the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 04 Oct 2023
Sentinel-1 detection of perennial firn aquifers in the Antarctic Peninsula
Abstract. In recent years, the existence of perennial firn aquifers in the Antarctic Peninsula (AP) has been confirmed by in situ observations. Previous studies have suggested that these subsurface aquifers, together with meltwater ponds at the surface, provide a reservoir of liquid water to feed propagating fractures, promoting hydrofracture-driven ice-shelf disintegration. This study maps perennial firn aquifers in the AP from space using C-band Synthetic Aperture Radar imagery from ESA's Sentinel-1 (S1) mission. With these observations, we detect firn aquifers at 1 km × 1 km resolution, for the period 2017 to 2020. Existing methods, that use S1 data and rely on a backscatter intensity difference threshold approach, are prone to misclassify late-melt events as aquifers, when applied to the AP. Therefore, we have developed and evaluated a new approach that is better suited to the Antarctic environment. The new method exploits the characteristic, gradual backscatter increase during the (partial) refreezing of the liquid water in the firn layer after the peak of the melt season. Most firn aquifers are detected in the north and northwest of the AP, as well as on the Wilkins and George VI ice shelves. Aquifer locations detected with the present methodology agree with in situ observations and with model simulations of firn water content.
- Preprint
(3244 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 16 May 2024)
-
RC1: 'Comment on egusphere-2023-2000', Anonymous Referee #1, 31 Oct 2023
reply
This manuscript details a new method for detecting firn aquifers over the Antarctic Peninsula (AP) from Sentinel-1 synthetic aperture radar data. The detection algorithm is based on the idea that in regions with firn aquifers, summer meltwater will refreeze more slowly within the firn due to high melt and warm subsurface temperatures. This refreezing process is reflected as a rebound of the SAR surface backscatter value from low values of wet firn in the melt season, to higher values over dry firn in the refreezing season. The authors set a detection threshold of that firn aquifers exist where backscatter reaches 80% of its average September value on or after the 97th day of the year. Using this threshold, they map aquifers in the 2017-2020 seasons. They compare these mappings to predictions from the IMAU-FDM firn model as well as an S1 mapping using an established firn aquifer detection algorithm developed in Greenland. They show that their new algorithm is more consistent with field observations from the AP than the Greenland algorithm and in generally good agreement with the IMAU-FDM predictions.
The authors have worked hard to address comments from a previous round of review, and the manuscript has significantly improved. Unfortunately, mapping firn aquifers with S1 is always going to be tricky because the limited penetration depth means that the water table cannot be sensed directly. Instead, these algorithms rely on delayed refreezing in the overlying firn as a proxy for the presence of aquifers. On the AP, this is further confounded by the lack of extensive validation data with temporal overlap with S1. In light of all of this, I think the authors can do more to be cautious in their wording and acknowledge that while their detections are consistent with extended retention of liquid water in the shallow firn, it remains unclear if these areas are truly firn aquifers in the traditional sense of a deep zone of perennially fully saturated firn.
Major Comments:
[1] The fact the IMAU FDM simulations fail to capture the location and extent of Wilkins Firn Aquifer, where there is extensive field and ice-penetrating radar evidence for the aquifer, suggests that comparison with the model may not be the best form of validation. At a minimum, the authors ought to acknowledge this discrepancy (which they show in Figure 8a) and the limitations of the model.
[2] I do not fully understand why the authors do not compare their S1 detections to the OIB-detected extent of the Wilkins Firn Aquifer. While the observations are not contemporaneous (2014 for OIB vs 2017-2020 for S1), this comparison would still provide valuable confirmation that S1 detects the generally correct extent and region of the aquifer. The field measurements of the Wilkins Aquifer seem to suggest that it should be quite stable, given the large water volumes retained. This seems like an oversight, particularly given that the authors later call out the OIB data as something that should be analyzed in the future!
[3] The time series from the Muller Ice Shelf in Figure 10d almost seems to suggest a very long melt season on that shelf, given that the time series starts off in December with backscatter of ~ -15dB and already a high LWC. I think this is indicative of a general issue that some of these regions on the AP could be seeing surface melting throughout a large fraction of the year, which would negate a recovery threshold algorithm that assumes recovery occurs. The -12 dB secondary threshold is used to include these areas in the firn aquifer extent, but this then raises the question of whether the map is now only firn aquifers, or firn aquifers and regions of extended surface melting. It would be enlightening to actually look at the length of the melt season across the entire AP during the study period and understand if there are regions where extended surface melting is going to mask detection of any subsurface structure.
[4] I think the text could do more to clearly acknowledge that it is not possible to do reliable year-to-year mapping of aquifer extent with this algorithm (as shown by the absence of the Wilkins aquifer in the 2018 and 2019 data, even though folks were in the field actively observing the aquifer in December 2018!). This is consistent with results from Greenland (at 4-5 years of aggregated data was used for both the S1 and SMAP mappings there).
Minor Comments:
The introduction could use some paragraph breaks.
Line 34 – I think this citation should be Miller et al (2020b)?
Line 63 – technically this is 410 km for the swath width (see table here, even though they say 400 km in the paragraph above it: https://sentinels.copernicus.eu/web/sentinel/user-guides/sentinel-1-sar/acquisition-modes/extra-wide-swath)
Line 73 – I think this is the same method originally described in:
Lievens, H., Demuzere, M., Marshall, H.-P., Reichle, R. H., Brucker, L., Brangers, I., et al. (2019). Snow depth variability in the Northern Hemisphere mountains observed from space. Nature Communications, 10, 4629. https://doi.org/10.1038/s41467-019-12566-y
It was also used in Brangers et al. (2020), so seems appropriate to cite those papers unless this is such a popular technique as to be common knowledge at this point.
Line 80 – perhaps add the field site dots also to Figure 1 so that the reader does not have to flip to the end of the paper to find Figure 5.
Line 120 – should be “time series” not “time-serie”.
Line 123 – can you describe why you chose these microstructural parameters?
Section 2.5 – how does this change if saturation is greater than the irreducible liquid water content? For example, an actual aquifer near the surface should be fully saturated, which would be at least 20% LWC by volume given a density of 600 kg/m^3. It is not clear to me why the LWC is set to the irreducible water content when making comparisons with real-world data where saturation can be higher, rather than the IMAU simulations.
I wonder if Section 5.2 would be better presented in the methods around Section 2.5.2 when you first discuss setting the DOY80 parameter? My first thought when reading the methods was that there as not justification for the chosen thresholds and this comes quite late in the paper to fill that gap.
Line 310 – is this supposed to refer to new OIB flight collection, or a comparison with the past data? If past data, that is all public and could be analyzed right now and would significantly strengthen a paper like this by providing some degree of quantitative validation data. If new collection, that will not be OIB and it would be better to say that new collection is needed to better validate these satellite system algorithms.
Citation: https://doi.org/10.5194/egusphere-2023-2000-RC1 -
RC2: 'Comment on egusphere-2023-2000', Anonymous Referee #2, 07 Nov 2023
reply
Buth et al., describes the development of a Sentinel-1 detection algorithm that uses a set of thresholds and fixed dates to map perennial firn aquifers in the Antarctic Peninsula, and then compares these results to IMAU-FDM. The manuscript further details the simulation of backscatter time series using IMAU-FDM and the SMRT radiative transfer model to validate the comparison.
This is a resubmit of a manuscript that I have already reviewed. Unfortunately, there are still significant theoretical and technical issues throughout the manuscript. With the exception of the inclusion of a second Greenland-derived algorithm with equally unvalidated results - this is basically the same paper.
The majority of my previous comments – all of which have been pointed out for a second time by Reviewer 1 -- have not been addressed.
https://egusphere.copernicus.org/preprints/2023/egusphere-2023-2000/
Critically -- the algorithm is not capable of mapping perennial firn aquifers in areas that have been fully validated via field expeditions and airborne radar surveys. The Southern George VI has also been drilled and airborne-validated. The is no evidence of perennial or seasonal firn aquifers in this area, which is mapped by the algorithm. As a reviewer – it’s game over at this point. However, thousands of additional square kilometers are mapped in areas of the Northern Antarctic Peninsula -- in complex terrain with steep slopes, accelerated ice flow, extensive crevassing and fracturing, and rock outcrops --these features are easily observed by simply looking at visible satellite imagery. These areas are unlikely to host perennial firn aquifers and will confound any radar system. The embayment of the former Wordie Ice Shelf, in particular, has decades of ice thickness measurements (i.e. there are bottom echoes from bedrock and no perennial firn aquifers).
Citation: https://doi.org/10.5194/egusphere-2023-2000-RC2 -
RC3: 'Comment on egusphere-2023-2000', Anonymous Referee #3, 23 Apr 2024
reply
This paper presents a new method for mapping perennial firn aquifers on the Antarctic Peninsula from Sentinel-1 data. The new approach is developed to mitigate issues due to misclassification from surface melt events in a previously published method. While the methodology is clearly described and the manuscript is well written, I have major concerns regarding the indirect nature of the retrievals and the approach to validation which I believe seriously limit the meaningfulness of the results.
- Physics of the retrieval. The retrieval is indirect because, as the authors point out, “the S1 method does not directly estimate the liquid water content”. While the time series in Figure 10 are encouraging, the absence of surface temperature information in this plot makes it impossible to identify the influence of surface melt. As the authors point out, the presence of surface melt is problematic for the pre-existing backscatter intensity difference threshold algorithm, but this will also hinder the performance of the recovery time threshold algorithm. Furthermore, I would be curious to see a difference map between Figure 5 and 6, especially if you remove the single year retrieval in Figure 5 (as was done in the central panel of Figure 6). There are some obvious problematic differences, e.g. the Wilkins Ice Shelf, but otherwise it would be insightful to see a map of the actual difference in the two algorithms.
- Approach to validation. The justification for providing “a good estimation of the extent of firn aquifers in the AP” (line 274) is based on the combined use of in situ observations, a firn model, and regional climate output. Each of these methods is problematic. The comparison with in situ observations in Section 4.1 is anecdotal, and based only on a very limited number of locations with known aquifers. The comparison with RACMO simulations in Figure 7 is also indirect. The finding that “Detected aquifers are preferentially located in regions where both surface melt and accumulation rates are high.” is somewhat encouraging but not a form of validation. The firn model comparison is similarly indirect, making it challenging to truly diagnose the algorithm performance. No assessment of false positive or false negative retrievals is provided so it’s not possible to assign a level of confidence to the retrieved locations of firn aquifers. These limitations to the validation could be addressed using IceBridge data (despite a temporal offset) to perform a direct validation of perennial aquifers, as the authors point out in the final line of the manuscript.
Citation: https://doi.org/10.5194/egusphere-2023-2000-RC3
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 444 | 157 | 26 | 627 | 20 | 16 |
- HTML: 444
- PDF: 157
- XML: 26
- Total: 627
- BibTeX: 20
- EndNote: 16
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1