewhere in the manuscript the edge of the ice shelf is referred to as the “ice shelf front”.  The correct term for that feature is the “ice front”.

In figure 3, is there a “black line” showing the track of the ALR (third line of the caption)?  I couldn’t see one.

On lines 222-223 it is stated that water at the ice front is colder and lighter than that in the cavity.  Does that refer only to measurements made with ALR?  Was the warmer, saltier water apparent in the section observed with the ship?  If not, I think it deserves some comment about where that warm, salty water may have come from?  Waters in the cavity must be cooled and freshened, so the observation must say something about variability at the ice front.  If, on the other hand, an equally warm, salty water mass is present in the ship CTD data, then the statement in the paper is a little misleading.

On lines 228-229, and elsewhere, it is stated that the observations reported in the paper are important for establishing mixing rates that can be “incorporated into numerical models”.  It is not clear to me how these data would be incorporated into a model.  Perhaps the point could be clarified?

On line 230, mention is made of a 100 m thick “melt layer” observed through a borehole.  What feature are you referring to?  The upper 100 m of the borehole data shown in Figure 2 appear to indicate the presence of less meltwater than deeper in the water column.  Why is that?  A shallow intrusion of WW along the ice shelf base?

Lines 197-203 draw comparisons with observations made at Pine Island Ice Front but point out differences in the physical setting.  One difference that might be relevant, but which appears to have been overlooked, is that in the case of Pine Island there is a neighboring ice shelf to the north, so the northern sidewall of the channel confining the Pine Island Ice Shelf does not extend all the way to the ice front.

The last four paragraphs compare findings with other AUV based observations of microstructure beneath ice shelves.  However, elsewhere the manuscript highlights the differences between those regions.  That makes the discussion feel like one that is motivated by common methodology rather than common physical setting.  Why overlook borehole measurements of turbulence that have been made within cavities?  Later in the section it is suggested that the AUV track beneath FRIS is 9 km long, but that does not seem to fit with the figures in the cited paper.  At the end the of the section the text again talks about improving parameterisations of mixing in models, but again, I don’t really see how you would use the data for that.

Lines 334-335 suggest that the small vertical heat flux observed means that a lot of ocean heat can be used to melt ice at the grounding line.  But how much is used there?  The outflows at the ice front remain above the freezing point, so some ocean heat that enters the cavity exits it without being used for melting.  Again, can you estimate some global budgets for the amount of heat used for melting and the overall average vertical heat flux that could put your spatially-limited observations in a DIS-cavity-relevant context?

This paper presents an interesting set of observations, in an environment difficult to access. The analysis is solid. It would be good to put these observations in the context of the previous work that has been done around Dotson - I understand that there are little observations in the cavity, but are the conditions along the face 'unusual'? It's hard to tell, and I acknowledge that this is not about long-term observations at Dotson, but it would be useful to put these observations in a broader context. 

As pointed out by the other reviewers, some of the key results are a bit either overstated, or unclear. For example, on L193, one would be hard-pressed to directly identify the 'enhanced mixing at the inflow' and it being over a larger area than that of the outflow from the section alone - it might useful to show a profile or two of dissipation rates. Sampling (station spacing) might be important when talking about "area", which is not discussed here. 

Overall, I don't have many comments that were not captured by the other reviewers. This is an interesting paper and it should be published.