Identifying 2010 Xynthia Storm Signature in GNSS-R-Based Tide Records

Round 1

Reviewer 1 Report

From my application-oriented standpoint this novel application of GNSS is very promising and opens up new opportunities for multidisciplinary research. The paper is interesting to a wide audience, from GNSS experts to climate change and disasters experts.

The quality of presentation would be high, but it loses some of its overall merit due to language issues, although the main messages are clear enough.

Therefore extensive editing of English language and style is required throughout the text. Here are some indicative language editing remarks in the abstract, just to make the point (but there are many more grammar, syntax and vocabulary corrections to be made in the main text):

§  Line 21: Please replace “…for the monitoring changes…” with “…for the monitoring of changes…”

§  Line 22: Please rephrase “Thanks to the deployment in many countries of permanent GNSS networks” to “Owing to the deployment of permanent GNSS networks in many countries”

§  Line 26: Please rephrase “GNSS geodetic” to “geodetic GNSS”

§  Line 27: there is one extra space after the parenthesis. The correct should be “…RGP (Réseau…”

§  Line 33: Please rephrase  “are able blindly separate” to “are able to blindly separate”

Author Response

Please find our responses to your valuable comments in the enclosed file.

Author Response File: Author Response.docx

Reviewer 2 Report

The manuscript attempts to demonstrate an interesting original application, namely, the detection of storm surges with GNSS reflectometry (GNSS-R). More specifically, with ground-based GNSS-R using signal-to-noise ratio (SNR) interference pattern technique (IPT).

Unfortunately, the demonstration fails to convince.  For example, if we accept authors' statement that "surge is defined as the difference between observed and astronomical modeled tides", then the main result is the detided residual sea-surface height (SSH) time series, buried in Fig. 4 as the red lines.  While the tide gauge (TG) residual series exhibits a clear anomaly in the first half of the Xynthia storm window (Fig.4 top panel), the GNSS-R residual series shows no sign of the storm.

What we see in the rest of the manuscript is a miscellany of techniques being attempted haphazardly: Singular Spectrum Analysis, followed by a moving average and tidal analysis, then Continuous Wavelet Transform and another band filtering.  There is a strong result buried in Fig.9, that the performance of GNSS-R depends on the period considered, with significant deviations from the ideal 1:1 diagonal lines at lower frequencies, but authors seemed to prefer not emphasize it in the text.

Therefore, I recommend rejection and resubmission. I'd suggest authors to reframe the work as a more neutral assessment of the strong and weak aspects of GNSS-R.  Below I point moderate and minor issues.

-------------------

MODERATE

The abstract should have no paragraph breaks.

The article needs tables:

- Section 5.1: a table summarizing the statistics would be valuable.

- Fig.9: statistics printed on the image should be extracted and moved into a proper table.

It is unclear how authors associated SSA modes to semi-diurnal tides (RC1), diurnal tides (RC2), inverted barometer effect and wind (RC3), and noise (RC4). Is that defined a priori, from the SSA definition, or is it an a posteriori conclusion reached inspecting the SSA results?

Could the more complicated SSA and iCWT be replaced for a simpler moving average with the different window widths applied to the raw original sea surface height series?

Normally tidal analysis is performed on the raw data; I see no valid reason for performing the tidal analysis on the SSA modes.

More details are necessary about the SNR data processing, such as which SNR signals (carriers and codes) were used, what was the update interval for height retrievals, the azimuthal and elevation mask, etc.

Section 3.1. please state how high is the GNSS antenna above mean sea level.

Section 3.2: the horizontal distance between GNSS and tide gauge must be mentioned in the paper.

Conclusions state that "GNSS-R approach is able to estimate SSH with a similar accuracy as the tide gauges", but the RMSE values reported (20-30 cm) is one order of magnitude worse than what conventional tide gauges are able to deliver.

Authors further conclude that "main additional value [of GNSS-R] is its ability for storm detection", and allude to elsewhere in the text that "classical tide gauges are dampened to limit the effect of the waves"; this is an overstatement, as modern tide gauges such as radars can be configured to sample more or less rapidly.

Conclusions should summarize the main results more quantitatively.

The discussion on future work included in Conclusions should focus on the technique employed in the manuscript, namely, ground-based GNSS-R using SNR IPT; comments about space-based platforms and more advanced correlation waveform are planted but really do not belong here, as they do not follow from any result in the present work. I recommend removing these unrelated extrapolations or at the very least segregating them into a separate paragraph.

---------------------

MINOR

5.1. SSH_GNSS time series -> GNSS SSH time series

Fig. 9: x- and y-axis labels could be "Tide gauge SSH (m)" and "GNSS-R SSH (m)".

As equation for the inverted barometer effect would benefit the reader.

Figure 1.b: a better photograph of the building surroundings, including obstructions, is necessary.

Introduction third paragraph (l.53-77): split in two, speaking of GNSS-R only in the second part.

p.1, l.18: Sea surface height -> Sea surface height (SSH)

p.2, l.49: They trend -> They tend

p.2, l.53: strongly storms -> strong storms

p.2, l.71: mostly used -> most used

p.4, l.131: direct and multipath and signals -> direct and reflected signals [multipath refers to the combination of multiple paths, not just to the reflected path]

Eq.(1): SNR should not be squared on the left-hand side.

Eq.(3): h_eff could be simplified to h, since there is not ambiguity.

Eq.(4), especially the height-rate correction, was first published in 2013, at <http://doi.org/10.1109/LGRS.2012.2236075>.

p.4, l.150: citation to [3] seems mistaken.

After most equations, there should be no paragraph identation, especially when the sentence is not interrupted.

p.12, l.337: beneficiate -> benefit

p.12, l.338: the oldest -> one of the oldest

p.13, l.353: but the time - please rewrite.

A citation is needed after "3CAT-2 satellite".

Author Response

Please find our responses to your valuable comments in the enclosed file.

Author Response File: Author Response.docx

Reviewer 3 Report

RemoteSensing-446052, my Review

Identifying 2010 Cynthia storm signature in GNN-R based tide records
P. Lan Vu et al.

Context & General comments
The question of using the GNSS-R signals in series of geodetic GNSS receiver/antenna located along the coast
is becoming more and more interesting and pertinent, even on the long term and, above all, to detect storm signature.

There is a growing interest and many papers have been published on coastal flooding. However, several questions
are still open today in terms of technology(ies) as well as modeling. The GNSS-R technique in addition to GNSS
receiver/antenna correctly located along the coast should be an opportunity to detect as early as possible the effect
of a storm or a tsunami.

It is true that the west part of France is equipped with a robust GNSS network (GPS at its beginning and NOT GNSS)
which is used by different applications (solid earth, sea level, tides and loading effects). The present paper is thus
perfectly fitted to investigate the GNSS-R signal instead of (or in addition to)

Abstract: I think the texte is too long; things should start at Ligne 26.

1.Introduction
This part is very interesting from the point of view of natural phenomena. The number of references is important.
Nevertheless at the end of the Introduction, the text is a bit confusing concerning both the long term and short term
sea level monitoring (L69-L77, P 2). The authors should improve this part; in particular on the long term, the respective
role of tide gauges, of GNSS networks (or DORIS) and the reflectometry technique should be clearly stated from
past and current result.

At the end, the reader is not really driven by the classical description of the paper content; sections, sub-sections, etc.


2. Study area
OK

3. Datasets
The location of Socoa (latitude and longitude) is not the same between Line 104 and L114 ?


4. Methods
This part is, for me, very clear with the idea of explaining the inversion technique as well as the SSA and wavelet transform.
It is probably the main part of this paper considering its role in opening a new way/approach with GNSS-R signals.  


5. Results
On the contrary, this section is a bit confusing; it’s like a warm result to be described without taking enough time to separate
things (what is important or not). It is particularly true in subsection 5.1 concerning analysis from the SSA method. At this
stage, authors are using to much acronyms; whereas the figures a very welcome, the number of details (e.g. Page 8) do not
give the necessary fluidity in the text and thus a comprehensive way.

But subsections 5.2 and 5.3 being shorter are much better.  

Subsection 5.4 is not as optimistic as the introduction and Conclusion; it gives the idea to the reader of a great complexity
to detect the Xinthia storm; the number of frequencies which are possible to detect or not, the addition of a combined
approach based on both SSA and iCWT.

So, I recommend to strongly improve section 5. (Results) in order to be clear enough and to permit further approaches
and future developments (including the deployment of other antenna along coastal areas (altitude, distance, etc.).
 

7. Conclusions
Same remark as before; there is a gap between the results, as stated by subsections 5.1 to 5.4 and the Conclusion
(L335 to L340).
They are many reasons to be excited about GNSS-R, but authors should be more realistic in their results knowing the
difficulties to detect very small signatures (atmosphere, sea level, etc.).

Author Response

Please find our responses to your valuable comments in the enclosed file.

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

Authors have added significantly new material, including whole new text sections and recalculated some results.  Especially the new Discussion subsections are very welcome.  I am surprised that the residual GNSS-R SSH changed so much, between the old Fig.4 and the new Fig.6 (red line in center panel).  I am also appalled by how different the tidal fit is between GNSS-R and tide gauge (blue lines in Fig.6), as Fig.7a shows good results for a direct comparison between raw GNSS-R and tide gauge.  Considering the point that the wave-protected tide gauge is not supposed to record the storm signal well, authors go on to try to model the storm combining the tide-gauge residual (that they call surge) and atmospheric pressure (IB).  In parallel, authors try to decompose the GNSS-R SSH series, to find some component that correlates well with the storm model.  Alas, a SSA component (RC3) was found to match reasonably well the storm signal, although it is unclear how much tuning is involved in the SSA.  I still find the whole exercise a bit of data massaging (or data torturing), but I guess authors did manage to demonstrate it is possible to extract a storm signature from GNSS-R, at least when you know what you are looking for.  It remains to be seen how the SSA "blind extraction" can be generalized to other events and sites.

Reviewer 3 Report

Thank you for aswering my questions and comments. In particulkar, Section 5 (Results) is know much better.