the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Observed impact of the GNSS clock data rate on Radio Occultation bending angles for Sentinel-6A and COSMIC-2
Abstract. Space-based Radio Occultation (RO) experiments currently require the tracking of signals from the Global Navigation Satellite System (GNSS) by a Low-Earth-Orbit (LEO) satellite as the signals travel through different layers of the atmosphere. The orbit and clock solutions for the GNSS constellations affect these experiments in two ways: They are needed to obtain a zero-differencing GNSS-based orbit and clock solution for the LEO, and they enter directly the processing of each single radio occultation profile, where the orbit and clock information for the transmitter (GNSS satellite) and receiver (LEO) is required. In this work, we investigate how different GLONASS and GPS orbit and clock solutions affect the statistical properties of RO profiles by comparing our results with forward-modelled bending angle profiles obtained from the European Centre for Medium-Range Weather Forecasts (ECMWF) short-range forecasts. Given that GNSS orbits are relatively smooth, the focus will be on the effect of different transmitter clock data rates, and we tested the range from 1 to 30 seconds. The analysis is based on the reprocessing of Sentinel-6A data (four months in 2021, or about 110k occultations) and of a smaller sample of recent COSMIC-2/FORMOSAT-7 data (about 9k occultations). We find that at impacts heights above about 35 km GLONASS bending angles statistics markedly improve with the use of high-rate clock information. For GPS, not much is gained by using rates higher than 30 s, and the statistics are better for more recent GPS blocks. These results are likely the manifestation of the different short-timescale behaviour of the atomic clocks onboard the GPS and GLONASS constellations.
- Preprint
(2711 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Review of manuscript amt-2024-80', Anonymous Referee #1, 19 Jul 2024
General comments:
The paper describes the impact of using different clock rates in the correction/processing of GNSS observations and retrieval of bending angles (BA) from radio occultation (RO) data. It is shown that using increased clock rates for GLONASS occultations reduces the standard deviation in Sentinel-6 and COSMIC-2 BA relative to ECMWF forecasts at high altitudes. This is an important result for both NWP and climate applications. It is also shown that for GPS occultations, the residual clock noise mostly depends on the GPS transmitter type (and to a smaller extent the used clock rate), with newer GPS Blocks having less clock noise.
The paper is well organized and well written, and I recommend publication with minor revision. My comments below are mostly suggestions for text improvements. One suggestion (see below) goes to providing one example (one occultation) of the impact on the excess phase as a function of time for different clock rates. I think it would be interesting to see this and it would in my view improve the paper, but it is only a suggestion.
Specific comments:
Line 40: I would add "and results are presented in section 4" after "set-up is described in section 3". I know it is implicit when you have parenthesis "(section 4.1)" and "(section 4.2)" in the next sentence, but it becomes more clear with an explicit statement. Also the parenthesis "(section 2)" in the previous sentence could be "as described in section 2" to be explicit. Normally parentheses should just contain clarifying information, not changing the meaning of the sentence if they were to be left out.
Line 58: "forward-modelled bending angle profiles extracted by the European Centre for Medium-Range Weather Forecasts (ECMWF) short-range forecasts" I suppose what is meant here is that the bending angles are forward-modelled based on temperature, pressure, and water vapor profiles extracted from ECMWF short-range forecasts. But that's not what the sentence says. The sentence should be revised.
Line 78: slight -> slightly
Line 114: It is unclear to me what "processing-center-independent cross comparisons" mean. If it was the opposite, i.e., processing-center-dependent cross comparisons, I would maybe understand it as comparisons of solutions from different processing centers. However, strictly speaking you are saying that it is the comparisons that are processing-center-independent. I'm not sure if this is really what you want to say. I have the same issue regarding "SW-" and "receiver-" if the dash refers to "independent". There is also "receiver-independent way" in line 110 that I don't fully understand. Overall, could the sentences be re-formulated to become more clear?
Line 119: I would say "(cf. section 1)" if what you mean is that you already stated something similar. In section 1, the similar statement (line 20-22) refers to both positions and clocks, but here only to clocks. Maybe the reference to section 1 here is not needed at all.
Line 120: "One is obtained using JPL RT GPS products with clocks at 30 seconds and orbits at 15 minutes". I suppose you here downsample to 30 seconds and 15 minutes, respectively. I think you should include that in the sentence, since these numbers are different from those given in the introduction. E.g., ... using JPL RT GPS products, downsampling clocks to 30 seconds and orbits to 15 minutes".
Line 136: fit -> smooth (I suppose that is what you mean).
Line 136: Not sure if this is necessary: "For example, if every 30th data point were an outlier, the 30 s downsampled product would contain little useful information. With this caveat in mind, the direct decimation represents a conservative choice since it could increase the relative percentage of outliers, thus adding to the loss of information in the downsampled products." It seems obvious that this could in principle happen, but is it a real concern? Is there any reason to believe that the fraction of outliers would increase? If there is no real concern, I think these sentences could be removed.
Line 147: a RINEX -> RINEX data
Line 148: Also here "(cf. Section 3.1)". Should it be "Section" or "section"? I'm not sure what the AMT guideline says.
Line 149: illustrated -> mentioned.
Section 4, before Section 4.1: Would it make sense to give one example of excess phase as a function of time over about half a minute at the beginning of a setting occultation where the excess phase is close to zero (with not too much ionosphere) to show how the noise in excess phase is reduced when the clock rate increases? I haven't seen that before, and I think it would be very illustrative. It would help to understand the characteristics of clock errors (and corrections) on the phase data. Excess phase differences using two clock solutions at different rates might also be very illustrative - I have never seen it.
Line 153: "following the approach of Figure 1". Do you mean "similar to Figure 1"? Maybe it is not needed in the sentence.
Figure 5: It is difficult to distinguish between 01s and 30s, as well as between 02s, 05s and 10s (also in the right panel). Maybe choose more distinguishable colors.
Comment to the results in Figure 5: Although there is not much gained in terms of standard deviation using higher clock rates for GPS, there could be a difference in error correlations. This in turn could affect standard deviation in derived refractivity (or in data assimilation of bending angle). I'm not asking for refractivity statistics here, but maybe my suggestion above of giving an example in excess phase could shed some light on this.
Figure 6 caption: hardware -> GPS transmitter types (I suppose), not unique -> not consistent with the other GPS Blocks (or something similar).
Figs 4,5,6,7.8: Discussions of results in the last sentences of the figure captions should be moved to the text (or just removed if already discussed in the text).
Line 177: "there is an improvement in going from Block-IIR, to Block-IIF, to Block-III". But is the picture for a given Block about the same for COSMIC-2 as for S6A?
Line 184: I couldn't find anything about clock noise in (Harnisch et al., 2013). Please check if it is the right citation/reference.
Line 190: Maybe the part on ROMEX needs update, or could a more general statement be made?
Line 194: I couldn't find the word "sweet-spot" in (Kursinski et al., 1997). Maybe this citation is not needed here.
Citation: https://doi.org/10.5194/amt-2024-80-RC1 -
AC1: 'Reply on RC1', Sebastiano Padovan, 21 Nov 2024
The comment was uploaded in the form of a supplement: https://amt.copernicus.org/preprints/amt-2024-80/amt-2024-80-AC1-supplement.pdf
-
AC1: 'Reply on RC1', Sebastiano Padovan, 21 Nov 2024
-
RC2: 'Review of manuscript amt-2024-80', Anonymous Referee #2, 24 Jul 2024
Summary
The present paper discusses the impact of different rates (1 to 30 seconds) from the GPS and GLONASS constellation clocks on radio occultation (RO) bending angle profile statistics for Sentinel-6A and COSMIC-2 RO missions. The authors conclude that higher GNSS clock rates lead to improved bending angle profile statistics with decreased standard deviation in comparison to ECMWF short-range forecast data for GLONASS, while not much is gained by using higher rates than 30 seconds for GPS (with slight differences between different GPS blocks). This is supposedly related to known lower short-term stability of GLONASS clocks.
General comments
Admittedly, the findings in this paper are not entirely new and the authors should include previous work conducted with regard to this topic in their discussion even if from preliminary studies, e.g., Yao et al. 2023 (also reference [1] therein) who investigated the effect of higher-rate GPS and GLONASS clocks with respect to COSMIC-2 in a similar study setup. However, the work presented discusses the issue in more detail than previous studies (at least to my knowledge, the authors are advised to conduct a proper literature review for other related publications) and the authors present the results in a well-structured way providing extensive illustration supporting their argumentation.
In order to strengthen and enhance the leverage of the publication the authors should therefore expand their discussion for the following points. Since the focus of the study is on the impact of GNSS clock rates on RO data processing, it would be of interest to discuss and show how the different clock rates are applied and manifest in the RO excess phase calculation, as the point where GNSS clock data enter the RO processing. Additionally, the applied interpolation method from the GNSS clocks to the high-rate RO measurement time stamps should be included in the paper and possible implications of different interpolation methods, if there are multiple to choose from, should be discussed (in dependence of the clock rate, if relevant.
Yao, Jian, Weiss, Jan-Peter, VanHove, Teresa, "Impacts of High Rate GNSS Satellite Clock Estimation on Radio Occultation Bending Angle Retrievals: Preliminary Report," Proceedings of the 2023 International Technical Meeting of The Institute of Navigation, Long Beach, California, January 2023, pp. 995-1001. https://doi.org/10.33012/2023.18621
Line per line and figure specific comments:
Figures (general): On my printout the graphics are slightly blurred, please provide the figures in higher resolution. Use intermediate minor tick-marks and provide major tick-marks with shorter intervals to support the reader with the identification of relevant values in figures (applies basically to all figures except Figure 2).
L2: “Space-based RO experiments …”. For my understanding this sounds a bit too “experimental”, RO is a proven and well advanced remote-sensing measurement technique, but maybe this is commonly recognized designation. This applies to other occurrences in the text as well.
L2: What is meant by RO experiments “currently” require tracking of signals from GNSS? I suggest to remove “currently”.
L5-6: Radio occultation was already introduced as acronym, use RO acronym here.
L6: Update to “…, where the orbit and clock information for the transmitter (GNSS) and receiver (LEO) satellites is required.”
L9: “… the study focused on the effect”, instead of “…, the focus will be on the effect”. Remove coma after “data rates”.
L10: “… range from 1 to 30 seconds”. State which exact rates were applied if this can be stated generally. This is more informative.
L11: State which four month served as test data period for Sentinel-6A and also which dates served as input for the COSMIC-2 analysis.
L13: Depending on the context 30 seconds can already be considered high-rate compared to e.g., 5 or 15 minutes orbit sampling. Better to use “… higher-rate clock information”. This might apply to other occurrences in the text as well.
L17: I suggest to replace “estimation” by “calculation” and rephrase the following part “… BA profiles based on signals from ...”.
L18: Remove “an” in “requires an accurate”.
L25-26: Unclear language. Rephrase to something like: “Due to the random stochastic noises that affect GNSS atomic clocks, a smaller sampling interval is required to obtain accurate interpolations, ...”
L29: Aren’t the CODE final GNSS orbits provided with 15 min sampling? Please check.
L35-37: This sentence does not seem logical, please state clearly what you are trying to say. Are you saying these stability analyses are “used” for high-rate corrections in order to obtain high-quality BA products. If yes, in which way are they used?
L42: Add verb. “The discussion and conclusions are presented in Section 5”. Makes sure to follow the journal guidelines for the upper/lower case notation of keywords like Section, Figure, etc. and follow them consistently throughout the manuscript.
L45: Better to use past tense: “… was equipped … was built ...”.
Figure 1: Please include OPE and STC in the title of the two figure panels, respectively. This way the reader has a direct connection to the acronyms used in the text. Also add it in “Left: Operational data stream (OPE) ...”
L58: Please correct: “… bending angle profiles are extracted from …”.
L62: Please indicate if you are you using matching occultation events for both processors or if they differ.
L64-67: You mention JPL final clocks used for OPE for GPS. Where have the JPL final products been introduced? I can’t find them in the paragraph from L23-32. Also, why are you using different inputs for OPE: GPS (JPL final 30 s) and GLONASS (JPL RT 1 s)? In my view it would be clearer and more consistent to use the same input for both, GPS and GLONASS. Please clarify.
L71: Did you look for more up to date references? I did not check but Jaduszliwer et al. 2021 and references therein might be a starting point.
Jaduszliwer, B., Camparo, J. Past, present and future of atomic clocks for GNSS. GPS Solut. 25 , 27 (2021). https://doi.org/10.1007/s10291-020-01059-x
L80: Rephrase “… described in the following are based on the re-processing of RO data …”.
L81: Please provide more details which RO receiver type is flown on S6A and COSMIC-2 and provide a reference.
L84: Is there any reference for the YAROS software?
Figure 3: Please add a grid and intermediate tick-marks on the y-axis to support the viewer. Add “Sentinel-6A” to the title.
L96-98: So BSW5.2 was used for OPE processing and BSW5.4 for STC processing? Please clarify.
L103: Remove punctuation: “See, e.g, …”.
L104: Plural: “… the reduced-dynamic and the kinematic orbits …”.
L104: What exactly do you mean by the data handling component? The same software using the same processing setup? If so, please revise to improve clarity.
L112: What are the different characteristics of the two oscillators?
L115: You state that typically the spread of the different orbit solutions is below 3 cm 3D-RMS. Did you check this for the investigated time periods in this paper as well? Did you find any noteworthy deviations? It would be a valuable addition for the reader and further improve the manuscript to add a time-series plot of the 3D-RMS of the different comparisons for Sentinel-6A and the same time period as in Figure 3.
L117: Better: “… (cf. Figure 1) …”.
L120: In the paragraph from L34-32 you introduce JPL RT products with 1 min orbit and 1 s clock sampling but here you are using JPL RT products with 15 min orbit and 30 s clock sampling? Also, since it is hard to keep track which GNSS product is provided with which orbit and clock sampling rates one could consider to collect this information in a concise table at the beginning of the manuscript.
L124: This has been shown by Fernandez et al., 2024? If yes, make this more clear by saying “… it has been shown by Fernandez et al 2024 that not much is gained in this case”.
L127: Provide proper figure reference: “It is clear from Figure 3 that …”
L133: Again I am confused by 15 min orbit sampling here and the 1 min orbit sampling of JPL RT data at the beginning of the paper.
L135-139: You state that using a fit would be more stable against outliers and therefore the better option, still you use straightforward decimation. What is the estimated difference between the two approaches and can you confirm that your choice does not impair the results?
L141: “… obtained by”
L146: Doesn’t Sentinel-6A track GPS and Galileo on the POD antenna?
L149: Better: “… as discussed in Section 3.2.”
Figure 4: Please include the time period of the underlying data in the figure caption. Also in Figure 5, 7, and 8.
L160: Add the height range to the text for convenience: “… at high impact heights between 50 and 60 km”
L176: You limit the illustration of COSMIC-2 GPS occultations to 30 s clock products in Figure 8. For consistency and to underline your findings it would be interesting to include the full range from 1 s to 30 s clock rates, in the way how it was done for Sentinel-6A or at least state that the COSMIC-2 analysis shows similar characteristics if this is true.
L180: Best performance in terms of standard deviation of what? Remind the reader that your analysis is based on bending angle statistics here.
L186: What is meant by “real” RO observations, are there any other? I suggest to remove “real”.
Figure 6: Why does the standard deviation for different clock rates reach different impact heights? E.g., in the left panel (Block-IIR) the green line (5 s) reaches about 58.7 km while the blue line (10 s) ends at 58 km?
L191: It would have completed the picture to include Galileo and Beidou occultations in this study. Nevertheless, do you have any expectation on the behavior of the use of higher rate clock data from those constellations and their impact on RO bending angle statistics, given their clock stability?
L193-195: You mention the RO sweet-spot down to 5 km, but what about the increased standard deviation shown in your figures at these altitudes?
L202: Acknowledgements
L221: Remove repetition of “https://doi.org/” and what is “112 395”, seems odd.
L223: Remove repetition of “https://doi.org/”.
L241: Remove repetition of “https://doi.org/”.
L257: Remove repetition of “https://doi.org/”.
L259: Add DOI.
L266: Add DOI. This is still in “Atmos. Meas. Tech. Discuss.”, update to published paper if available.
L273: Is there a online resource available?
L275: Remove redundant “2019” before DOI.
Citation: https://doi.org/10.5194/amt-2024-80-RC2 -
AC2: 'Reply on RC2', Sebastiano Padovan, 21 Nov 2024
The comment was uploaded in the form of a supplement: https://amt.copernicus.org/preprints/amt-2024-80/amt-2024-80-AC2-supplement.pdf
-
AC2: 'Reply on RC2', Sebastiano Padovan, 21 Nov 2024
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
218 | 56 | 21 | 295 | 15 | 17 |
- HTML: 218
- PDF: 56
- XML: 21
- Total: 295
- BibTeX: 15
- EndNote: 17
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1