Mapping the Antarctic Grounding Zone from ICESat-2 Laser Altimetry

This paper uses ICESat-2 data to identify the grounding zone of the Larsen C Ice Shelf, Antarctica. It builds off of methods developed for ICESat and automates some of these processes. The authors describe a modified repeat-track analysis (that uses 6.5 months of ICESat-2 data) and a crossover analysis (that uses 1 year of data) to determine a GZ that is in close proximity to the continuous GL assessment from the DInSAR GL of Rignot et al., 2016; this agreement is heavily based on the increased resolution of the ICESat-2 data, and not a measure of precision, as the authors contest.


Lines 136-160 -In
the authors show the position of Ib, the ice-sheet-shelf inflexion point (otherwise known as the break-in-slope), which is obtainable from optical-, DEM-and altimetry-based methods. As the authors are aware, Ib has been used extensively in the literature (including Christie et al. (2016) as mentioned in the text, but also e.g. Bindschadler et al. (2011, The Cryosphere) and Scambos (2007, Remote Sensing of Environment)) to supplement our understanding of the GZ in areas where DInSAR coverage is poor. I was therefore a little surprised to see no discussion or method of deriving this important GZ component using ICESat-2 in the text. This is despite the break-in-slope being clearly visible in e.g. Figures 5a and 5c. Was any attempt to automate this component of the GZ made? If not, why not? I think a brief discussion of Ib, its manifestation in the ICESat-2 data and any attempts to automate it should therefore be added to the Section 3.1. If easily obtainable, then an additional figure similar to Figure 5c could also be added to show any differences in ICESat-2-derived Ib and the break-in-slope mapped by ICESat-1 (Brunt et al., 2011) and/or the ASAID team (Bindschadler et al. 2011).
On a related note, I have noticed occasional large errors in the positioning of DInSAR-derived GLs (Rignot, ESA CCI) over the Antarctic Peninsula (including Larsen C), presumably related to the geo-coding of SAR data over the Peninsula's steep terrain using old DEMs, and/or the misattribution of the Point F on ice landward of nunataks or other mountainous exposures at the GZ. These positional errors are not seen in e.g. the ASAID product due to the on-nadir viewing angle of Landsat and the ability to distinguish between mountains and ice more easily. Therefore, any comparison of ICESat-2-and ASAID-derived Ib picks may be more representative of the true GL (in lieu of Point F) in these areas. (As a relatively slow flowing region with a well-defined break-in-slope, Ib will fall close to Point G (as it does in Fig. 5), making it reliable proxy for Point G and F).
In light of the above and if not already done, the authors should thoroughly QC the DInSAR product for such offsetstheir omission could actually improve your mean absolute separation statistics! Lines 150 and 156 -How sensitive is the pinpointing of Point H to the prescribed variance threshold used in the error function, and why was the fourth derivative of this function chosen? In other words, how would e.g. a 3 rd order fit look on Figure 5d, and where would H be placed as a result? Some discussion of your choice is needed in the text here.
Line 190 -With the exception Churchill Peninsula (Figure 6a) where *one* ICESat-1 F point falls very close to the ICESat-2 F, spatially coincident ICESat-ICESat-2 observations do not exist anywhere else along the Larsen C coastline. I would therefore excise any comparison between the two datasets from the text as I don't think such statements can be made with confidence. Instead, I'd solely report on the close agreement between ICESat-2 and DInSAR observations. (the discussion of the enhanced coverage between ICESat-1 and ICESat-2 is okay, however!). Dawson and Bamber (2017) refers to Siple Coast, not Larsen C. Are the reported separation values mentioned in the text derived from those observations over the Siple coastline, or has the methodology of Dawson and Bamber been extended to Larsen C? This is essential information to state here.

Sentence beginning line 196 -The GZ information contained in
Lines 217-226 -In Figure 8 (b and c especially), I see numerous instances of dark blues over the ice shelf which, following the main text, readers might incorrectly interpret as grounded ice. The opposite is true over grounded ice, which may be interpreted to be floating. What is the explanation for these observations? Surface processes? A few sentences explaining these blips are essential here.
Lines 233-235 -Similar to my comments on the potential derivation of Ib from ICESat-2 data and its comparison with e.g. the ASAID product, I am surprised to see no comparison of the authors results with ASAID-derived H. At present this our best pan-ice sheet understanding of H, so it would be very useful to see a discussion of how your results compare here.
Line 275 -As no reference is included to support this assertion, the discussion of crevasses and their influence of GZ reads like something of an afterthought, potentially degrading the reader's confidence in this otherwise highly impressive technique. Since this is predominantly a methods paper, I'd encourage the authors to elaborate on a) why they didn't (couldn't?) clean up these data, b) exactly how they think these phenomena may influence the overall accuracy of their technique and, if needed, c) briefly mention how future work may address this issue. I imagine a lot of this could (and should) go into Section 3 so that the reader can bear these uncertainties in mind while reading Section 4. In addition, it is generally poor grammatical form to begin new sentences with 'Also', 'However', 'It' etc., so the readers should consider revising the structure of the next sentence (and several other similar instances throughout the manuscript) to address these minor niggles.
Line 277 -Good point. Does the processing chain provide a means in which the user can obtain the number of repeat-track observations your GL picks are based on? Such information could be used to calculate a 'GZ pick reliability rating', and/or inform data users about the range of tidal frequencies the GZ is mapped over.

Specific Comments
Line 6 -While it's beyond the scope of this review to provide detailed corrections to the style of English and grammar used throughout the manuscript, I've re-structured the abstract to greatly condense and more effectively convey the key selling points of paper here. The authors should feel free to amend/disregard as necessary. Authors should also note the omission of greater 'precision' relative to ICESat-1, as I strongly believe this is not convincingly addressed in the manuscript (see my comment above on line 190).
"We present a new, fully automated method of mapping the Antarctic Ice Sheet's grounding zone using repeat-pass and cross-over analyses of newly acquired ICESat-2 laser altimeter data. Our method recovers the position of the landward limit of tidal flexure and the inshore limit of hydrostatic equilibrium, as demonstrated over the mountainous and hitherto difficult to survey grounding zone of Larsen C Ice Shelf. Since launching in 2018, our method has already doubled the number of grounding zone observations acquired from ICESat-1, which operated between 2003 and2008. Acting as a reliable proxy for the grounding line, which cannot be directly imaged by satellites, our ICESat-2-derived limit of tidal flexure locations agree well with independently constrained measurements, with a mean absolute separation and the standard deviation of 0.29 km and 0.31 km, respectively, from interferometric synthetic aperture radar-based observations. Our results demonstrate the efficiency, high spatial precision and density in which ICESat-2 can image complex grounding zones, and its clear potential for future pan-ice sheet grounding-zone mapping efforts." Lines 18-21 -I think these opening sentences could be worded more eloquently to say something like: "Long-term satellite observations have linked the on-going thinning of Antarctica's ice shelves (Paolo et al., 2015) to enhanced rates of ice mass discharge beyond the grounding line (hereafter GL)the point where the grounded ice sheet first detaches from the bedrock and begins to float (Fricker & Padman, 2006). Ice discharge calculations are sensitive …".
Line 28 -Provide references for mass balance studies and those assessing overall ice sheet stability. The latest offering from the IMBIE crowd (Shepherd et al., Nature, 2018) would be good here.
Line 33 -Following my comment regarding Line 275, the authors should aim to avoid ambiguous terms like 'Its' to begin sentences. Suggest rewording to "The precise width of the grounding zone depends on …".
Lines 35-36 -This is a good example of many instances of repetition in the manuscript. The true GL (G) and limits of tidal flexure (F) have already been defined on line 30/31, so don't need to be repeated here. Saying something like the following would be much more concise and easier to understand. "… While Point G cannot be detected directly from satellite-based observations, Point F resides within close proximity of this location, and is thus widely considered to be the most robust satellite-observable proxy for the "true" grounding line (insert references here)." (Note that this rewrite also gets round the incorrect notion that the limit of tidal flexure is something only present on the ice sheet surface, when reality it is the surface manifestation of a variety of mechanical processes going on at the ice-bed-ocean interface) Paragraph beginning Line 93 -This is a good example of the 'back-and-forth' organization of the paper I mentioned in my general comments. That is, whereas reader expects this paragraph to involve a discussion of steps 1 and 2 detailed in line 91, the reader now has to go back to a discussion of repeat cycles and the grounding line reference product. In accordance with processing steps shown in Figure 3, I suggest the authors separate this part of the methods section into several distinct subheadings possibly labelled "3.1.1: Repeat-track data preprocessing" (to include discussion of nominal ref track calculation and search window size from reference GL dataset), "3.1.2: Cross-track slope correction" and "3.1.3: GZ component identification". If the authors choose to do this, then each section should also be explicitly cross-referenced in line 91. The processing steps corresponding to each sub-section could also be indicated on Fig. 3 by enclosing each sub-section in a box.
Lines 93-96 -Be sure to emphasize here that you used all three beam pairs in your analyses, as readers may interpret from the current version of the text that only one beam was used.
Line 99 -As per my comment on line 69. Should 'RPT' be 'PT'? Also consider putting inverted commas around 'nominal reference track' to emphasize that this is a newly introduced term.
Line 101 -Needed to do what? I suspect set a GZ detection search window size, but this isn't obvious from the text.
Line 103 -Why was the Mouginot dataset used over the ice rise information documented in Depoorter? Is it more accurate? More up-to-date? Explicitly state why you used this product here.
Line 104 -What is an 'initial GL product'? I see no mention of it in the text or Figure 3, and don't know how it differs from the 'reference GL product' discussed in the same sentence.
Line 111 -In 2017 Larsen C calved massive iceberg A-68 from its ice front, exposing an ~6000 km2 area of previously ice-shelf covered ocean water. Mouginot's coastline precedes this calving (and the ICESat-2 observational period), meaning that the authors have inadvertently included this huge section of open water in their analyses. This is clearly seen in Figure 2. I suggest the authors re-clip their dataset using a more up-to-date coastline, such as the one available at: https://www.add.scar.org/. The authors may also wish to consider clipping the velocity dataset shown in Figure 5 for consistency between figures.
Line 112 -This sentence is highly repetitive and includes 'elevation change' three times. Consider condensing.
Also note here the irregular ordering of the Figures in text. Figures should generally be referenced in the text in alphabetical order, so in this instance I suggest swapping Figures 5a and b in Figure 5, and changing Line 131 to say '5a'. Same for line 132 ('5b'). Also, remove "color-coded dot curves" here and in line 132; this is non-essential information best left for the caption.
Line 133 -Suggest: "To estimate Points F and H from the elevation anomalies … was calculated from the corrected …" Lines 136-160 -Following my general comment, these paragraphs are very difficult to follow, as a) they flick back and forth between the identification of F and H, and b) it is highly unclear what 'new' identification procedures the authors have used versus that discussed in Fricker & Padman (2006). There is also a lot of repetition of content already mentioned in the introduction. I strongly suggest re-structuring these paragraphs to closely follow the structure outlined below: First paragraph: Very brief overview of the anatomy of the GZ with respect to Points F and H and the work of Fricker & Padman (1-2 sentences). Then, a few sentences on how *they* identified the grounding zone.
Second paragraph: A discussion of how the authors expanded upon this earlier work and identified Point F using new slope-corrected techniques, MAEA and its second derivative. No mention of H here.
Third Paragraph: Discussion of how using MAEA alone is likely more difficult to identify Point H, and subsequently: how the authors used MAEA together with their error function to locate H. When discussing the error function, elaborate more on what "as a guide" means (readers should be able to replicate this method, so as written this seems vague/ambiguous).
Line 163 -unneeded 'the' before elevation change. Also, this and next sentence are almost complete repetition, and should be condensed (and possibly merged with the following sentence). Also define 'KDTree' (and any other abbreviation) in full when using for the first time. Sentence beginning line 180 -It is implicit in the above sentence that you did this, so suggest the removal of the sentence from the text.

Sentence beginning line 168 -
Line 183suggest rewording to better match the structure of my suggested abstract changes. Suggest something like: "Using our newly developed ICESat-2-based GZ detection algorithm, we recover a two-fold increase in the number of GZ features identified over Larsen C Ice Shelf by ICESat-1 (69 and 71 picks of Points F and H, respectively, versus 30 of each by ICESat-1). The spatial distribution of each platform's GZ picks is shown in Figure 6, together with a comparison of Point F as determined from independent DInSAR observations. The improvement in our ability to image the GZ using ICESat-2 data is especially notable in heavily crevassed regions like Jason Peninsula and Churchill Peninsula (Jansen et al. 2010), which were previously difficult to image using single-beam ICESat-1 observations alone". Lines 189-195 -As per my technical comments on Lines 183-216. Suggest restructuring this paragraph and Figure 5 to only make reference the more up-to-date ESA CCI product.
Lines 189-201 -This paragraph is another example of the somewhat unorganized writing style discussed above, this time pertaining to the back-and-forth comparison of different sensor observations. Mirroring my suggested edits to the methods section regarding the derivation of F and H, I suggest re-structuring this and the following paragraph in the following order to greatly improve clarity.
First para: Discussion of how the author's results compare with DInSAR observations (ESA CCI only). (Any comparison of ICESat-1 and -2 GZ points are largely unconvincing and should be excised from the text here).
Second para: Discussion of where your DInSAR and ICESat-2 F points don't agree (i.e. the para beginning line 202. Nice findings, by the way!).
Third para: Discussion of how your F points compare with those imaged by other sensors, including CryoSat-2 (nb. A discussion of CryoSat-2 vs. ICESat-1/DinSAR observations is outwith the confines of this study, and as a predominantly ICESat-2 methods study, should be removed. Also note the incorrect spelling of CryoSat-2). In accordance with my earlier comments, I think this paragraph would also be the place to add in comparison of your Ib picks vs. those of the ASAID product).
Fourth Para: Discussion of Point H, encompassing my comments on lines 233-235 above.

Lines 206-210 -"(vertical red dash line)", "(blue dash curve)" etc. is non-essential information best placed in the caption instead.
Lines 209-210 -Highly verbose sentence, and repetitive use of 'left' and 'right'. This sentence should be condensed.

Line 220 -Insert comma before and after |dh|.
Line 221 -Change "the repeat track analysis" to "our repeat track analysis" Line 223 -What does "same time difference" mean? Suggest rewording for clarity.

Line 224 -Suggest changing to "The locations of Point F identified using our repeat-pass technique (black dots in Figure 8c) are located in the middle of high and low |dh| observations, indicating the presence of floating and grounded ice either side of Point F, respectively".
Paragraph beginning line 227 -For simplicity and clarity, I would restructure this paragraph to first mention why we're interested in knowing the width of the GZ, then discuss how it has been calculated in the past and then how the authors have calculated it in this paper. Then finish with a discussion of your results. Similar to the discussion of Points F and H within the context of independent observations, some brief words on how measured GZ widths compare with e.g. ASAID GZ widths would also be well placed here. Also, the first sentence of this para is repeated directly from the introduction, so suggest rewording.
Line 239 -Suggest changing to: "… for recovering several commonly imaged components of the ice-sheet-shelf grounding zone, including the inland limit of tidal flexure (Point F)

and the limit of hydrostatic equilibrium (Point H). The new method presented in this study should …"
Line 240 -"The new automated method presented in this study should allow a more efficient and consistent mapping of the Antarctic GZ features". I see what the authors are trying to suggest here, but don't think it's articulated quite correctly. The authors should bear in mind that due to recent advances in polar-orbiting EO capabilities (incl. the launch of Sentinel-1a/b), we can in theory now generate a precise understanding of the GZ every 6 days using DInSAR. The spatial-temporal coverage of e.g. DInSAR-derived F will therefore always be better than ICESat-2 F, so I think this sentence (and any similar arguments made further down in the discussion and conclusion sections) should be re-worded to reflect this. What I think the authors want to say is that ICESat-2 provides an automated, efficient means of characterizing the Antarctic Ice Sheet's GZ, to complement the high-accuracy observations afforded by other modern Earth Observation sensors.
Line 241 -This sentence should go into more detail about why we are interested in GZ position change and width (e.g. for improving numerical models aiming to constrain SLR predictions, identifying Antarctica's most vulnerable regions to climate change etc.).

Line 244 -Suggest rewording to "To maximize repeat-track coverage over the GZ, our method considers the ground tracks of one ICESat-2 repeat cycle beam pair as two individual repeat tracks".
Line 247 -Should say "… which was an issue when using ICESat-1 data to identify the GZ (Brunt et al, 2010) …". Estimation implies imprecision, so identify or similar is better".
Line 248 -"In addition, this also allows us to calculate the GZ features from just two time stamps of tidal amplitudes". It's obvious to me why the authors say this, but to some this might imply an element of inaccuracy in the author's results. This is because recent research has shown that the GL can undergo significant changes in its position over a range of tidal frequencies, in some cases by as much as 4 km (e.g. Milillo et al., 2017, GRL).
I suggest the authors acknowledge this important point, and mention that while their technique does indeed work using only 2 repeat cycle's worth of data, the identification of longer-term GZ change completely distinct from tidal variability (e.g. Friedl et al. 2019) may only be possible as/when more ICESat-2 observations become available.
Line 240see my comments re: Line 240. While this was perhaps once true, with the launch of Sentinel-1a/b (and soon -1c) I think the authors should reword this to state how the two datasets will complement each other.
Paragraph beginning Line 257 -This paragraph is a bit 'matter-of-fact' or bullet-point-like, and at times not merited within the context of this paper (e.g. the review-like discussion of CS2 vs. ICESat-1 observations, which are independent of the method presented here). In any case, for simplicity I would reword the paragraph to echo the structure the results section (see my comments on Lines 189-201), and again excise any discussion of ICESat-2 vs. ICESat-1 precision from the text. As this is the discussion section, a regurgitation of the exact separation distances is also not essential here. Also, to avoid any possible confusion, I would refrain from referring to 'GL detection' when referring to non-Point G components of the GZ. Line 268 -"…, it demonstrates". What demonstrates?
Lines 270/271 -Where are these tidal estimates derived from? CATS2008a? Cite the appropriate reference here. Also, remove the space separating this and the next paragraph.
Lines 279 -Suggest rewording to something more concise like "Surface elevation changes derived from ICESat-2 cross-over data also provide valuable information about the approximate location of Antarctic grounding zones".
Line 281 -should say "… separation afforded by repeat track observations, it can still reveal the approximate location of the GZ". I also suggest the next sentence reads: "In doing so, this method has the potential to provide important validation of repeat-track-derived GZ features, including along fast-flowing ice streams where the GZ often undergoes rapid changes …" or similar.
Line 283 -Restructure sentence to condense and avoid the use of 'Also…'. This and the next sentence could also be easily merged.
Line 288 -This sentence is rather verbose, and I think it could easily be shortened and/or restructured to convey the key selling point of paper more effectively (see my wording of the abstract above).

Sentence beginning Line 288 -'the 8-month period'.
Sentence beginning Line 293 -Larsen C Ice Shelf is a pronoun and so the preceding use of the word 'the' is incorrect and should be removed. Next sentence once again begins with 'it'. What does 'it' refer to? Reword for clarity.
Sentences beginning Line 294 -I would let the readers decide if the separation between your ICESat-2 and DInSAR observation are small, so remove 'small' here. Also change all instances of 'GL' to 'GZ' for accuracy, and hyphenate all uses of the word "derived".
Line 296 -Again, I'm just not sure you can confidently talk about ICESat-2 vs. ICESat-1 precision that way. Also, as far as I can tell, in the current version of the manuscript the authors make no direct comparison between their ICESat-2-and any CryoSat-2-derived F products, so don't think any robust statements can be made regarding precision in that sense either. As a result, I think this sentence should be restructured to echo my suggested restructuring of the results section.
Line 297 -Change 'this method' to 'our method' and remove 'the' before Larsen C Ice Shelf.
Line 301 -Suggest changing to "…Antarctic Ice Sheet, similar cross-over-based analyses have the potential to provide more accurate depictions of the GZ".
Sentence beginning line 301 -This is almost complete repetition of lines 293-295, so I suggest removing it from the text.
Line 310 -Where can the ESA CCI DInSAR data be found?
Line 333 -Incorrect dataset cited.    Borstad et al., 2017;GRL), so it'd be interesting to see any available GZ picks here. I suggest also standardizing the font size of all labels on the main map, and instead assigning track labels a different font color to better contrast against place-names. I also suggest center justifying the scale bar relative to the other items of the legend, and moving all items slightly to the left (at the moment the 1000 m label intercepts the figure border and looks untidy). Taking potentially color-blind readers into consideration, I'd also encourage the authors to choose another, non-jet/rainbow color ramp.    Figure 5 -Why is repeat cycle 5 not shown? In the main text it is stated that this dataset is used, so its omission here is rather odd. Suggest also ordering the repeat cycle legend in numerical order (3l, 3r, 4l, 4r and so on) Also, I don't quite understand the inclusion of elevation by latitude (dashed profiles) here?
The only reason I can think of is because track 1192 overpasses two GZs on the north and south of the unnamed ice rise seen in Fig.2 (and so should have some easily identifiable break-in-slope), but even then it's unclear where this happens. In any case, when superimposed over the anomalies and other info it makes for a rather busy diagram, so I'd suggest removing it for simplicity. If the authors feel this is essential information, then I recommend they include it either as an additional sub-panel to this figure (cf. Fricker et al., 2006) or as a supplement.
In all panels, some of the axis and other labels appear cut-off, so care should be taken to fix these minor blips. Figure 5 caption -This is a very long and difficult to comprehend sentence (and caption overall). Can it be condensed somehow? I'm also not sure the second sentence makes grammatical sense, and could be shortened to: "in the legend, c03_1r refers to the right ground track of ICESat-2 pair 1, repeat pass 3".
After the Padman reference, insert "following Fricker & Padman, 2006". Information about the derivation of MAEA has already been discussed in the text, and so should not be included in the caption. Capitalize all instances of the word 'Point'.     In c, a black circle appears to be obscuring a blue circle. Suggest using unfilled black circles (i.e. outline only) in all subplots to avoid this. Suggest also scaling the black circle in the legend to match the its size as shown in the sub plots. Figure 8 caption -Incorrect grammar. Suggest "Spatial distribution of ICESat-2 crossovers analyzed in this study. The spatial extent of plots a-c is shown in Figure 6c. At each location, the absolute change in elevation, |dh|, is shown. Background is …".