Tabular icebergs in the Southern Ocean are primarily calved from three large ice shelves in Antarctica: the Ross (Lazzara et al., 1999;Joughin and MacAyeal, 2005), Filchner-Ronne (Scambos et al., 2005), and Amery Ice Shelves (Fricker et al., 2005). In the last two decades, ice shelves in the Antarctic Peninsula have become more unstable and experienced rapid calving (Scambos et al., 2004;Rignot et al., 2004;McGrath et al., 2012). After calving from an ice shelf, an iceberg starts its journey through the Southern Ocean, along the way playing a significant role in sea-ice formation (Massom et al., 2001(Massom et al., , 2018)), polynya occurrence (Robinson & Williams, 2012), dense shelf water export (Williams et al., 2010), and ocean primary production (Arrigo et al., 2002;Arrigo & van Dijken, 2003). An iceberg may even affect penguin colonies when grounded on a continental shelf (Kooyman et al., 2007). Driven by wind forcing, ocean currents, and the Coriolis force (Gladstone et al., 2001;Wagner et al., 2017), Antarctic icebergs usually drift vast distances around the continental shelf, occasionally, drifting north and eventually melting into the surrounding ocean (Merino et al., 2016;Silva et al., 2006). During drift, tabular Icebergs survive rapid surface melting (Jansen et al., 2005), basal melting (Russell, 1980;Jansen et al., 2007;Braakmann-Folgmann et al., 2022), ocean wave rush (MacAyeal et al., 2006), and ocean current shear (Huth et al., 2022). Eventually, they break up and release smaller icebergs (Wagner et al., 2014), hurrying up their melting into the ocean. Breakup is a natural process for tabular icebergs drifting in the ocean, but the breakup mechanism of tabular icebergs has yet to be fully understood.
Rapid change and breakup of tabular icebergs in the Southern Ocean are frequently monitored by remote sensing satellites that observe lateral changes in areal extent (Stuart and Long, 2010;Scambos et al., 2005;Braakmann-Folgmann et al., 2022).
Further combining altimetry measurements, vertical changes of tabular icebergs can be accurately quantified (Braakmann-Folgmann et al., 2022), providing important ancillary information to understand an iceberg breakup event. Breakup results from ice fracture and subsequent rift formation in response to stresses that cause crevasse propagation (Benn et al., 2007;Cuffey & Paterson, 2010;Alley et al. 2008). When stresses exerted on an iceberg exceed some threshold, fracture tends to occur and develop perpendicular to the maximum tensile stresses. As the highest frequency of remote sensing observation is at best daily, the rapid ice fracture and breakup process are difficult to capture by such. Because of a lack of observation data, studies have yet to be conducted to understand the breakup of tabular icebergs (Bouhier et al., 2018).
A calving event on the Larsen C Ice Shelf off the Antarctic Peninsula in July 2017 gave birth to one of the most enormous tabular icebergs ever recorded, named A68, with an area of almost ten percent of the ice shelf (Hogg & Gudmundsson, 2017). In early 2021, the main body of A68 (hereafter A68a) drifting in the Southern Ocean disintegrated and disappeared quickly in the ocean. Before its rapid vanish, several major breakup occurred (Braakmann-Folgmann et al., 2022) which may have played important role in its rapid disappearance. When drifting in the ocean, two major breakup of A68a when A68a drifting away from the South Georgia Island was attributed to the large shear forces caused by differences in ocean currents (Huth et al., 2022). When A68a melting in the ocean, it cooled down and diluted the surrounding ocean water (Smith and Bigg. 2023), causing deepening of the underlying water mass (Tarling et al. 2024), having significant influence on the surrounding marine ecosystem. However, the first breakup when A68a drifting close to the South Georgia Island was not fully understood, but likely to be caused by collision with seafloor (Huth et al., 2022). More investigation and evidence is required to better understand this breakup as this collision may have weaken the iceberg, causing its rapid melting into the surrounding ocean.
This research uses available multi-source remote sensing data to investigate the breakup of A68a when it drifted close to the South Georgia Island in the Southern Ocean. We begin by revealing its grounding status through remote sensing data. Then, we turn to an established ice-flow model to calculate the tensile stresses exerted on the iceberg under various drift scenarios while grounded. We uncover the breakup mechanism of the large tabular iceberg A68a due to its interaction with seafloor shoals on its drift across the Southern Ocean.
We use Sentinel-1A/B Extra Swath GRD data from July 2017 to February 2021 and MODIS image from December 2020 to May 2021 (Fig. 1, 2, Table S1 and S2 Cryosat-2 Synthetic Aperture Radar (SAR) Geophysical Data Record (GDR) data during November 1 to December 16, 2020 are used to extract the freeboard of A68a (Fig. 4).
The freeboard data were used to invert ice thickness by using firn air thickness data (Holland et al., 2011). Seafloor Bathymetry data from the General Bathymetric Chart of the Oceans (GEBCO) is used to investigate the grounding section of A68a (Fig 3b). Tide height data from CATS2008 (Padman et al. 2002)
In this study, we use remote sensing images to detect area, drifting trajectory, drifting and rotating speed of iceberg A68a first, and then use satellite altimetry to detect freeboard, invert thickness of A68a when the iceberg drifted close to the South Georgia Island. The mass of A68a before breaking up was predicted and the rift line where A68a calved along was extracted. Grounding status was detected using thickness and seafloor bathymetry data. Finally, using all parameters extracted above, Ua model was https://doi.org/10.5194/egusphere-2024-2790 Preprint. Discussion started: 11 November 2024 c Author(s) 2024. CC BY 4.0 License. set to investigate the deviatoric stresses exerted on iceberg A68a. The methods for all the processing are described as follows.
Part of the drifting trajectory of A68a from the Antarctic Iceberg Tracking Database until July 2019 is shown in Fig. 3a. Using Sentinel-1A/B GRD and MODIS data, the boundaries of iceberg A68a at different observation dates are extracted by human interpretation (Wang et al. 2016) using ArcGIS, and multiple polygon shapefiles are generated. Then, the centroid locations of all shapefiles are calculated using the Geometry Calculator with ArcGIS. Finally, the drifting trajectory of iceberg A68a between July 2017 and April 16, 2021, is generated using these centroid locations (black star in Fig. 3a). In Fig 3a, the red (A68 iceberg) and the grey (A68a iceberg) circle points are obtained from the Antarctic Iceberg Tracking Database (Budge & Long, 2018) . The black star points (A68a) are extracted from Sentinel-1 and MODIS satellites (Supp. Material). The observation dates of the A68a position are formatted as 'yyyymmdd' and illustrated in black. AP stands for Antarctic Peninsula. The grounding line of the Antarctic Ice Sheet is marked with a blue curve. The background is topography (Arndt et al. 2013), including the Antarctic Peninsula, South Georgia Island, and intervening the seafloor.
Insert Figure 3 The boundary polygon shapefiles extracted from remote sensing images are projected to an equal area projection (Lambert Azimuthal Equal Area) using ArcGIS and the area change of A68a is fitted with the least square method (Fig. 3c). The areas of derivative icebergs A68b, A68c, A68d, A68e, A68f, and A68g during several major breakup events are indicated with color circles. The accelerating areal-losing rates for three distinct periods (referred to as the first, second, and third phases of the iceberg's journey) are illustrated with grey line segments and the rates specified in km 2 /a.
The centroid location and the time interval of both observations of A68a are used to calculate the drifting velocity, assuming a linear drift of A68a. The drifting velocity of iceberg A68a is obtained by dividing the time interval with drifting distance. Besides drifting, A68a also rotated in a clockwise direction (Fig. 3b). It is assumed iceberg A68a rotated around the center of mass (centroid). The rotating velocity is obtained by following two steps: (1): centroid normalization, and (2): rotation angle extraction.
Centroid normalization is first conducted to make sure both boundary files have (0, 0) as https://doi.org/10.5194/egusphere-2024-2790 Preprint. Discussion started: 11 November 2024 c Author(s) 2024. CC BY 4.0 License. their centroid. Rotation angle calculation is decided by how many degrees are required to be rotated until the best match of both outlines is found. Then the angular velocity of iceberg A68a is calculated by dividing the drifting period with rotation angle.
Cryosat-2 measurements less than 55 m (Holland et al. 2011) over A68a can be identified virtually from five different tracks in Fig. 4 (A68a is the only large iceberg that appeared in this region). Following the method from Wang et al. (2016), the sea surface height and top of the iceberg are determined for each track, and mean freeboard of A68a from each track of Cryosat-2 is calculated by Eqn. 1.
(1)
where 𝐸 , 𝑆 and 𝐹 indicates the ice surface height, sea surface height and mean freeboard of iceberg A68a from track i of Cryosat-2.
We assume iceberg A68a has a uniform freeboard here. The final freeboard of A68a, which is 32.5± 1.0 m (1.0 m as the standard deviation of freeboard, Table S3) is calculated as the mean of all the five freeboards estimated (Eqn. 2). F indicates the mean freeboard of the iceberg A68a.
Insert MODIS observation on December 16, 2020 (Fig. 2e), was contaminated by clouds and not used to extract the outline of A68a. However, the A68a outline extracted from Sentinel-1A images (December 15, 2020) was shifted and rotated to match A68a shown in MODIS observation using the method introduced above. In this way the outline of A68a corresponding to MODIS observation time on December 16, 2020, is extracted.
The Sentinel-1A image on December 15 was georeferenced to MODIS observation time using 17 Ground Control Points (Table S4, Fig. 2e). The georeferencing reached an accuracy of 36.1 m.
In the same way, the outline of A68a extracted from Sentinel-1 image on December 21, 2020, is shifted and rotated back to position on MODIS observation time, December 16, 2020, by using 11 Ground Control Points (Table S5). The georeferencing reached an accuracy of 5.8 m. One rift line is extracted by comparing two different boundaries of A68a before and after the breakup (Fig 5 and 6a).
Image of A68d from Sentinel-1 on December 21, 2020, was georeferenced to MODIS observation time using five Ground Control Points (Table S6). The georeferencing reached an accuracy of 51.6 m. In this way, another rift line is extracted (Fig 5 and 6a).
Insert Figure 5 here
The firn layer thickness must be considered to invert ice thickness but is seldom known as no simultaneous measurement of the iceberg A68a is available. However, the firn air content of Larsen C ice shelf varied from 8.0 to18.0 m at the calving front (Holland et al. In this way, the thickness of iceberg A68a is inverted using Eqns. ( 3) and (4), which is 195.5 ± 19.1 m, and the ice draft is 163.0 ± 19.1 m.
The uncertainty of ice thickness inversion is calculated with Eqn. ( 5)
where 𝑑 and 𝑑 are the uncertainty of air layer thickness and freeboard of A68a, respectively.
Before the breakup, the area of A68a was close to 3825±89.0 km 2 (December 15, (3), the total mass of iceberg A68a was 642.2±69.0 Gt (1Gt=10 9 t).
The method for grounding detection introduced by Wang et al. ( 2016) is adopted in this study. It is also assumed that the iceberg is floating first, and compare the ice draft inverted from the freeboard with seafloor topography. The region with ice draft lower than seafloor bathymetry indicates grounding (negative comparison result in Fig. 6b-c).
A large negative value indicates a heavily grounding section of an iceberg. This process is performed by following Eqns. ( 6) and ( 7)
𝐸 and 𝐸 are the elevations of ice draft and seafloor bathymetry respectively. D is the elevation difference of ice drift and seafloor bathymetry.
Insert Figure 6 here
Ice tends to fracture under critical tensile stresses forming rifts along which ice breakup occurs. In this study, a 2d shallow-ice-shelf approximation glacier model Ua (Gudmundsson. 2008(Gudmundsson. , 2012(Gudmundsson. , 2013) ) is employed to investigate the deviatoric stresses when iceberg A68a interacts with seafloor shoals. Because A68a was drifting and https://doi.org/10.5194/egusphere-2024-2790 Preprint. Discussion started: 11 November 2024 c Author(s) 2024. CC BY 4.0 License.
rotating before colliding with the sea mountain, the drifting and angular velocity of A68a must be inverted first by exerting an appropriate drag force on the iceberg in Ua.
Since the drifting of large icebergs (length>10 km) is primarily affected by ocean currents (Wagner et al. 2017), it is assumed that the drifting and rotating of A68a was a consequence of ocean currents only. In Ua, the ocean drag term is applied by following Eqn. ( 8)
where 𝑡 stands for ocean drag, 𝒢 floating/grounding mask of the iceberg. The drag coefficient is indicated with 𝐶 , where 𝑣 and 𝑣 stand for ocean current and basal ice velocity, respectively.
In Ua, the freeboard and thickness extracted from Section 3.3 and 3.5 was used to configure the surface (32.5 m) and bottom (-163.0 m) elevation of iceberg A68a. The seafloor bathymetry was input to set the lower boundary for ice-seafloor interaction modeling. The temperature of iceberg was set as -20 °C. The parameter A and n in Glen flow law was set as 2.9377×10 -9 and 3 respectively. Surface and basal mass balance of the iceberg when interacting with the seafloor was set as 0 and neglected.
Since the drifting of large icebergs (i.e., length >10 km) is primarily affected by ocean currents (Wagner et al. 2017), an appropriate ocean drag parameterization is of critical importance in Ua. The ocean drag coefficient C0 was set as 0. The air drag was not considered in the ice-seafloor interaction modeling. Detailed setting of ice-seafloor interaction modeling in Ua can be found from Table 1.
Insert Table 1 here 4. Results
After the birth of A68a, it drifted north and finally disappeared in the Southern Ocean in May 2021 (Fig. 3, 2a-2d). The drifting trajectory and areal changes of A68a from remote sensing (Fig. 3a, 3c) indicate that iceberg A68a has lasted almost four years, from July 2017 to May 2021. The maximum latitude A68a reached was 52 °S (Fig. 3a, and Fig. 2a-2d). A68a experienced six major breakup events (Fig. 1). The total areal ice loss in generating nascent icebergs during these breakups was > 2000 km 2 (Fig. 3c), more than a third of the original iceberg A68a. This suggests that the decimation of a large tabular iceberg in Antarctica occurs through a sequence of breakups.
The lifespan of A68a shows three different areal-losing rates (Fig. 3c) during its evolution, which are -156 km 2 /a, -1204 km 2 /a, and -11,990 km 2 /a, sequentially. The increasing areal-losing rates demarcate three distinct evolutionary periods of this well-documented, large tabular iceberg adrift in the Southern Ocean. The ratio of areallosing rates for three different phases is approximately 1:8:80, which signals that once an iceberg has arrived at the third phase, it is well on its way to rapidly vanishing into the surrounding ocean. The survival time ratios of 6:2:1 and total ice loss ratios of 1:3:11 of the three different phases combine to indicate that the rate and volume of ice loss of the A68a iceberg accelerate over time.
A major breakup event ensued when A68a eventually drifted to South Georgia Island, and A68a gave birth to another iceberg, A68d (Fig. 1c, and 3b). Moderate Resolution
Imaging Spectroradiometer (MODIS) data show that the breakup event occurred on December 16-17, 2020 (Fig 2e , 2f). Sentinel-1A/B synthetic aperture radar data reveals that after the breakup, A68a decreased to ~3582±8 km 2, and the area of A68d was 147±1 km 2 (Fig. 3c). Using satellite observations from MODIS, Sentinel-1A/B, and adding in Cryosat-2 radar altimeter data (Fig. 4) the bulk properties of A68a were established during December 15-16, 2020. At that moment, it was ~195±19 m thick, weighed some ~642±69 Gt, drifted at a speed of 24±2 cm/s, and rotated with an angular velocity of ~15±2°/d, clockwise. Furthermore, the drifting speed during the previous two days, December 15-16, 2020, did not change much from the days preceding them, December 13-15, 2020 (Table . S7). This indicates the sizeable tabular iceberg had large inertia to sustain a relatively constant motion. Bathymetric data illustrates that in the vicinity of South Georgia Island, there is a seamount with a subsea elevation rising from about -250 m to the surface (Fig. 3b, 6a).
MODIS geospatial data (Fig. 2e, 6a) shows that A68a hovered over part of the seamount on December 16, 2020. The grounding status of A68a at this juncture in time has been assessed by assuming hydrostatic equilibrium, employing firn-layer thickness (Holland et al. 2011) and ice thickness inverted from remote sensing. The result indicates > 3 km 2 of the iceberg (18 pixels) was grounded on the seamount (Fig. 6b and 6c). At its most severely grounded point, the iceberg was estimated to be lifted about 30 m out of floatation. Grounding on the seamount resulted in a sudden basal drag on the rapidly drifting and rotating iceberg. A seafloor drag's abrupt appearance dramatically altered the dominant balance of forces of the wind and currents and the iceberg's internal stresses.
To visualize the rift along which A68a calved during December 16-17, 2020, boundary outlines of A68a and A68d, extracted from Sentinel-1A on December 21, are georeferenced to MODIS data on December 16 by matching ground control points (Fig 5d and 5e). The georeferenced icebergs reflect two rifts (named Rift 1, Rift2) on December 16 (Fig. 5e and 6a). During the breakup event, a region formed between the two rifts, and about 2 km 2, was lost. The timing of the breakup, as seen from remote sensing, suggests this modest loss was a contributing factor in the more significant breakup. Coincidently, analysis reveals that Rift 1 was overlaying the most severely grounded region (Fig. 6a and 6e), which suggests that the area is critical to the breakup of A68a.
Using the standard Ua parameter settings (Table . 1), the basal ice speed, ice surface elevation, and deviatoric stresses exerted on A68a are simulated in Fig. 7a-d.
Examining rifts extracted from remote sensing, the maximum surface elevation of A68a is observed to be located on Rift 1. Furthermore, the simulation indicates that the maximum tensile stresses appear over grounded regions, and the location of maximum tensile stresses is on Rift 1 (within <200 m, Fig. 7b-c). The rift observed from remote sensing crosses the simulated location of maximum tensile stresses exerted on iceberg A68a which support the theory from (Benn et al. 2007) that the tensile cracks tend to occur perpendicular to the maximum tensile stresses.
Insert Figure 7 here
The drag coefficient is set differently for iceberg and sea-ice studies (Bigg et al 1997;
Flato & Hibler 1992). Since Ua is a two-dimension glacier model and external forces exerted on the sidewall of an iceberg are not considered, we assume an iceberg is a thick piece of sea ice and only consider the basal drag applied to the iceberg in this study. Flato and Hibler (1992) suggested the ocean drag coefficient was 0.0055 and the ocean drag was quadratic to ocean current velocity, from which one can get the following settings: 𝑚 = 0.5, 𝐶 =0.4 (unit: (m/s)/Pa). . This result indicates that the setting about 𝐶 =0.4 is not appropriate because no significant drag at the boundary would occur when putting an iceberg in a still ocean. The variation of maximum deviatoric tensile stresses exerted on A68a is shown in Fig. 8 when is set to various values.
Insert Figure 8 here Without more observation data related to iceberg drifting, it is difficult to determine the values of 𝐶 accurately. However, Fig. 8i indicate that smaller setting of 𝐶 leads to larger stresses ocean drag exerted iceberg and larger tensile stresses at the boundary of the iceberg. When 𝐶 >0.4×10 4 , the maximum tensile stresses exerted on A68a in the still ocean is approaching that from diagnostic runs without applying ocean drags.
Based on this, 𝐶 =0.4×10 4 is assessed as an appropriate setting since neither significant direction change of deviatoric forces occurred inside of the iceberg A68a, nor large tensile stresses at the boundary.
The Ua glacier model is used to investigate further the relationship of maximum tensile stresses to grounding, drifting, and rotation of A68a. As a sensitivity experiment, the model seamount is lowered by 50 m so that iceberg A68a experiences no drag in the vicinity of the seamount. Tensile stresses no more significant than 0.1 MPa are simulated inside the iceberg (Fig. 9). In the next experiment, restoring the seamount to its observed height, with A68a drifting with a speed consistent with observations from remote sensing but with no rotation, the maximum tensile stress jumps to 9 MPa over the grounding region (Fig. 7d). An increase of almost a factor of one hundred. When further incorporating the rotation of A68a, the maximum tensile stress increases to 10
MPa (Fig. 7d), a further twenty percent increase. These sensitivity experiments demonstrate that grounding a rapidly drifting and rotating iceberg on a seafloor shoal leads to significant tensile stresses inside the iceberg. Since icebergs tend to fracture under significant tensile stresses, and the rifts from remote sensing and maximum tensile stresses from glacier modeling coincide, it is concluded that substantial tensile stresses associated with grounding triggered the breakup of A68a.
Insert Figure 9 here
Modeling results indicate that grounding only A68a on the seamount, with no motion, would generate maximum tensile stress < 0.2 MPa, even considering seafloor changes by ±20 m. However, the maximum tensile stress increases to about seventy times when additionally applying half of the drifting speed observed from remote sensing, reaching eighty times than grounding only when full drifting speed is simulated. Simulations suggested that larger tensile stresses could occur in A68a when further raising the seafloor, thus simulating more severe grounding. When lowering the seafloor by 20 m (other settings the same as Table 1), the maximum tensile stress was about 6.0 MPa.
When raising the seafloor by 20 m, the maximum tensile stresses experienced by A68a increased to > 20 MPa. This confirms that grounding an iceberg with a considerable drifting speed generates significant tensile stresses inside the grounded section of the iceberg. Less obvious, perhaps, more severe grounding non-linearly increases the maximum tensile stresses.
The rotation of the iceberg is another factor influencing the maximum tensile stresses.
When iceberg A68a is simulated to rotate as observed by remote sensing data (15.1°/d, clockwise direction, other settings the same as Table 1), a maximum tensile stress > 10
MPa appears over the grounding region of A68a. Modeling results indicate that for the same seafloor bathymetry, more significant angular velocity exerts more considerable tensile stresses on A68a. Using all data from different grounding statuses of A68a, the maximum tensile stresses increase by about 10, 20, and 30 percent when additional angular velocity at 8 °/d, 15°/d, and 23°/d are applied. In short, rotation enlarged the maximum tensile stresses experienced in A68a by about twenty percent, contributing to the breakup during December 16-17, 2020. averaging over the region from Latitude: -55° to -56°, to longitude: -37.25° to -37.75°
(iceberg location seen in Fig. 10b) and is indicated with blue curves. This region's standard deviation of pressure is used as the error bar. When the polar cyclone moved close to A68a, the mean sea level pressure dropped about 30 hPa (Fig. 10e), which could significantly raise the sea surface by approximately 0.3 m (Fig. 10f) and facilitate A68a climbing up to a higher region of the seamount.
Large tabular icebergs experience breakup while drifting in the ocean, ultimately melting and vanishing into the ocean, the destiny of all icebergs. The breakup generates many smaller icebergs and increases the area for iceberg-ocean interaction, speeding up the iceberg melting. Seafloor shoals, such as a seamount, lead to iceberg grounding and are a barrier to iceberg drift. When an iceberg drifts toward a seafloor shoal with a relatively large speed, if the seafloor is higher than the iceberg draft, a grounding event occurs, even despite an iceberg having a large mass and inertia to maintain its speed.
Grounding of rapidly drifting and rotating icebergs tends to create significant deviatoric stresses inside an iceberg, which may trigger the formation of rifts and facilitate iceberg breakup.
For an iceberg that drifts towards a seafloor shoal, the sea surface change caused by a rising tide and weather events, such as a polar cyclone, may make the iceberg drift to a higher region of the seafloor shoal. After the passing of a polar cyclone and during low tide, a follow-on significant sea surface drop can make iceberg grounding more severe on the seafloor shoal, generating even more enormous tensile stresses in the iceberg and facilitating the breakup of the iceberg across the maximum tensile stress point.
The later, rapid vanishing of iceberg A68a into the ocean is fully explained by the breakup event that occurred on December 16, 2020, caused by the collision of A68a with a seamount. We demonstrate the first successful case of using a large tabular iceberg as a natural laboratory, with remote sensing observations as a data source and (a) (b) (c) (d) (a) (b) (c) (c) (a) (b) (a) (b) (c) (d) (e) (f)
https://doi.org/10.5194/egusphere-2024-2790 Preprint. Discussion started: 11 November 2024 c Author(s) 2024. CC BY 4.0 License.