the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 08 Sep 2026
GNSS Reflectometry on the McMurdo Ice Shelf
Abstract. The ability to monitor glaciated surface types, including snow, ice and meltwater, at a high spatial resolution, is critical for monitoring overall glacier mass balance. Using an Antarctic ice shelf as a study site, we tested the feasibility of Global Navigation Satellite System reflectometry (GNSS-R) and GNSS interferometric reflectometry (GNSS-IR) in monitoring glacier surfaces. During the 2023–24 Antarctic austral summer, we conducted a field campaign to collect GNSS-R front-end samples from one 9-m tower and GNSS-IR samples from one 1.5-m tower offset 14 m away horizontally, at each of two sites located on the McMurdo Ice Shelf near McMurdo Station, Antarctica. The towers near Phoenix airfield were on snow-covered ice. The towers near McMurdo's former Pegasus airfield were on a heterogeneous surface of bare ice and snow-covered ice, which varied temporally. From mid-November to early December 2023, we collected GNSS-R, GNSS-IR, camera imagery, and lidar data of the glacier surface within a 20 m radius at each of the two 9-m tower sites. All data were returned to the lab for post-processing.
We show our results processing the GNSS-R front-end samples to estimate the surface reflectivity from the signal-to-noise ratio. We find that the GNSS-R signal-to-noise ratio is insufficiently strong for estimating surface reflectivity. We also show our results processing the Earthscope geodetic receiver data for interferometric reflectometry, looking at the spectral peak amplitude as the indicator of surface type. We find that the GNSS-IR spectral peak amplitude mean and bias differ between the two sites, but cannot rule out the role that hardware differences between the sites may have in this.
- Preprint
(28278 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 18 Nov 2026)
- RC1: 'Comment on egusphere-2026-5057', Anonymous Referee #1, 24 Oct 2026 reply
-
RC2: 'Comment on egusphere-2026-5057', Anonymous Referee #2, 26 Oct 2026
reply
Dear authors and editor, thanks for letting me review this paper. I considered a really top advance to understand which pioneer method combining GNSS-R and GNSS-IR on the McMurdo Ice Shelf can be more approppriate. However, the current draft has some structural improvements, excessive low-readability figures, and a lack of clear separation between methodologies and results.
- Reorganize the manuscript to clearly separate the GNSS-R and GNSS-IR workflows, methodologies, and results. Remove distracting operational field logs (e.g., minute-by-minute camera anomalies) from the main text.
- I would move secondary camera projections, detailed tower schematics, and operational flowchart boxes to the supplementary material.
- Improve the legibility, formatting, and axis label sizes of key plots (e.g., Figures 10, 11, 13–15, 17–20) to meet high-impact journal standards.
- Explicitly emphasize in the Discussion and Conclusions that differences in spectral peak amplitudes (SPA) between sites (Phoenix vs. Pegasus) may be confounded by distinct receiver-antenna hardware (Septentrio vs. Trimble), rather than solely environmental surface variability.
- Check more comments in my attached pdf.
Data sets
Antarctic GNSS Reflectometry 2023 (D-556) Datta-Barua and Banwell https://doi.org/10.7283/61M5-8Z10
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 354 | 179 | 59 | 592 | 60 | 70 |
- HTML: 354
- PDF: 179
- XML: 59
- Total: 592
- BibTeX: 60
- EndNote: 70
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Overall recommendation
Major Revision
The experimental dataset is potentially valuable, and the publication of inconclusive GNSS-R results can give useful information for future reflectometry experiments. However, I believe additional validation of the receiving chain and acquisition methodology is necessary.
General comments
This paper presents an experimental investigation of GNSS reflectometry (GNSS-R) and GNSS interferometric reflectometry (GNSS-IR) for monitoring 2 points : snow and ice surface conditions on the McMurdo Ice Shelf in Antarctica.
The experimental campaign is interesting, and I can appreciate the effort involved in deploying several instruments in such a challenging environment. The combination of GNSS measurements with camera imagery and LiDAR observations is particularly valuable and could provide an interesting dataset to study the response of GNSS signals to different cryospheric surfaces.
However, I have several concerns regarding the instrumental configuration, the processing methodology and the interpretation of the results.
My main concern is related to the GNSS-R experiment. The authors were unable to reliably detect reflected GPS signals, despite using a dedicated LHCP antenna, significant amplification and coherent integration. Considering the presence of exposed ice and the relatively favorable geometry of some satellites, I find this result really surprising.
It is always hard to review an unsuccessufl experiment but I am convinced it can certainly provide useful scientific information. However, before concluding that reflected signals are too weak to be detected, I think it is important to demonstrate that the complete receiving chain and the acquisition algorithm were functioning correctly.
Regarding GNSS-IR, the authors obtain spectral peak amplitude estimates and identify differences between the two sites. However, the use of different receiver and antenna models makes it difficult to determine whether these differences are related to surface conditions or instrumental effects. It could be interesting to use similar experiment if we want to conclude.
I believe this work has a very high potential, particularly because of the experimental dataset collected, but several technical points should be clarified or further investigated before publication.
Specific comments 1. GNSS-R receiving chain and signal detection
My main concern and doubt is the complete absence of reliable GPS signal acquisition from the reflected channel.
The authors seems to use an Antcom LHCP antenna connected to an Ettus USRP N210, with 40 dB of inline gain and 31 dB of RF gain. Coherent integration was performed over 10 ms.
Under these conditions, I would have expected at least some reflected GPS signals to be detectable, particularly over relatively smooth ice surfaces and for satellites at high elevation angles.
Have the authors estimated the expected received signal power or carrier-to-noise density ratio (C/N0) for this configuration?
Even a simplified link budget, considering the antenna gain, receiver noise, cable losses, reflection coefficient and observation geometry, would help determine whether the reflected signals were expected to be detectable.
I also wonder whether the acquisition chain was validated in laboratory independently, using a known GPS signal, before or after the Antarctic campaign. An easy solution is simply to use the incident signal in the receiver channel. Are you sure about the experimental setup ?
This seems particularly important since the main conclusion of the GNSS-R experiment is based on the inability to reliably acquire reflected signals. We have to remove problems coming from hardware.
2. RF filtering and receiver performance
I have some questions regarding the RF filtering used in the receiving chain.
From my experience with GNSS instrumentation, placing a narrow-band cavity filter directly after the antenna can make a significant difference in reception performance, particularly when relatively high amplification gains are used and out-of-band interference is present.
The authors mention a 40 dB inline amplifier and 31 dB of USRP RF gain, but I could not clearly identify whether a dedicated band-pass filter was installed before the amplification stage.
Was the input spectrum examined for possible interference, receiver saturation ?
It would be useful to provide more information about the RF front-end, including the filter characteristics, amplifier noise figure and gain distribution.
A properly selected band-pass filter could help reject out-of-band interference and will improve receiver performance, although its insertion losses and bandwidth must also be considered.
I think this point deserves attention before suggesting additional amplification as a possible solution.
3. GNSS-R antenna radiation pattern and polarization
I think the radiation pattern of the downward-looking LHCP antenna should be provided in the manuscript. It is a missing information which can explain the absence of signal.
The antenna gain depends on the direction of arrival of the reflected signal, and therefore on the satellite elevation angle.
Without the antenna radiation pattern, it is difficult to determine whether the expected reflected signals were received within the main lobe and with sufficient gain. From my experiment, Antcom had a non uniform radiation pattern. For this particular experiment, it would have been useful to use maybe a specific antenna.
Then, I strongly recommend providing the antenna gain pattern at GPS L1 for the LHCP port.
The cross-polarization discrimination is also important, since insufficient rejection of RHCP signals could lead to unwanted reception of direct GPS signals. I do not know if this antenna has two ports but it could be also given if you have these data.
It would be particularly interesting to relate the expected reflection directions of the selected satellites to the antenna radiation pattern.
This information could help determine whether antenna directivity or polarization losses contributed to the unsuccessful acquisition.
4. Use of direct GPS observations to assist reflected signal acquisition
I have some questions regarding the GNSS-R acquisition strategy.
From my understanding of this experimental setup, only the reflected LHCP signal was recorded as raw I/Q samples by the USRP, while the upward-looking RHCP antenna was connected to the Trimble receiver and the USRP GPSDO.
In our previous GNSS-R experiments, we used the direct GPS signal to determine the Doppler frequency and code delay before searching for the reflected signal. This allowed us to considerably reduce the delay-Doppler search space, which is particularly useful when dealing with weak reflected signals. IT is not the case here. The acquisition strategy should be more complete for this type of experiment
5. GNSS-R coherent and incoherent integration
It seems clear to me that no GPS signal is present in the acquired data. I assume that the authors have extensive experience with GNSS signal processing and are using a well-established processing chain. Therefore, I have no particular concerns regarding the software or processing methods employed.
However, using an independent tool such as GNSS-SDR could also be helpful to verify that the absence of a detectable signal is not caused by an issue in the processing software itself.
6. Interpretation of GNSS-R SNR and surface reflectivity
The manuscript discusses the relationship between the SNR calculated from the acquisition maps and surface reflectivity.
However, the acquisition SNR depends on several instrumental and processing parameters, including antenna gain, receiver noise, integration time and observation geometry.
In the absence of a clearly identified reflected signal, I think the calculated SNR should primarily be considered an acquisition diagnostic rather than a direct indicator of surface reflectivity.
This distinction should be more clearly explained throughout the manuscript.
7. Representativeness of the GNSS-R measurements
I find it surprising that the analysis is based on only two weeks of data. Why was the analysis limited to such a relatively short period? A longer observation period would have been valuable for assessing the performance and reliability of the system.
it might have been always useful to process and analyze the data while the experiment was still running (something like in real time just to ensure everything is ok). This would have allowed the authors to identify potential issues at an early stage and, if necessary, adjust the experimental setup or acquisition strategy accordingly. No information is given about that.
8. Potential contribution of other GNSS constellations
The GNSS-R processing for this paper is restricted to GPS L1 C/A signals.
I understand this choice, but considering the difficulties encountered with signal acquisition, it could be interesting to discuss the possible contribution of additional GNSS constellations.
For example, Galileo E1 could provide additional satellites and potentially more favorable reflection geometries. Furthermore, Galileo pilot signals could offer opportunities for longer coherent integration, provided that an appropriate processing strategy is implemented.
I imagine that the raw USRP data recorded with sufficient bandwidth to include Galileo E1 signals. Why did the authors limit their analysis to the GPS constellation? Including other GNSS constellations, such as Galileo, GLONASS, or BeiDou, could have increased the number of available satellite signals and potentially improved the chances of detecting reflected signals. It would be useful for the authors to explain why only GPS satellites were considered in this study.
If so, this could represent an interesting possibility for future analysis of the existing dataset.
9. GNSS-IR comparison between sites
The GNSS-IR results show differences in spectral peak amplitude between Pegasus and Phoenix.
However, the two stations use different antenna and receiver models, with Trimble equipment at Pegasus and Septentrio equipment at Phoenix.
As the authors already mention, these differences may influence the measured amplitudes independently of the surface properties.
I wonder whether it would be more appropriate to focus on relative temporal variations within each station rather than comparing absolute spectral peak amplitudes between sites.
I would also appreciate additional information about the number of observations retained after quality filtering and the influence of the selected azimuth sectors and satellite geometries.
These aspects are important for assessing the robustness of the comparison.
10. Relationship with camera and LiDAR measurements
The camera imagery and LiDAR observations represent an interesting aspect of the experiment, but I feel they could be used more extensively in the analysis.
However, since this is not the main focus of the paper, I do not consider this section particularly useful at this stage. In my opinion, the priority should be to understand why no signal was detected. This paper has to give a reason.
If this issue has been clearly identified , explained and resolved, the authors could then consider whether these data combining camera and LIDAR can be exploited.
11. Interpretation of temporal variations and conclusions
One interesting observation is the change in spectral peak amplitude following the modification of the antenna height at Pegasus.
This clearly demonstrates that the estimated amplitude is sensitive to the measurement geometry.
I think this point deserves more discussion, since other amplitude variations observed during the experiment might not necessarily be related to changes in surface conditions.
It would be useful to better distinguish between variations caused by known instrumental modifications, possible environmental changes and variations that remain unexplained.
Overall, I would encourage the authors to be more cautious in their conclusions and to better separate instrumental effects from possible geophysical signals.
The study demonstrates that GNSS-IR measurements can be collected and processed in this environment, but does not yet establish a reliable method for discriminating between snow and ice surface types using spectral peak amplitude.
Technical corrections
Overall recommendation
Major Revision
The experimental dataset is potentially valuable, and the publication of inconclusive GNSS-R results can provide useful information for future reflectometry experiments. However, I believe additional validation of the receiving chain and acquisition methodology is necessary, together with a more cautious interpretation of the GNSS-IR observations.