Responses to Reviewers, JAMC-D-17-0131

We greatly appreciate the many comments from the three reviewers, which have significantly improved the original manuscript. The anonymous reviews are now credited in the acknowledgments.

Reviewer 1:

The manuscript could be improved by formulating scientific questions more explicitly and by emphasizing differences to previous studies more clearly, i.e., by emphasizing the novelty of this study. The structure of the abstract and the conclusions could be improved as well.

We have now rewritten the abstract and conclusions. Our responses to the scientific questions comment are addressed below.

Main comments:

1. The manuscript is rather long and certain parts/descriptions are somewhat long-winded. The number of figures is rather high too. Are all figures and descriptions of details necessary to derive the conceptual model of Fig. 15?

We tightened up the text and have been able to reduce the word count significantly despite additional text added to meet reviewer suggestions. We have now omitted Figs 4, 8a, 12a, b, and c and have made minor revisions to other figures. Figure 14 was split into two figures (now Figs. 13 and 14).

2. The abstract immediately starts with a detailed description of the observed flow over an obstacle (crater rim) into a basin. A clear structure including parts such as short motivation and goals are missing. Similarly, the title describes the observed process (i.e., lifting and penetration) but not really the goal of the study. Further, the last section (summary and conclusions) contains one long paragraph (except from the short introduction and a final remark). I suggest to divide this paragraph into smaller pieces of information, i.e., concise conclusions.

Motivation and goals have been added (lines 80-85). The title has been changed to "Katabatically driven cold air intrusions into a basin atmosphere". The text in the conclusions has now been moved to a new section (Conceptual model), with the material broken into additional paragraphs. New text has been added to the conclusions to summarize what is new and how the findings relate to our earlier METCRAX cold air pool paper (Whiteman et al. 2010).

3. Clear goals of this study or research questions are missing at the end of the introduction or background. They should be developed based on gaps in knowledge, i.e., at the end of the literature review. Here it is important to point to open questions that could not be answered in METCRAX I but will be answered in the present manuscript. Just another case study with more sophisticated instrumentation does not imply a clear goal and does not justify another

publication. More specifically: is there any difference between the mechanism of “cold-air intrusion” proposed in Whiteman et al. (2010) and the penetration of the katabatic flow proposed in the present manuscript?

There have been few previous studies on cold air intrusions into basins, so that gaps in knowledge come primarily from the previous METCRAX study where the instrumentation was focused mostly inside the crater. In this paper we add information on atmospheric structure outside the crater, overflow over the rim and the atmospheric structure this produces inside the crater. These general topics are now mentioned at the end of the Introduction (lines 80-85) and in the Background (lines 103-107). The bifurcation, cavity and hydraulic jump were unanticipated new findings not covered in the Whiteman et al. (2010) cold air intrusion paper. Another motivation (producing a dataset to test numerical models) is now mentioned in lines 184-185 and 656-662. The conceptual model and Conclusions sections now better emphasize these new findings.

4. The authors propose a conceptual model (Fig. 15) for the penetration of a katabatic flow into a basin based on a single case selected from a rich dataset of several IOPs. What was the motivation for choosing this “golden day”? How representative is this flow pattern? How often does it occur on other quiescent nights? I suggest to add some motivation for choosing this case to the last paragraph of section 3 and some information about the representativity and/or frequency of the proposed flow pattern at the end of section 8.

We added a motivation sentence and information on representativity at the end of section 3, as suggested (lines 177-185).

Most IOPs were disturbed to some extent by larger-scale (both synoptic and mesoscale) flows above the crater or by residual cloudiness. IOP7 was relatively undisturbed and thus the most suitable for investigating unperturbed mesoscale katabatic processes and mesocale/microscale processes in and above the crater. This case study is thus the most suitable for future modeling. Unfortunately, we have no long-term data sets at the Meteor Crater to help with the determination of representativity and we were largely unsuccessful in the field in using synoptic charts to forecast whether an upcoming event would be fully quiescent or somewhat disturbed by other mesoscale influences.

Minor and technical comments:

1. Page 3, line 84: Why “unusual”, i.e., in which respect?

Done. We wished to emphasize the small scale of the topography, which allowed the use of many surface-based measurement instruments. Lines 85-87.

2. Page 7, line 168 and Fig. 2: Include description of primed quantities also in caption of Fig. 2

Done. Lines 781-785.

3. Page 7, line 169: “... sampled every 2.5 minutes”: instantaneous values or averages over 2.5 minutes?

Done. Instantaneous. Line 154.

4. Page 8, line 187: Potential temperature is usually defined for a reference pressure of 1000 hPa. Please, mention more explicitly that a different definition for a certain reference height is used.

Done. Lines 173-174.

5. Page 8, line 187: Add units to temperature gradient in the formula (K m-1) or alternatively replace it by g/cp.

Done. Lines 171-176.

6. Page 8, line 189-190: It is indeed convenient to use “temperature” as a synonym for “potential temperature” in this manuscript, however, it is confusing. It might lead to misinterpretation of figures if somebody is not aware of this convention. Hence, I suggest to use potential temperature everywhere.

Done. We now use potential temperature throughout the manuscript.

7. Page 8, line 202-204: Is this result surprising? The dependence of height and magnitude of the wind maximum on the slope angle has been shown by numerical simulations (Schuman 1990) as well as linear theory (Prandtl 1944).

Schumann's simulations were for upslope flows. According to Prandtl (1942), downslope flow speeds should be greater on steeper slopes, and jet max heights should be greater on shallower slopes. The high speeds on our shallow slope are in disagreement with Prandtl's equation. The stronger flow speeds on shallower slopes are supported by other research, as reported by Zardi and Whiteman (2012).

8. Page 9, line 215: What about the effect of surface friction on wind direction (Ekman layer)?

We have now included this (Lines 207-208). In responding to this comment we also found and corrected a mistake in our use of our backing/veering terminology.

9. Page 10, line 229: Does this decrease in static stability imply vertical stretching of isentropes as the flow approaches the rim? Wouldn't a hydraulic-type of flow rather be characterized by squeezed isentropes? Such squeezed streamlines can be seen, e.g., in Fig. 3.28c of Chow et al. (2013) upstream of the obstacle.

When we assumed that the elevated layer flowing over the rim was lifted adiabatically, the destabilization could be attributed entirely to column stretching. We have performed some new analyses of heat fluxes at the lowest levels of the RIM tower and the levels

just above the dividing streamline on the NEAR tower, finding that the heat flux divergence increases with height in these shallow layers at both sites. This heat flux divergence would tend to cause a destabilization of the profiles at the lowest levels of the RIM tower. We have revised the text to include this information. Lines 218-221.

10. Page 11, line 268: “amount of lifting”: how calculated?

Done. Lines 253-254.

11. Page 11 line 269-270: Avoid referring to Fig. 7 before Fig. 6.

We were not able to easily change the order of callout for this figure but have fixed the problem by stating "to be seen later in Fig. 6d". Line 255.

12. Page 11, line 272: “These outflows are warm” – appears to be not true for NR which is slightly colder than WR (see Fig. 5).

NR is actually inside the crater a few meters below the rim. We think this is the reason why it is slightly colder than the other downwind rim sites. We have now mentioned this in the text. Lines 260-261.

13. Page 12, line 281: I can see variable wind directions at FLR (i.e., larger scatter) but no clear/regular oscillations. What is the oscillation period?

The oscillations are not expressed very well in this IOP and we have no figure that really focuses on this phenomenon. We are preparing a separate paper on seiches (first author is Manuela Lehner and we plan to submit this to BLM), although that focuses on a different IOP when the seiches were more regular. The oscillation period then was about 10-20 minutes.

14. Page 12, line 294: Any explanation for this short-term drop in wind speeds at the rim? It occurred at the time when winds at about 100 m above the rim were strongest (Fig. 6g) and temperature at the rim showed a local maximum (Fig. 7b). Any connection between these features?

We think this is a bluff-body separation that occurs in the lee of the rim when winds aloft accelerate the overflow. This is the subject of another paper that Dr. Lehner is preparing, which discusses the phenomenon when it is better expressed in IOP4. To save space and keep our focus we prefer not to comment on it in this paper. 15. Page 13, line 302: Reference to Fig. 7a not clear. Where do we see the arrival/start of the katabatic winds at the top of the tower? Change to south-westerly wind direction?

We have added that the arrival occurred with a shift in wind direction to WSW. Line 289.

16. Page 14, line 321-324: This sentence is difficult to follow.

We have rewritten this section to make it easier to follow and to decrease the word count. Lines 308-324. We have also moved some text to the Fig. 7 caption.

17. Page 15, line 350: Add units to the coefficients of this temperature formula, i.e., 4.82°C and 4.39°C.

Done. We also added that to the text (line 318) as well as to the equation in Fig. 7.

18. Page 16, line 384-385: Provide arguments for the assumption that the NE HOBO line represents undisturbed conditions. Can we indeed assume that the pseudo-vertical profile NE along the sidewall is about the same as the vertical profile at the crater center (e.g., at TS-C)? For example, lifting of the airflow at the downwind part of the crater (i.e., in the outflow region) would change the temperature structure compared to the crater center.

We have added text to justify the use of the NE profile as the relatively undisturbed background profile in the crater (lines 342-346), as well as providing a better explanation of what the pseudo-vertical profiles represent (339-342). The NE and SSW profiles are both on a slope where some of the same physical processes are operating (e.g., radiative cooling to similar sky view factors and nearness to the underlying surface). The mean NE profile is slightly (0-1 K) colder than the mean free air TS-C profiles. Also, the NE profiles have much better time resolution than the TS-C profiles.

19. Page 17, line 408: Please explain zUi'.

We have been able to dispense with this terminology. (Line 435)

20. Page 19, line 434: I may have misunderstood the method of LNBE calculation (Fig. 11). Nevertheless, it appears that the authors assume that in the layer (determined from the rim profile) does not change along the sidewall. However, turbulent mixing (as indicated on page 17, line 407) may have changed the stability and, hence, along the sidewall. This in turn would affect LNBE.

For the LNBE calculation we assumed that Δ does not change along the sidewall. As this reviewer stated, this approximation does not account for the weak warming caused by turbulent mixing as the cold air intrusion comes down the sidewall. This is now stated more clearly in the text. Lines 449-451.

21. Page 20, line 472-473: Why are radial velocities less affected by ground clutter than wind vectors derived from dual-Doppler retrievals? Does this imply that the lidar located at the crater rim is stronger affected by ground clutter than the lidar located at the crater floor?

To retrieve two-dimensional dual-Doppler winds over the sidewall it is necessary that the range gates from the two lidars overlap and neither is affected by ground reflections (clutter). Because the N rim lidar looks down on the slope at a high incidence angle, the range gate intersecting the slope has to be discarded. This means that dual-lidar winds

cannot be retrieved close to the slope. On the other hand, the RHI scans from the floor lidar include low angle scans that are nearly parallel and close to the underlying slope. We have rewritten the description to make this clearer. Lines 543-547.

22. Page 21, line 493: Could this video animation be provided as a supplement on the journal's website? (https://goo.gl/83rfBJ)

JAMC, unfortunately (and unlike BAMS), does not have a website for supplemental material. But we have found an alternative long-term website and have put the video animation on that website with appropriate links in the article.

23. Page 22, line 514: First time mentioning “extrusion”. Please explain and refer to Fig. 13 if appropriate.

The term (extrude) is actually first used two paragraphs above (line 526), where it refers to new Fig. 13b.

24. Page 23, line 530: “... is shown in the figure” – which one?

Done. We have now referred to new Fig. 12. Line 474.

25. Page 23, line 530-531: Not clear which type of momentum equation has been used here – please provide appropriate reference. Is it formulated for rotated (slope-parallel) coordinates? “Vertical” on line 530 appears to be rather “horizontal” or more specifically “along-slope”. Furthermore, the equation shown is not the momentum equation itself but the integrated version. Is u here the slope-parallel component rather than the truly horizontal component? If so, has this been considered when using wind measurements for the calculation of s0?

We have now done a better job introducing this equation. Line 474-476.

26. Page 23, line 538: Where do we see the result of this analysis? Fig. 12?

Done. We have added a reference to Fig. 12. Line 482.

27. Page 24, line 569: “... places it just above ...” – referring here to surface-based wind layer?

Done. Lines 373-376.

28. Page 25, line 580-581: Reinecke and Durran (2008) studied the effect of nonuniform static stability rather than the effect of nonuniform wind speed. In fact, wind speed was assumed to be constant.

We have omitted the erroneous reference to Reinecke and Durran.

29. Page 25, section 7: What about the relation between the nondimensional ridge height and the amount of blocking? Could the former be used to diagnose the height of the dividing streamline?

Is there an agreement with observations?

We tried this approach earlier, but found no agreement with observations and decided not to discuss this analysis so as to save space.

30. Figure 1: Please explain black dotted lines.

Done. Line 775.

31. Figure 2: Please explain primed letters in caption.

Done in the text and now, also, in the figure caption.

32. Figure 8: Please add units to coefficients in formulas shown in panel (a) and (b). See similar comments further above.

We have eliminated panel (a), but have made the requested change in panel (b), which is now Fig. 7.

33. Figure 9: Reference to i=1, 2, and 3 is not clear. These numbers/indices are not explicitly shown in the figure.

The mathematical indexing has now been removed from this figure, although it is required and, now, better explained in the following figure (now Fig. 10).

34. Figure 12: Please explain zp.

Done. Now in caption.

35. Figure 13: Please increase figure size. Labels, arrows and other features are small and difficult to read. For better orientation please use same type of x-axis on all sub-figures, i.e., with the origin at the crater center.

We have separated Fig. 13 into two figures (now, Figs. 13 and 14), the first displaying the dual-Doppler images and the second displaying the single Doppler RHI scans. The first figure is now set in a single column so that the text size is acceptable. For Fig. 14 we have retained the standard method of displaying the images so that the x-axis origin is the lidar location. The blue dots on the basin floor in Fig. 13 locate the x-axis origin for Fig. 14.

Reviewer 2:

1. Your study is based on observations on one particular night that had clear sky and undisturbed background condition.  Over the duration of your experiment, I am sure you had other nights with similar conditions.  The question is how representative is this case?  In other words, how often does this occur?

See the response to Reviewer 1, major comment 4. This was the best night relative to lack of disturbance by significant background winds above the crater. We have added text regarding the representativeness of this case study. Lines 177-185.

2. I understand this may be beyond the scope of the paper, but can you at least speculate, based on your observations on other days, how the behavior of the flow might change with changes in cloudiness, background wind and stability?

Cloudiness would be expected to decrease the nighttime cooling and thus the strength, depth and overall stability of the katabatic layer on the plain. We prefer not to speculate on this in the current paper in view of the space limitation. We have submitted a separate paper on IOP4, which had higher background winds. This paper is now referenced in the Conclusions (Line 663).

3. A concern I have (other readers might also have) is that although the phenomenon is well described and the paper is well written, why would one care about something that happened on a particular location with unique topography (crater with rim of particular height).  Can you try to generalize the results?

A new paragraph in the Conclusions section (lines 656-662) now states clearly that the Meteor Crater is an idealized setting, but that the findings and the dataset could be used to make future parameterized model simulations that will help to generalize the findings, which may be applicable not only to atmospheric but also to other fluid settings. See also lines 85-90 and 180-185.

4. Although I appreciate the amount of details the authors put into the description of the experiment and the phenomenon, the writing could be tightened up somewhat.  I understand the authors have published a series papers on the observed katabatic flows with different focuses.  It would be nice to see some discussions that highlight the differences between this and previous papers.

We have rewritten several sections of the paper and tightened up the text significantly, reducing the work count. We have also now highlighted the differences between this and previous papers in the Conclusions section. Lines 632-662.

5. Panels in Fig. 6 could be enlarged somewhat to make each individual plots easier to seen.

This figure is very busy, with quite a bit of small type. As you can see, we have pushed the 8 sub-figures together as much as possible to eliminate extra space. We will ask the

editor to set the figure as page width, which we think will solve this problem.

Reviewer 3

Major comments

1. Incomplete treatment of the lidar data: While I appreciate that the primary analysis in this paper is from the in situ sensors, the inclusion of the lidar data, while very interesting, seems incomplete in its treatment. Many of the inferences and conclusions drawn from the lidar data are not fully supported by the presented snap shots. Rather vague suggestions based on the authors “behind the scenes” review of a fuller set of data are used to support statements that include “frequently”, “tend to”, “often”, etc. While I don’t doubt that many of these conclusions are justified it is difficult for a reader to reach the same understanding based on the limited “snap shots” at interesting times. My primary recommendation is to include additional lidar time-mean and time-variance information for longer periods (e.g., the steady state portion of the night). Such analyses would quantify the persistence and transience of the flow regimes discussed. For example, variability in the location of the hydraulic jump and near surface convergence might be better supported (maybe the variance of the radial velocity would show this?). To accomplish these analyses I’d recommend interpolating the lidar data to a common polar coordinate grid (assuming the raw data is not perfectly gridded) to facilitate the computation of temporal statistics. Presumably the dual Doppler data are already on a common grid and time-statistics can be computed in a straightforward manner. I do think that the snap shots themselves are interesting, but need to be seen in the context of the “steady state” period as a whole.

As suggested, we have now added 5-h-mean figures for the dual -Doppler and single-Doppler data and added additional text to better highlight the Doppler lidar findings. We have also put an animation of the entire sequence of 5-min dual-Doppler lidar images on a webpage with links in the paper so that the full variability can be seen by readers. We have also submitted a second paper dealing with the higher wind conditions in IOP4 that has a more complete treatment of lidar data in it. Animations of that lidar data will also be placed online. Space requirements did not allow us to include in the manuscript some of the new analyses that we performed in response to this reviewer comment. But we include these below for the reviewer.

2. Super adiabatic layer: I have a few of issues with this characterization. First, this is only a very slightly super adiabatic layer compared to what are typically discussed as daytime super- adiabatic layers (a few K). This should be acknowledged. Secondly, and to this end, the HOBO accuracy is listed as .25 K, suggesting that this very slightly non-adiabatic variation with height (or distance as it were) is within the uncertainty of the measurement. Thirdly, since the layer in question is a sloping layer, not a vertical profile, I’m not sure it is fair to characterize it in terms of a lapse rate. It is indeed a pseudo-vertical profile, but as you rightly point out the air mass is somewhat modified in its descent of the slope via turbulent mixing, which should warm the layer slightly as it descends. Since the observations do seem to generally support this warming (small as it is) I think that is a good characterization. However, this does not imply a convective overturning and mixing mechanism since the warmed air does not directly underlie potentially colder air (as is the case in the daytime CBL). I believe all of these caveats need to be discussed when the super-adiabatic layer is first presented to readers. Further, perhaps super adiabatic is

the wrong term since it is not a true lapse rate. Would it be better to characterize it simply as diabatic temperature variation along the slope? Or do you have evidence from direct vertical profiles that a super-adiabatic layer exists?

The weakly super-adiabatic layer is a persistent feature in pseudo-vertical profiles from both METCRAX and METCRAX I, is found on all upwind HOBO lines and cannot be attributed to accuracy of the HOBO data. We now explain that this is a modified along-slope stability obtained by plotting the near-surface observations as a function of their elevations and that these "along-slope" but weak super-adiabatic profiles have a different meaning than free atmospheric profiles (which do not have super-adiabatic layers). The pseudo-vertical super-adiabatic profiles do not mean that convection will occur up the slope or vertically above the slope. This is now carefully explained in the text. Lines 339-340, 362-368, 417-419, and 434-443.

3. Turbulent mixing and neutral buoyancy: Since turbulent mixing modifies the descending airstream that seems to inhibit the applicability of the buoyancy sorting mechanism (at least as I understand it). The parcels are being mixed with their surroundings and thus their level of neutral buoyancy presumably changes with time! The LNBE approach probably mitigates some of this error since the mixing and warming of the descending air is relatively minor (i.e. the along slope potential temperature is only very slightly “super-adiabatic”). That the two analyses generally agree is probably sufficient, but you might mention this as a possible limitation.

The weak rate of warming at the surface due to near-surface shear-induced turbulence as the cold air intrusion comes down the sidewall would have a relatively minor effect on the surface temperature used to calculate Δ. The potential temperature at the top of the cold air intrusion was not measured, nor could we determine directly whether the buoyancy sorting mechanism was occurring. Thus, our estimate of LNBE assumes that the potential temperature difference of the intruding air does not change coming down the sidewall. We have added text to make it clear that that is an implicit assumption. Lines 449-451.

4. Lifting heights and dividing streamlines: I was a bit confused at times by some of the description of the various measures of flow lifting and blocking. I don’t have a coherent explanation of exactly what the problem is, but I think some overall revision to aid in interpretation and simplification of these metrics is warranted.

We have omitted old Fig. 8a dealing with dividing streamline heights. The text dealing with lifting heights and dividing streamlines has been rewritten and put in a separate, shorter and, hopefully, clearer section (Section 4c, esp. lines 301-324) that now emphasizes temperature changes around the rim.

Specific/minor comments:

1. Figures: A few of the figures seem unnecessary and/or confusing. Figs. 4, 9, 10, and 12 can probably all be simplified and improved. Fig. 4, in particular, is difficult to interpret and does not seem to add significantly to the manuscript. In Fig. 9 the inclusion of various times (multi-

panel) is probably only interesting if you are demonstrating notable changes from one time to another. As I read the text I didn’t really get the sense that there were such key changes. As such this might be reduced to one time mean (plus standard error) set of profiles. In Fig. 10 I found the use of brackets a bit confusing... is there a different convection that might be used? In Fig. 12 I wasn’t ever sure I understood the insight provided by time-height temperature variation panels (a-c), whereas the last panel seemed to contain most of the information discussed in the text. It might be possible to remove the time-heights.

See response to Reviewer 1, no. 1. We removed a figure and several sub-figures. We removed the brackets in what is now Fig. 9 (was Fig. 10) and the three time-height sub-figures in Fig. 7 (was Fig. 8). We felt that the hourly profiles in Fig. 9 provided the reader an idea of the time deviations and also illustrated NBEs. Fig. 8 (was Fig. 14), which shows mean temperature profiles over the 5-h period, has been moved forward in the manuscript. We have also put an animation of all 5-min profiles on a website with a link in the paper so that the highest time variations can be viewed by readers.

2. Line 174: is NEAR an acronym?

No. We used NEAR and FAR to designate two of the sites. Unfortunately, we couldn't think of better names at the time that NCAR needed site names.

3. Line 187: it should say K not deg C

Potential temperature is reported in °C using the basin floor as the reference level. To reduce confusion, we have now changed the text so that potential temperature differences, gradients and tendencies are reported in K. Lines 171-176.

4. Line 230-236 (Fig. 4): This section is confusing in wording, and I’m not entirely sure how it adds to the interpretation that the static stability in the layer is decreased? To this end, Fig. 4a,b is difficult to interpret. I personally think the figure and these lines could be removed as they are not a central component of the analysis or the conclusion presented at the end of the paper.

We have now removed the figure and revised the text. See also the response to Reviewer 1, no. 9.

5. Lines 248-261: This is a very compelling set of observations and nicely summarized here. My only comment would be that on line 261 you might add a caveat that the processes leading to the vertical mixing in the wake zone are as of yet unknown (could be all sorts of interesting wake effects).

Done. Lines 246-247.

6. Line 309-313: This spatial analysis of dividing stream line height is somewhat confusing in terminology. The dividing streamline concept was clear earlier for the along wind approach, but is somewhat unclear here.

We have rewritten section 4c to clarify the concept. Lines 322-324 give a better description of the concept. Incidentally, Fig. 4 in Smith (1989) is a nice illustration of this concept.

7. Line 314-316: This seems to assume that the flow is lifted adiabatically, but it that isn’t strictly true. The RIM profiles presented earlier showed that the static stability in the lower portion of the profile was decreased, which is very likely a result of turbulent mixing, which from a parcel perspective is not adiabatic. In other words, I’m not sure that this simplistic approach of how far the air was lifted to get a given temperature is valid.

We have now computed turbulent sensible heat fluxes for the lower levels of the RIM tower and the portions of the NEAR tower just above the dividing streamline, finding that turbulence could destabilize these portions of the overflow. We have accordingly changed the text. The bulk of the upstream profiles are lifted adiabatically, although the lowest 10s of meters do seem to be affected by diabatic processes that destabilize them as they are lifted over the rim.

8. Line 342-343: Can you clarify what delta theta is? I don’t quite follow, and thus the following analysis isn’t clear either. Perhaps and example would help here. (for example, delta that of X means that ....). This will help establish why delta theta =0 reflects the coldest temperature (I kind of get it but it took many reads).

This section has been rewritten to better explain the approach and the results of the analysis. Line 318.

9. Line 350: I think that units must be attached to both right hand side terms of this equation (e.g., K). More broadly, is this equation necessary. This is not a fit that other researchers might apply universally to cold pool data sets and thus it isn’t particularly interesting or useful. I think it is enough to not the angular dependence of the data.

We have removed Fig. 8a, its equation and discussion of the adiabatic lifting, focusing on the angular temperature dependence, as suggested. We have left the equation for delta theta in the paper as we think this is an interesting result, although it also is not directly applicable to non-circular craters.

10. Line 297-358: This is the stickiest part of the manuscript for me. It’s tough to get through, and while the presented data are valid, I think it somewhat obfuscates the primary processes, which is the piling up of cold air on wind normal slopes.... I’d consider revising this section in favor of one metric showing this. As it is there are three closely related metrics and it gets muddled.

We have now removed the lifting height metric figure and have rewritten this section (now section 4c). We think the new text will be much easier to follow.

11. Line 368: If the profiles vary so little with time why present so many panels in Fig. 9, see follow on comment below.

See Reviewer 3 comment 1 above.

12. Line 371: I think the degree sign should be K. You’re using potential temps so the differences are in K, not degrees.

Done. See response to your specific comment 3.

13. Line 378: The sidewall profiles are only very slightly superadiabatic, and please see one of my primary comments above with respect to this point.

See Reviewer 3 comments 2 and 20 above. We have changed the text to mention that the profiles are only slightly super-adiabatic, as suggested. They are nonetheless of meteorological importance, indicating the domain of the cold air intrusions on the slope.

14. Line 382: I don’t understand what the upper dot on each profile is? I understand the NBE dot, showing the crossing of the profiles... can you explain this?

We have removed the upper dots, which were there to emphasize the rim potential temperatures used in the analysis that followed. We agree with the reviewer that the rim temperatures dots are not really necessary.

15. Line 405-408: Again in reference to super-adiabatic layer. Since the profile is not vertical I have a hard time allowing it to be “destabilized” by the mixing process. In a vertical profile I can not envision how (and am not aware of examples of) turbulent mixing producing superadiabatic profiles, rather it tends to mix toward an adiabatic lapse rate. The along slope flow is different, as it is modified during its descent. This does not imply that the air arriving at the lower slopes is unstably stratified above the inversion (i.e. dtheta/dz is not necessarily negative in the vertical... or at least you don’t show that it is). While I appreciate that the psuedovertical profiles have been used in many previous studies, it seems to me that such a slight variation in stability can not be adequately assessed in this manner. That said I think all of this can be addressed with some revised wording and caveats added to the description.

We have revised the wording, as suggested, and have now clearly stated that the pseudo-vertical stability is an along-slope gradient and not to be used to assess vertical atmospheric stability. See the response to your major comments 2 and 3 and comment 20.

16. Line 408-411: I do not follow how these lines follow from the superadiabatic destabilization argument. These seem to be the inverse... perhaps I’m confused but some clarification is required.

This section (now section 5c) has been rewritten in response to comment no. 15. We believe that, when pseudo-vertical measurements are made on slopes that the existence of a pseudo-vertical super-adiabatic layer is a reliable indication that cold air intrusions are occurring on the slope.

17. Fig 9. You show many profiles in Fig. 9, but never really discuss relevant differences or consistencies amongst them. Why show so many? Is there a point you’re trying to make about the variability? It seems like this could be done with just one of these panels, or a longer time mean along with the variance (or standard error at each elevation). Reducing this figure is in line with the general concept that the paper could be more concise.

These panels were designed to give readers an idea of the variation during the analysis period and to illustrate the NBE concept. The mean profiles are already included in the paper (now Fig. 8, was Fig. 14). We have now uploaded an animation of all 5-min profiles on a website for readers so that they can assess the full variability.

18. Fig 10. I find the use of brackets here a bit confusing.

We have removed the brackets.

19. Line 420: It might be useful to indicate to the reader which color line(s) you are discussing in the parenthetical figure reference. (e.g., “the red line in Fig. X”).

The text here refers to all upwind sites rather than to individual sites, so we cannot refer to individual line colors. We were able to improve the caption, however, and there is a legend stating the line colors.

20. Line 428: The “superadiabatic” along slope variation is very minor. How do the magnitude of the temperature differences between the rim and NBE compare to the sensor uncertainty?

We revised the text to state that the pseudo-vertical profiles are only slightly super-adiabatic and that this super-adiabatic profile is a persistent feature of upwind soundings in both METCRAX and METCRAX II. The sensor accuracy is stated in section 3. Data points on the HOBO lines all fall closely on a super-adiabatic line. See our responses to other comments on the super-adiabatic layer.

21. Line 436-447: I’m not sure I understand why the time-height sections are presented. This LNBE approach seems to be well summarized by the time series on Fig. 12d. As the time-heights are not extensively discussed I think it might be wise to simply remove them.

We have now removed the three time-height sub-figures.

22. Line 469-471: Here an on many subsequent points (lines 469-506) you say “often”, “tends to”, and “frequently” but only show snap shots. If there is a persistent horizontal axis eddies, flow intrusion, hydraulic jump type features they should all be apparent in a time-mean or time- varying perspectives of the flow. See major comment above for more on this point.

These terms are now better supported by our animation of the full set of dual-lidar images that is now provided as supplemental information.

23. Line 478: “usually present” Line 479: “typically”

See response 22 above.

24. Line 480: “frequently”Line 485: “tends to”

See response 22 above.

25. Line 486-487: I think I can see the feature you’re indicating in Fig. 13f, but not in 13e.

We now reference only Fig. 13f (now Fig. 13b). Line 528.

26. Line 488: It would be straightforward to actually compute the flow convergence along the lowest radials. How does the convergence vary in time? Is it linked to other observed quantities?

As suggested, we have computed the radial mean flow convergence for the 5-h period (see Fig. B3 below) using the standard convention that flow towards the lidar is negative (thus, flow convergence is negative or red). There is a broad area of convergence (units are m s-1 km-1) above the upper slope where the horizontal flow coming over the rim slows and sinks into the basin. A second convergence zone occurs on the lower slope at ~1825 m MSL (just above the cold pool) associated with the hydraulic jump. Because the hydraulic jump moves up and down the slope, the mean convergence (Fig. B1) is diffused along the slope.

27. Line 544: I think the figure numbers should be 14, not 1

Fixed.

28. Line 559-560: See primary comment about “super-adiabatic”

We have modified the sentence to emphasize that the super-adiabatic gradient is an along-slope (i.e., pseudo-vertical) gradient and shouldn't be interpreted as a vertical stability parameter. See response to comments no. 2 and 20.

29. Line 584-586: This is a bit misleading in that your earlier analysis clearly shows that a portion of the incident flow is blocked by the Rim topography. That portion of the flow is clearly not supercritical.

The reviewer is correct that the flow below the dividing streamline is clearly not super-critical. Our calculations are intended to determine whether the flow approaching the crater above the dividing streamline is super-critical, as it is this flow that comes over the rim. The Lehner et al. (2016a) idealized simulations did not, in general, include a crater rim. This is now noted in the text.

30. Line 595-596: This statement would be better supported if you include additional lidar analysis that shows the mean structure of the jump for the analysis period.

We have now included a mean view of the cross section in new Fig. 13, adding explanatory text to the manuscript. Further information for this reviewer is in Figs. A1-4 below. The mean dual-lidar figures show the upper current overflowing the crater, the wave in this flow, the top part of the cold air intrusion descending the sidewall, the low-wind speed cavity, and the high variance rising motions on the lower slope indicating the hydraulic jump. The non-stationarity of the hydraulic jump de-emphasizes this feature in the mean portrayal. We have not plotted winds or variances for grid points with an inadequate number of samples.

31. Line 633-634: I don’t think you’ve really supported this conclusion. Again, further lidar analysis is required.

The reviewer is correct that we have not shown this. We have a second paper that focuses on warm air intrusions, and this conclusion came from analyses performed for that paper. We have removed the statement.

32. Fig. 15. It seems like the conceptual model figures should indicate that the cavity is considerably warmer and more adiabatically mixed than the downstream crater profiles. (as per Fig. 14).

We have added a sentence to the figure caption indicating the relative temperature and mixing state of the cavity. Line 921-922

REFERENCES:

Smith, R. B., 1989: Hydrostatic airflow over mountains. Adv. Geophys., 31, 1-41.

Zardi, D., and C. D. Whiteman, 2012: Diurnal Mountain Wind Systems. Chapter 2 in: Mountain Weather Research and Forecasting (Chow, F. K., S. F. J. DeWekker, and B. Snyder (Eds.)). Springer, Berlin, 35-119.

ADDITIONAL FIGURES FOR REVIEWER 3:

Figure A. Means and variances of the dual-lidar retrieved winds in the 2-D (u-w) cross section over the 5-hour period, from 5-min retrievals. 1) Mean winds, 2) speed variance, 3) u variance, and 4) w variance.

Figure B. Radial wind means and statistics from single Doppler RHI scans for the 5-hour period, from 2.5-min retrievals (120 possible). 1) Mean radial winds, 2) variance, 3) along-beam speed convergence, and 4) sample size.