General comments:
My comments below relates to either new text in the revision or to replies to comments in the first review. There are several places where there are differences between the tracked changes provided by the authors and the revised manuscript. Also, a new figure (Figure 3) was added together with a new paragraph discussing it, but the new text was not marked as new in the tracked changes. I ended up not relying on the tracked changes. Comments below refer to what is in the revised manuscript.
Specific comments:
L6-8: I urge the authors to remove the claim in the abstract that the data are 'high-quality' and that the variability is comparable to spaceborne RO missions. In their answer (2a) to my comment on this in the first review, the authors refer to the paper by Schreiner et al. (2020) and write: "Between 10 and 20 km, the COSMIC-2 refractivity difference standard deviation was shown to be approximately 1%, slightly smaller than our observed variability estimates." However, 1% is not 'slightly smaller' than 2.3%. The authors also write in their answer: "Given the size of the receiver and the payload being constructed of commercially available parts, we consider this to be high-quality." Whether something is 'high-quality' cannot depend on how it comes about. There is a very recent paper by Cao et al. (2022) (https://acp.copernicus.org/articles/22/15379/2022/) showing that BRO can provide refractivity profiles with a standard deviation less than 1% between 10 and 15 km, in comparison to ERA5 (their Figure 8). The fact that it is not possible to provide this kind of quality with commercially available parts (at least not in this study) is an important message to get across. In any case, the abstract should objectively state the results of the study.
L39: Perhaps missing an 'and' before "by fuel range of the aircraft."
L124 (and many other places): I think the word 'delay' refers to the time delay of a signal, but for the phase the word 'excess' seems more appropriate (the signal is delayed in time, but there is an excess in the phase due to the neutral atmospheric influence). Please consider to remove 'delay' when talking about the phase.
L142: Like the other reviewer, I find it relevant to discuss the impact of the POD uncertainty for BRO in step (b). In their answer (3a, bottom of page 9) to the other reviewer the authors write that "a brief discussion was added to reflect the concerns of the POD uncertainty.". But where? I don't see it.
L172-177: It is still unclear to me how many ERA5 profiles are used. The authors answer (15a) to my comment on this in the first review, and their revised manuscript, did not help. In their answer they write: "... median profile used to run the initial ROSAP simulation, run the initial Forward Abel Integrator, and determine the time series of refractivity at the receiver during the occultation.". So one profile for all this. But in the revised manuscript it says: "... one referencing refractivity profile .... at the zero-elevation ... used to compute the time series of refractivity at the receiver.... Furthermore ... we use a median refractivity profile ... surrounding the zero-elevation TP location for input into the initial ROSAP and FAI simulations." Thus two profiles as I read it, one at zero elevation TP to determine the time series of refractivity at the receiver, and another one (taking the median of grid points surrounding the zero-elevation TP) for ROSAP and FAI simulations. If their answer to me is correct, then the text needs to be updated to make it clear that it is the same profile used for these three tasks. I understand that a second (or third) profile at the 5 km TP location is used for comparisons because the lower TPs are drifting away from the zero-elevation TP location.
Figure 5: I believe the dN (d for deci) in the x-axis label is the wrong unit. In the metric system 20 dm = 0.2 m. Thus 20 dN-units = 0.2 N-units. As mentioned in my first review, I think it should be deca N-units, so that 20 deca N-units = 200 N-units.
Figure 5: The dashed lines need to be mentioned in the Figure caption. In their answer (16a) to my comment on this in the first review the authors write: "The dashed lines indicate the boundary layer height detected with the gradient method based on temperature, specific humidity, and refractivity (Nelson et al., 2021; Ao et al., 2012, 2008; Winning et al., 2017; Xie et al., 2006).". However, these papers are not referred to in the manuscript, where it just says (L189) "... weak gradients in specific humidity, and refractivity ...", which is not very understandable. I cannot see these weak gradients at 0.9 km in particular. Please mention the gradient method and the references in the paper. It is still not clear why there are two lines if they are based on temperature, specific humidity, and refractivity. Please make that clear in the paper. In the Figure caption it could say something like "The dashed lines at approximately 0.9 km indicate the PBLH (see text for details)."
L191-195: Please mention the sampling rate of the observations here or earlier in the paper. Are the simulations done at the same sampling rate? In their answer (19a), the authors tell that the simulations has now been run through the same smoothing process as the observed excess phase data to remove high-frequency variability. This needs to be told in the paper.
Figure 6: The ROSAP results look much better now. Not only is the high-frequency noise removed, but it also seems that the excess phase has come much closer to the calibrated observations than in the original figure (where the two curves crossed near 1000 sec). What is the reason for the latter? I don't suppose the smoothing can change the excess phase like that.
L199-202: I don't want to debate whether 141 and 700 is of the same order of magnitude as claimed by the authors in their answer (20a), and in the manuscript. To me the text is still misleading when it says: "The overall mean SNR from the GROOT receiver (141.79 V/V) is on the same order of magnitude (order of 100) as the mean SNR from the COSMIC-1 and SAC-C GNSS RO satellite missions (approximately 700 V/V ...)". I would rather say something like: "The overall mean SNR from the GROOT receiver is 141.79 V/V. This is relatively small compared to the mean SNR from the COSMIC-1 and SAC-C GNSS RO satellite missions, which is approximately 700 V/V." However, I think the COSMIC-1 and SAC-C SNR's are a bit smaller than 700 V/V due to defocusing at comparable tangent point altitudes (which could be relevant when you do such SNR comparisons). Please check. On the other hand, since the signal only passes through about half of the atmosphere in BRO relative to spaceborne RO, the defocusing effect in BRO will be smaller. This makes SNR comparison to spaceborne RO tricky. And how about the L2 signal? These discussions on SNR seems to focus only on L1, but the data quality also depends on the L2 signal. It should be mentioned in the paper that the discussion is about the L1 signal. It would be interesting to see also the SNR for the L2 signal in Fig. 6c. Please consider to add this.
Figure 7: How is the ROSAP bending angle calculated? From simulated excess Doppler? Or ray traced bending angle? The authors answer (22a) to this question did not help to clarify this. Please clarify in the paper.
L211-214: "Differences between the retrievals and the ROSAP simulation between ~6 and 8 km are most likely caused by differences between oblate and spherical Earth geometry assumptions. While we perform an Earth oblateness correction as part of the transmitter/receiver geometry processing, these effects may not be completely removed.". I very much doubt that to be the case. Why would the oblateness create something so distinct between 6 and 8 km, and then even after the oblateness correction? Is there a reason why you don't think this could be due to variations in the real atmosphere? Or could such variations be caused by yaw instability? Or ionospheric effects not adequately removed? Please revise the text unless you have substantial evidence (if so, please provide it) suggesting that this comes from the oblateness correction (you could just remove these new sentences if you don't have a likely explanation, and say that the variations are not understood). There are also differences around 11-12 km that should be discussed if these around 6-8 km are discussed.
L225-230: In their answer (24a) the authors explain that the median and MAD values referred to in the text are calculated over certain altitude regions: "As these are individual profiles, the median 'above 10 km' is a single value calculated from all data points above 10 km as well as within other regions.". But this is not explained in the revised text. Please do that if you prefer to discuss these median and MAD values. Later on (in Section 4) the median and MAD values are based on ensembles of profiles, which is quite different. Please make that clear. Alternatively (which I think would be much better), the discussion of median and MAD values based on the results in Fig. 8b in Section 3 could be omitted. It doesn't seem to add any useful information.
L249-250: "The ZPM-1 refractivity differences from both retrieval methods show much more variability across all heights in both the median profile and the individual refractivity profiles." I disagree with this statement regarding the individual profiles. I see less variability in the grey lines in Fig. 10 than in Fig. 9. The document with tracked changes, and the authors answer (28a), mention "oscillations" that are not mentioned in the revised manuscript. Still, it is not clear what "oscillations" versus "variability" mean in this context. I think it would be better to acknowledge that there is less variability in the individual ZPM-1 results. The larger variability in the median seems to be a result of the lower number of profiles. In any case, there are so few profiles so that the results are likely not statistically significant.
L258: "Reasons for errors ... can come from ... causes." Sentence does not make much sense. Instead you could say: "Errors ... can come from ... causes".
L278: In this sentence the authors added 'and temporal': "The added benefit of using BRO platforms is the dense spatial and temporal sampling available due to the low platform velocities relative to the LEO-based RO satellites." It can be difficult to understand why the low platform velocities would increase the temporal sampling, but perhaps replacing the word 'available' with 'over targeted regions' in this sentence would more clearly describe what I think the authors mean.
Figure A1: I still don't see any distinction between Piksi 1 and Piksi 2 in the text. In their answer (33a) the authors explain it nicely. Please consider to put that explanation in the Appendix. |
