The Antarctic grounding zone, which is the transition
between the fully grounded ice sheet to freely floating ice shelf, plays a
critical role in ice sheet stability, mass budget calculations, and ice sheet
model projections. It is therefore important to continuously monitor its
location and migration over time. Here we present the first ICESat-2-derived
high-resolution grounding zone product of the Antarctic Ice Sheet, including
three important boundaries: the inland limit of tidal flexure (Point F),
inshore limit of hydrostatic equilibrium (Point H), and the break in slope
(Point I
With a global sea level rise equivalent of 58 m (Fretwell et al., 2013), the Antarctic Ice Sheet has been losing ice at an accelerated pace (Shepherd et al., 2018). This mass loss is largely driven by the ice dynamics of the marine ice sheet due to sustained and accelerated thinning of the ice shelves (Bamber et al., 2009; Paolo et al., 2015; Pattyn and Morlighem, 2020; Favier et al., 2014; Gardner et al., 2018) and rapid retreat of the grounding line (hereinafter referred to as the GL) (Point G in Fig. 1) (Christie et al., 2018; Milillo et al., 2019; Rignot et al., 2014; Scheuchl et al., 2016), which is the boundary between the grounded ice sheet and the floating ice shelves (Rignot et al., 2011a). The grounding line is identified as an essential climate variable that is critical in understanding Earth's climate by the Global Climate Observing System. Knowledge of its location is important in ice sheet numerical modelling and mass budget estimation as it controls the rates of ice flux from the grounded ice sheet into the ocean (Schoof, 2007), and it is a key indicator of the marine ice sheet instability (DeConto and Pollard, 2016; Joughin et al., 2014; Ritz et al., 2015). Therefore, continuous long-term monitoring of the GL location and its temporal migration is crucial for understanding ice sheet stability and assessing the Antarctic Ice Sheet's contribution to future sea level rise.
Schematic diagram of the ice shelf grounding zone (GZ)
structure adapted from Fricker and Padman (2006). Point G is the true
grounding line where the grounded ice first comes into contact with the ocean,
Point F is the landward limit of ice flexure caused by ocean tidal movement,
Point H is the seaward limit of ice flexure and the inshore limit of
hydrostatic equilibrium, Point I
The GL is located inside the grounding zone (hereinafter referred to as the
GZ; Fig. 1). The GZ is defined as the region between the landward limit of
tidal flexure (Point F in Fig. 1), where the ice is not influenced by ocean
tides, and the inshore limit of hydrostatic equilibrium (Point H in Fig. 1)
where the ice is floating in full hydrostatic equilibrium
(Brunt et al., 2010b; Fricker
and Padman, 2006). Inside the GZ, there is often a surface elevation minimum
(Point I
There are two established approaches for estimating the GL location using remote sensing techniques: (a) directly detect the break in slope (hereinafter referred to as the “static method”); (b) use observations of surface elevation change due to variations in ocean-tide-induced tidal flexure (hereinafter referred to as the “dynamic method”). The break in slope is mapped by identifying the inflection of the ice surface slope from a digital elevation model (DEM) (Brunt et al., 2010b, 2011; Fricker and Padman, 2006; Hogg et al., 2018; Horgan and Anandakrishnan, 2006) or the change in brightness in satellite optical imagery (Bindschadler et al., 2011; Christie et al., 2016, 2018). The satellite-optical-imagery-based approaches are able to provide complete coverage of the Antarctic Ice Sheet (Bindschadler et al., 2011; Scambos et al., 2007). However, they work best only when the ice thickness increases rapidly inland from the GZ and often fail to map the GL in areas of fast ice flow where the subglacial bed and surface slope are shallow (Christie et al., 2016, 2018).
Repeat-track and crossover analysis of satellite altimetry
(Brunt
et al., 2010b, 2011; Dawson and Bamber, 2017, 2020; Fricker and Padman,
2006; Li et al., 2020) and differential synthetic aperture radar
interferometry (DInSAR)
(Brancato
et al., 2020; Mohajerani et al., 2021; Rignot et al., 2016; Rignot, 1998;
Scheuchl et al., 2016) use the dynamic method to detect Points F and H. In
general, DInSAR has been the most successful method of capturing Point F
accurately and providing overall good spatial coverage. However, there are
relatively few regions that have been measured repeatedly by DInSAR
(Friedl et al.,
2020; Hogg et al., 2018), while some areas have not been mapped at all due
to orbital limitations of the satellites
(Mohajerani et al.,
2018). Satellite altimetry, therefore, can provide valuable information
where DInSAR measurements are not available. The existing satellite-altimetry-derived GZ products from ICESat (Brunt et al.,
2010a) and CryoSat-2 (Dawson and
Bamber, 2020) suffer from poor temporal and spatial coverage and are not
suitable to monitor changes in the GZ. ICESat-2, launched on 15 September 2018, however, has higher along-track resolution and better spatial coverage
compared with ICESat (Markus et
al., 2017). It can be used to map the Antarctic GZ with greater accuracy and
spatio-temporal coverage than previous satellite-altimetry-derived products.
Here we generated the first ICESat-2-derived Antarctic GZ product with high
spatio-temporal coverage using 18 months of ICESat-2 laser altimetry data
(Li et al., 2021), including three GZ features: Points F,
H, and I
This paper provides a detailed description of the ICESat-2-derived GZ product and the methodologies used to derive the dataset. We also discuss the associated uncertainties and validate the new GZ product with ICESat-2 crossover measurements and previous GZ products.
ICESat-2 measures the ice sheet surface elevation at a repeat cycle of 91 d. The Advanced Topographic Laser Altimeter System (ATLAS) onboard ICESat-2 has three beam pairs in comparison with the single beam of the Geoscience Laser Altimeter System (GLAS) onboard ICESat. The across-track spacing between each beam pair is approximately 3.3 km, with a pair spacing of 90 m. The along-track sampling interval of each beam is 0.7 m with a nominal 17 m diameter footprint (Markus et al., 2017; Smith et al., 2019).
In this study, we used version 3 of the ATL06 Land Ice Along-Track Height
Product (Smith et al., 2019) from 30 March 2019 to 30 September 2020 (Scheick et al.,
2019; Smith et al., 2020a) to map three different GZ features, including the
landward limit of tidal flexure (Point F), the inshore limit of hydrostatic
equilibrium (Point H), and the break in slope (Point I
We processed the ATL06 elevation data using the same methods described in Li et al. (2020). We did not apply the ocean tide correction to ICESat-2 ATL06 elevation, and we “re-tided” the ocean loading tide. Poor-quality elevation measurements caused by clouds or background photon clustering were removed by applying the ATL06_quality_summary flag (Smith et al., 2019). A neighbouring surface elevation consistency check was applied by using the along-track slope of each ground track. We only kept elevation measurements where differences between the original elevations and the estimated elevations from along-track slope were lower than 2 m. The reference segment locations of each ground track were also derived from the “segment_quality” group to calculate a reference track, which will later be used in the GZ calculation.
Our method of estimating GZ features utilizes ICESat-2 repeat tracks from different cycles (Fig. 2, Box 1). Following the steps of repeat-track generation described in Li et al. (2020), the surface elevation, elevation measurement geolocations, and the reference segment geolocations of 6 ground tracks along each of the 1387 reference ground tracks (RGTs) were categorized into 9 distinct repeat-track data groups, including 6 single-beam repeat-track data groups and 3 beam pair repeat-track data groups (Fig. 4a and b in Li et al., 2020). For each repeat-track data group, a “nominal reference track” was calculated by averaging the locations of reference segments from all repeat tracks inside this data group. A reference GL was also calculated as the intersection between the nominal reference track and a composite GL which was generated by merging the Depoorter et al. (2013) GL with the most recent GLs from different sources (Table A1). Allowing for a possible GL change between the current GZ location and the composite GL, we defined a 15 km calculation window landward and seaward of the reference GL along the nominal reference track; only ATL06 elevation measurements located within this calculation window were used in the GZ calculation (Li et al., 2020). This is to ensure the pre-defined calculation window can capture the GZ adequately in our study period due to potential GL changes during the past decade, especially for the fast-flowing glaciers.
The automatic workflow of identifying the grounding zone
(GZ) features from ICESat-2 data. Box 1: ICESat-2 repeat-track preparation;
Boxes 2 and 3: estimation of the landward limit of tidal flexure (Point F)
and the inshore limit of hydrostatic equilibrium (Point H) from the dynamic
method; Boxes 4 and 5: estimation of the break in slope Point I
We removed ATL06 data points with elevation higher than 400 m and data
points located in open water based on the coastline mask provided in the
SCAR Antarctic Digital Database (ADD)
(
The key feature of the dynamic method is to identify the temporal changes in ice surface elevation due to ocean tides between Points F and H from different repeat tracks (Brunt et al., 2010b, 2011; Fricker and Padman, 2006). The temporal ice surface elevation changes were derived from a set of “elevation anomalies” (Fig. 2, Box 2). For each single-beam repeat-track data group, the reference elevation profile along the nominal reference track was first calculated by averaging the elevations of each repeat track at the nominal reference track; then elevation anomalies were calculated by differencing the elevation profile of each individual repeat track and this reference elevation profile (Li et al., 2020) (Fig. 3c, h, m). For the beam pair repeat-track data group, the elevation profile of each individual repeat track was first corrected for the across-track slope onto the nominal reference track (Eqs. 1 and 2 in Li et al., 2020). The average of all across-track slope-corrected elevations from each track at the nominal reference track was then taken as the reference elevation profile. The elevation anomalies were calculated by subtracting this reference elevation profile from the across-track slope-corrected elevation profile of each repeat track inside the beam pair repeat-track data group.
Examples of repeat-track analysis for track 887
The estimation of GZ features Points F and H are based on extracting the transition points from the mean absolute elevation anomaly (MAEA) (Fig. 3d, i, n), which is defined as the average of the absolute value of all elevation anomaly profiles. The inland limit of tidal flexure, Point F, is identified as the point where the elevation anomaly of each repeat track exceeds a noise threshold (Brunt et al., 2010b, 2011; Fricker et al., 2009). The region where the MAEA is close to zero is regarded as the fully grounded ice (the region to the left of Point F in Fig. 1) as it is not influenced by tidal motion. Point F was then estimated to be the point where the gradient of the MAEA first increases from zero, and the second derivative of the MAEA reaches its positive peak (Li et al., 2020). The inshore limit of hydrostatic equilibrium, Point H, is identified as the location where the elevation anomaly of each repeat track reaches its maximum and becomes stable. It was estimated as the transition point where the gradient of the MAEA finally decreases to zero, and the second derivative of the MAEA reaches its negative peak (Li et al., 2020).
To select the correct transition points from the second derivative of the MAEA curve as Points F and H, previously we used an error function fit to the MAEA as a guide (Li et al., 2020). While the error function can reliably estimate Point H because the gradient of the elevation anomaly always changes smoothly to zero, it is unreliable in identifying Point F where there is a sharp transition on the MAEA curve, or the across-track slope-related noise on land ice is high (green dots in Fig. 3j and o). To solve the inaccurate picks of Point F under these circumstances, instead of using error function fitting, we used a three-segment piecewise function fitting only to the landward part of the Point H on the MAEA profile (Fig. 2, Box 3) (green lines in Fig. 3e, j, o). The closest positive peak of the second derivative of this piecewise function to the reference GL was taken as a guide point to find Point F. As a final step, all results are visually inspected due to the complex nature of the GZs, and ICESat-2 crossover measurements are used as a reference on the Filchner–Ronne and Ross ice shelves (Sect. 2.5). In the final GZ product, we also recorded the number of repeat cycles used and the ocean tide range calculated as the maximum elevation anomaly deviation from all repeat tracks at Point H.
The break in slope Point I
The break in slope Point I
After obtaining the reference elevation profile on the nominal reference track of each single-beam repeat-track data group (Fig. 4b, h, n), we first linearly interpolated the reference elevation based on sequential segments at an along-track distance of 20 m to fill the data gaps (cyan lines in Fig. 4c, i, o). To remove noise caused by small-scale topographic features such as crevasses, we applied a Butterworth low-pass filter with a normalized cut-off frequency of 0.032 and an order of 5 to the interpolated reference elevation profile (black lines in Fig. 4c, i, o). The low-pass filter removed the high-frequency noise without changing the shape of the reference elevation profile, therefore retaining the locations of GZ features.
Estimation of break in slope (Point I
We calculated the along-track rms height
The along-track surface slope (Fig. 4e, k, and q) and the slope break
(Fig. 4f, l, and r), which is the gradient of the along-track slope, were
calculated from the low-pass-filtered reference elevation profile. A group
of peaks were identified from the absolute values of the slope break as
potential break-in-slope features (red crosses in Fig. 4f, l, and r) as
they are the locations where the along-track slopes change most rapidly. The
break in slope Point I
To validate the repeat-track-derived GZ features, we calculated the elevation changes at crossovers from ICESat-2 ascending and descending tracks (Fig. 2, Box 6). This can be used to measure the grounding line (Li et al., 2020), which is the boundary between high elevation changes on floating ice due to tidal movement and low elevation changes on land ice not influenced by ocean tides. In this study, the crossover analysis was performed at the two largest ice shelves in Antarctica with the highest crossover densities, the Filchner–Ronne and Ross ice shelves. To calculate the elevation changes at crossovers, we closely follow the methodology developed in Li et al. (2020). When removing the crossovers with time stamps of the ascending and descending tracks in the same tidal phase on floating ice, we set a minimum threshold of elevation change due to ocean tides on floating ice to be 20 cm as the minimum detectable tidal amplitude from repeat-track analysis is around 10 cm over the two ice shelves. After deriving the mean elevation difference at each crossover, we interpolated them onto a 2 km regular polar stereographic grid using a distance-weighted Gaussian kernel. The correlation length of the Gaussian kernel is 5 km, and it uses the nearest 100 measurements. For the final gridded crossover elevation changes, we set a threshold of 20 cm for the location where the ice starts to be affected by ocean tides, which is Point F. We are aware that the elevation change threshold of Point F is not constant across all the regions of these two ice shelves. However the 20 cm threshold represents the most conservative estimation of Point F location, such that a crossover with an elevation change less than 20 cm should be grounded ice.
The highest absolute precision in identifying the GZ features Points F, H,
and I
To estimate the positional uncertainty in the GZ features, we compare the
results calculated along the left and right beams as well as the nominal
reference track in each beam pair. As the left and right beams are only
separated by approximately 90 m, and the GZ identified from the repeat-track
analysis for a beam pair is often located in the middle between the left and
right beams (
Mean absolute separations and standard deviations between the grounding zone features calculated from the single-beam repeat-track data group and beam pair repeat-track data group.
Using the GZ mapping techniques developed in this study, we produced a new
high-resolution GZ product (Li et al., 2021) by
identifying 21 346 Point F (Fig. 5a), 18 149 Point H (Fig. 5b), and 36 765 Point I
Spatial distributions of ICESat-2-derived grounding zone
features of the Antarctic Ice Sheet.
Compared with the ICESat-derived GZ product (Brunt et
al., 2010a), which has 1497 Point F, 1470 Point H, and 1493 Point I
Elevation changes at crossovers on the Filchner–Ronne and Ross ice
shelves were mapped in our study (Figs. 6, 7). The transitions from land ice
(low
On the main glacier trunk of the Support Force Glacier (Fig. 6d), the crossover-derived GL and ICESat-2-derived Point F align well with the ESA Climate Change Initiative (CCI) DInSAR-mapped Point F in 2016 and the CryoSat-2-derived Point F in 2017. On the western side of the glacier, the ICESat-2-derived Point F, crossover-derived GL, and the CryoSat-2-derived Point F show an approximately 10 km seaward migration compared with the ESA CCI DInSAR-derived Point F in 2016. On the main glacier trunk of Bailey Ice Stream, the crossover-derived GL, ICESat-2-derived Point F, CryoSat-2-derived Point F, MEaSUREs DInSAR-derived Point F, and the ESA CCI DInSAR-derived Point F in 2014 agree well with each other (Fig. 6e). However, on the northern flank of the Parry Peninsula (Fig. 6e), the ESA CCI DInSAR-derived Point F shows an approximately 10 km retreat compared with all the other GL measurements.
On Crary Ice Rise (Fig. 7c), ICESat-2-derived Point F locations agree well with the crossover-derived GL distribution but show a retreat of up to 15 km compared with all the previous GL measurements. On Mercer Ice Stream and Siple Dome, the ICESat-2-derived Point F and crossover-derived GL have good agreement with the previous GL products (Fig. 7b and d). On Echelmeyer Ice Stream, where there is only one ICESat-derived Point F, the ICESat-2-derived Point F locations show an approximately 30 km retreat compared with ICESat-derived Point F but agree well with the CryoSat-2-derived Point F in 2017 and the ICESat-2-crossover-derived GL (Fig. 7e), further confirming the conclusion that ICESat picked the wrong Point F in this region (Dawson and Bamber, 2017).
In addition to comparing the ICESat-2-derived Point F with ICESat-2
crossover measurements and the historic GLs on the Filchner–Ronne and Ross
ice shelves, we compared the ICESat-2-derived Point F to the latest
pan-Antarctic DInSAR-derived GL product, which was estimated from a
deep-learning-based approach by using Sentinel-1a/b synthetic aperture radar (SAR) images in 2018
(Mohajerani et al., 2021). With its acquisition
time close to ICESat-2 (up to 1 year apart), we do not expect large
separations in GL locations between these two products due to any changes in
the GL. The Sentinel-1a/b DInSAR-derived GL has a precision of 200 m
(Mohajerani et al., 2021); however due to
limitation of Sentinel's coverage in polar regions, this product does not
fully cover the Filchner–Ronne and Ross ice shelves. The absolute
separations between 2018 DInSAR-derived Point F with ICESat-2-derived Point F are shown in Figs. 8a and A3. Despite the relatively small difference in
measurement time, there may still be changes in Point F. In general, the
rapid retreat of the grounding line happens in fast ice flow
(Konrad et al., 2018). Therefore, we also divided the
GL separations into two categories: slow-moving regions where the ice
velocity is less than 100 m yr
In total, the mean absolute separation and standard deviation across the ice sheet between the two products are 0.02 and 0.02 km, respectively, comparable to the precision of the DInSAR GL product (Table 2). This indicates that the ICESat-2-derived Point F can achieve the same level of precision compared to DInSAR measurements. A total of 84 % of the surveyed GZ is located in slow-moving regions. As expected, the overall mean separations and standard deviations in slow-moving regions, where the GL is normally stable, are lower than in fast-flowing regions. The increase in GL separation in fast-flowing regions between the two products is possibly due to the reduced ICESat-2 GL measurements caused by low signal-to-noise ratio in elevation anomalies of repeat tracks and the fact that DInSAR often suffers from poor signal coherence due to high ice velocity. In the Amundsen Sea embayment and Bellingshausen Sea sector, which have been experiencing substantial mass loss and rapid GL retreat during the past 2 decades (Bamber and Dawson, 2020; Milillo et al., 2017, 2019; Rignot et al., 2014, 2019; Scheuchl et al., 2016), the mean absolute separations in fast-flowing regions are 0.17 and 0.24 km, respectively. The highest mean absolute separation and standard deviation, however, are located in Wilkes Land, East Antarctica (Table 2, Figs. 8a and A3). The Moscow University and Totten Glacier ice shelves in Wilkes Land are both narrow embayments with fast ice flow, where the ice may not be in full hydrostatic equilibrium, and the high ice velocity can often lead to DInSAR measurement errors.
Mean absolute separation (km) and standard deviation (km) between ICESat-2-derived landward limit of tidal flexure (Point F) and 2018 DInSAR-derived Point F (Mohajerani et al., 2021) in individual regions.
In slow-moving regions, we observed large deviations between the two products such as the Dronning Maud Land (Figs. 8b and 9a). They are possibly caused by the ephemeral grounding of ice on the scale of kilometres across the ice plain with low surface slope as the ocean tide rises and falls (Bindschadler et al., 2011; Brunt et al., 2011; Milillo et al., 2017). Here we took two examples to demonstrate the short-term GZ feature migration induced by ocean tide oscillation. On the Novyy Island of Dronning Maud Land, the distance between the ICESat-2-derived Point F along the right beam of track 145 is about 2 km compared with the 2018 DInSAR-derived Point F (Mohajerani et al., 2021) (Fig. 9a), while the ICESat-2-derived Point F along the left beam of track 153 in the same region is less than 100 m away from the DInSAR-derived Point F (Fig. 9f). The large difference in Point F location is not caused by errors in methodology but is due to the tidal variations on a lightly grounded ice plain in this region. The tidal range at Point F along track 145 is 0.41 m, while it is 1.03 m at Point F along track 153. The observation suggests that the ice shelf is grounded at low tide and floating at high tide (Brunt et al., 2011).
Comparison between the inland limit of tidal flexure
(Point F) from repeat-track analysis for two tracks located in the same region
on the Dronning Maud Land under different ocean tidal amplitude ranges. Same
as Fig. 3,
Although Point F and Point I
We also compared the ICESat-2-derived Point I
Mean absolute separation (km) and standard deviation (km)
between ICESat-2-derived break in slope (Point I
Mean absolute separation (km) and standard deviation (km)
between ICESat-2-derived break in slope (Point I
Detailed spatial distribution maps of the ICESat-2-derived Point I
Spatial distributions of ICESat-2-derived break in slope
(Point I
The inshore limit of hydrostatic equilibrium, mapped from the ASAID project, is the most complete product for Point H to date and was derived from ICESat-derived Point H and Landsat-7 imagery (Bindschadler et al., 2011). The positional error in Point H from the ASAID product is about 2 km. The absolute separation between the ICESat-2-derived Point H and ASAID Point H is shown in Fig. 13. The overall mean absolute separation and standard deviation for the whole Antarctic Ice Sheet between the two products are 1.65 and 1.29 km (Table 5), respectively, which are within the 2 km geolocation error in ASAID Point H. However, they vary by region (Fig. 13 and Table 5). The Larsen C Ice Shelf has the smallest mean absolute separation and standard deviation, while the Amery Ice Shelf has the highest mean absolute separation and standard deviation of 2 and 1.62 km, respectively.
The absolute separations between the ICESat-2-derived Point H and the Point H from the ASAID grounding line project (Bindschadler et al., 2011). Data are superimposed over recent ice surface velocity magnitudes (Rignot et al., 2017) and the IMBIE basin boundary (Shepherd et al., 2018; Rignot et al., 2011b) in Antarctic polar stereographic projection (EPSG:3031).
The mean absolute separation and standard deviations between ICESat-2-derived Point H and ASAID-derived Point H (Bindschadler et al., 2011).
The location of Point H is not stagnant but changes with ocean tides. On the western flank of the Skytrain Ice Rise on the Filchner–Ronne Ice Shelf (Fig. 14), the ICESat-2-derived Point F locations along the left (Fig. 14a–e) and right beams (Fig. 14f–j) of track 1071 are separated by 158 m. However, the distance between the ICESat-2-derived Point H locations is 6 km. The tidal range at the seaward Point H along the left beam of track 1071 is 3.3 m, while it is only 0.8 m at the landward Point H along the right beam of track 1071. This indicates that the ocean tide oscillation will not only influence the grounding point of the ice but will also change the point of hydrostatic equilibrium. More examples will be used to fully investigate the influence of ocean tides on the GZ width in future research.
Comparison between the inshore limit of hydrostatic
equilibrium (Point H) from repeat-track analysis for left and right beams of
track 1071 located on the Skytrain Ice Rise under different ocean tidal
amplitude ranges. Same as Fig. 3,
Although good agreement exists between the ICESat-2-derived Point F and DInSAR-derived Point F in 2018, large deviations have been observed in slow-moving regions due to short-term GL migrations over ice plains caused by ocean tides. The DInSAR-derived Point F using Sentinel-1a/b interferograms in 2018 sampled different GL positions with changes in ocean tides; however, it fails to capture the ephemeral grounding observed in this study (Fig. 9). This indicates that 1 year's worth of DInSAR data may not be fully adequate to address the migration of GL in different ocean tide amplitudes within a tidal cycle (Mohajerani et al., 2018).
By comparing the ICESat-2-derived Point F with ICESat-2 crossovers, as well as with several published GZ products on the Filchner–Ronne and Ross ice shelves, we are able to detect possible errors in different GZ products. The large landward deviations in the ESA CCI DInSAR-derived Point F on the western flank of Support Force Glacier in 2016 and the northern flank of Parry Peninsula in 2014, compared with all the other GZ products, indicate that the ESA CCI DInSAR-derived Point F locations are likely to be in error. A landward GL migration of up to 15 km was identified for ICESat-2-derived Point F at Crary Ice Rise compared with previous GL products, which is coincident with high mass loss in this region (Smith et al., 2020b), indicating that it can be a possible region of GL retreat. Further research is needed to fully understand the reason why the GL has been retreating in this region.
In highly crevassed and fast-flowing glaciers with low tidal amplitudes
(Padman et al., 2002), such as Pine Island Glacier
and Thwaites Glacier located in the Amundsen Sea embayment, it is difficult
to capture both Points F and H based on the dynamic method, which samples
elevation changes at different tidal phases using repeat-track analysis. The
fast movement of the glaciers can cause extensive advection of ice surface
features on the floating ice, such as crevasses and surface undulations
(Moholdt et al.,
2014; Khazendar et al., 2013). This
will result in high elevation anomalies not associated with ocean tides,
making it difficult to identify the limit of ice flexure. A Lagrangian
framework has been used to reduce the elevation change anomalies caused by
feature advection (Moholdt et al.,
2014; Dutrieux et al., 2013). This
method, however, requires the movement of ice features synchronized with the
ice flow, which is only applicable on floating ice shelves
(Marsh et al., 2016). Thus it is not
suitable for this study as we are only interested in the transition between
grounded ice and floating ice. Unlike the limit of tidal flexure Points F
and H that directly depend on the tidal variations, the break-in-slope point
is the location where the ice “feels” the bed sufficiently to react to the
stresses associated with this contact, and it is not influenced by the
temporal tidal variations
(Bindschadler et
al., 2011). Also the elevation differences measured by Points F and H are
always noisier than the absolute surface elevation measured by Point I
Compared with the ASAID break-in-slope delineation from Landsat-7 images,
ICESat-2-derived Point I
The dataset produced in this study is available at the University of Bristol
data repository, data.bris, at
We present the first ICESat-2-derived high-resolution Antarctic GZ product
using 18 months of data, including three GZ features (Li et al., 2021). This product has been derived using automated techniques
developed in this study based on ICESat-2 repeat tracks and has been
validated using a crossover analysis of ICESat-2 data over the
Filchner–Ronne and Ross ice shelves and against the recent DInSAR
measurements. A total of 21 346 Point F (the landward limit of ice flexure),
18 149 Point H (the inshore limit of hydrostatic equilibrium), and 36 765
Point I
Although our study period only covers 18 months, we are able to detect short-term GZ migration due to ocean tidal oscillation. Examples of repeat-track analysis in Dronning Maud Land and the Filchner–Ronne Ice Shelf show that the influence of ocean tide variations will not only change the grounding location of the ice but will also influence the point of full hydrostatic equilibrium for the floating ice. A more detailed analysis of the relationship between ocean tide variations, GZ width, and different geophysical factors is needed in the future. With more repeat cycles coming out in the next few years, we will be able to map the GZ features based on the same techniques developed in this study repeatedly and efficiently. This will allow for tracking GL migration at higher accuracy and provide more comprehensive insights into ice sheet instability, which is valuable for both the cryosphere and sea level science communities.
Spatial distributions of the absolute separations between the ICESat-2-derived landward limit of tidal flexure (Point F) and Sentinel-1a/b DInSAR-derived Point F in 2018 of individual regions in Table 2; the spatial extents of each region are shown as black boxes in Fig. 8. In all subplots, data are superimposed over the recent mass change map (Smith et al., 2020b); the blue line is the Sentinel-1a/b DInSAR-derived Point F in 2018 (Mohajerani et al., 2021), and the light-grey line is the IMBIE basin boundary (Shepherd et al., 2018; Rignot et al., 2011b).
Spatial distributions of the separations between the
ASAID-derived break in slope and ICESat-2-derived break in slope (Point I
Spatial distributions of the absolute separations between
the ICESat-2-derived break in slope (Point I
List of different grounding line (GL) products used to update the Depoorter et al. (2013) grounding line for the composite grounding line generated in Sect. 2.2.
TL developed the methods, produced the results, and wrote the paper. GJD and SJC assisted with data processing. JLB conceived the study and contributed to the interpretation of the results. All authors commented on the manuscript.
The contact author has declared that neither they nor their co-authors have any competing interests.
Publisher’s note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This article is part of the special issue “Extreme environment datasets for the three poles”. It is not associated with a conference.
We would like to thank the National Snow and Ice Data Center (NSIDC) for providing the ICESat-2 L3A land ice height data product. We greatly appreciate Pietro Milillo (University of Houston, USA) for providing the 2016–2017 grounding line dataset at Thwaites Glacier.
Tian Li received funding by the China Scholarship Council (CSC)–University of Bristol joint-funded PhD scholarship. Jonathan L. Bamber and Stephen J. Chuter received funding by the European Research Council (GlobalMass; grant no. 694188). Jonathan L. Bamber also received funding by the German Federal Ministry of Education and Research (BMBF) in the framework of the international future lab AI4EO (grant no. 01DD20001).
This paper was edited by Tao Che and reviewed by two anonymous referees.