remote Grounding Event of Iceberg D28 and Its Interactions with Seabed Topography

: Iceberg D28, a giant tabular iceberg that calved from Amery Ice Shelf in September 2019, grounded off Kemp Coast, East Antarctica, from August to September of 2020. The motion of the iceberg is characterized herein by time-series images captured by synthetic aperture radar (SAR) on Sentinel-1 and the moderate resolution imaging spectroradiometer (MODIS) boarded on Terra from 6 August to 15 September 2020. The thickness of iceberg D28 was estimated by utilizing data from altimeters on Cryosat-2, Sentinel-3, and ICESat-2. By using the iceberg draft and grounding point locations inferred from its motion, the maximum water depths at grounding points were determined, varying from 221.72 ± 21.77 m to 269.42 ± 25.66 m. The largest disagreements in seabed elevation inferred from the grounded iceberg and terrain models from the Bedmap2 and BedMachine datasets were over 570 m and 350 m, respectively, indicating a more complicated submarine topography in the study area than that presented by the existing seabed terrain models. Wind and sea water velocities from reanalysis products imply that the driving force from sea water is a more dominant factor than the wind in propelling iceberg D28 during its grounding, which is consistent with previous ﬁndings on iceberg dynamics.


Introduction
Grounding is a common process during the life course of an iceberg. When an iceberg drifts into shallow water areas, its keel cuts into sediments; then, a grounding event occurs. The process can last from days to years [1,2], during which the iceberg scours on the seabed under multiple environmental factors, such as the Coriolis force, wind, and currents around it [3]. The scouring creates plough marks that can provide insight into oceanic and glaciological conditions, revealing ocean-ice interaction procedures [4][5][6]. Investigation of continental land shelves and submarine plateaus presents plow marks of mega icebergs in the Late Pleistocene age indicative of submarine landform processes [7]. A recent study implies that iceberg grounding can trigger submarine landslides on certain occasions and can become hazards to coastal regions thousands of kilometers away [8]. In addition to modifying submarine topography, grounded icebergs can damage benthic communities through their interactions with the seabed, supported by in situ observations of several benthic habitats in polar areas [9][10][11][12].
Environmental factors control the motion of grounded icebergs in ways similar to determining their drifting pattern, but with the constraints of seabed topography [2,3,13].

Remote Sensing Data 2.2.1. Satellite Images
From 6 August to 15 September 2020, eight Sentinel-1 SAR images and three Terra MODIS images were collected to observe the iceberg. Sentinel-1 is a two-satellite constellation with C-band SAR on board, including Sentinel-1A and 1B [34]. Each satellite has a temporal resolution of 12 days, and the constellation can observe the same area every six days since they follow the same ground track. Every Sentinel-1 SAR image used was obtained with HH polarization in Extra Wide swath mode, with a spatial resolution of  20 × 40 m (range × azimuth). MODIS is embedded on a two-star constellation consisting of Terra and Aqua [35]. In our study, the 250 m-resolution images recording calibrated earth surface radiance (MOD02QKM) from MODIS Band 1 (centered at 645 nm) were served as a complement to Sentinel-1 images for a better temporal resolution. Since the backscatter coefficient and surface reflective property of ice differ from other surfaces, the iceberg shows a notable contrast brightness with its surrounding sea-ice and water. Thus, the boundary of iceberg D28 can be delineated manually.

Altimetry Data
Data from altimeters were used to get a precise estimation of the draft of iceberg D28. Considering the coverage and accessibilities of all altimeters, we used data from three satellites: Cryosat-2, Sentinel-3, and ICESat-2.
Cryosat-2 carries a SAR Interferometric Radar Altimeter (SIRAL) as a primary instrument to monitor variations in the thickness of the Earth's marine ice cover and continental ice sheets [36]. Baseline-D ice sheet elevation data from level 2 products were used in our study. Ice Baseline-D is the newest version of its processer, providing surface height over land ice with better accuracy than previous processers when compared with a reference elevation model over Antarctica [37]. The along-track spacing of Cryosat-2 level 2 data is about 350 m, and the cross-track spacing is about 2.5 km on average. Sentinel-3 is a two-star constellation with the Sentinel Radar Altimeter (SRAL) centered at Ku-band (13.575 GHz, bandwidth 350 MHz) onboard, providing data with an along-track resolution of about 300 m [38]. Similar to Cryosat-2 SIRAL, there is a continuously updating processer to estimate the surface elevation of land ice. The data used in our study were produced by the latest processer introduced in December 2020.
ICESat-2 has the Advanced Topographic Laser Altimeter System (ATLAS), consisting of two lasers at 532 nm [39]. The lasers send 10,000 pulses per second and take measurements every 0.7 m (2.3 ft) along the track. ICEsat-2 provides products with different retrackers to measure the elevations of various surfaces. ATL06 product estimates the median surface elevation of every 40 m-segment along track over land ice.
In all, we utilized Cryosat-2 Level 2 Baseline-D products, Sentinel-3 Level 2 LAN products, and ICESat-2 ATL06 products obtained from September 2018 to September 2019, which is a year before the calving event, to calculate the freeboard and draft of the origin area of iceberg D28. It can be regarded as an estimation of the iceberg thickness assuming that there was no significant thickness change during its first-year drift [40].

Environmental Parameters
To analyze the motion of iceberg D28 in the grounding area, wind and current forces were taken into consideration. The wind velocity of the study area in August and September of 2020 is was extracted from ERA5 hourly reanalysis data. ERA5 is the fifth generation of the European Centre for Medium-Range Weather Forecasts (ECMWF) reanalysis for global climate and weather [41]. There are four main subsets: hourly and monthly products-both on pressure levels (upper airfields) and single levels (atmospheric, ocean-wave, and landsurface quantities). Here, we used the 10 m-u/v components on a single level to represent the propelling force by wind. The data were re-gridded into regular latitude/longitude grids of 0.25 degrees.
The current force is represented by sea water velocities in both the east and north directions from the Coupled Model Intercomparison Project (CMIP6) Hi-res Ocean Models [42]. The grid size of the models was 1/12 degrees, and the variables were organized into standard NetCDF files.
Seabed elevation models of the study area were from Bedmap2 and BedMachine datasets. Bedmap2 is a suite of gridded products comprising the surface elevation, the ice thickness, the sea floor, and the subglacial elevation of Antarctica south of 60 • S [32]. BedMachine contains a bed topography map of Antarctica based on mass conservation [43]. The seabed elevation model from BedMap2 data has a spatial resolution of 1 km, and that from BedMachine data has a finer resolution of 500 m.

Iceberg Draft from Satellite Altimetry
The thickness of iceberg D28 can be estimated from the knowledge of its origin, assuming that there was no significant change in the thickness during its first-year drift. The same approximation is used by previous studies on iceberg keel and its impacts on the seabed [40]. Thus, if the freeboard and density model of the front part on Amery Ice Shelf are available, thickness can be calculated through the hypothesis of hydrostatic equilibrium.
The draft of iceberg D28 during the grounding event cannot be directly calculated using the assumption of hydrostatic equilibrium in the same way as for a floating iceberg, since part of the bottom was supported by the seabed. However, slightly grounded icebergs show a similar relationship between thickness and draft to floating ones [18]. Moreover, according to the principles of buoyancy, the water depth at the grounding points of an iceberg in touch with the seabed cannot be larger than the draft estimated with the presumption of floating. Thus, we calculated the draft of iceberg D28 through the hydrostatic equilibrium theory and treated the results as maximum water depths at grounding points.
Altimetry profiles collected from Cryosat-2, Sentinel-3, and ICESat-2 level 2 products provided surface elevations over Amery Ice Shelf from September 2018 to September 2019, in the format of time, height, longitude, and latitude, together with essential geophysical correction parameters. Ocean tides of all the altimetry data were corrected with the CATS2008 tidal model [44]. Considering the rapid ice flow of the front part of the Amery Ice Shelf, the altimetry ground points were corrected to simulated locations on 20 September 2019 with the ice flow velocity from MEaSUREs InSAR-based Antarctica Ice Velocity Map, Version 2 [45,46]. To reduce the uncertainties induced by the rapid propagation of the rift system, data within 1 km away from the edge of the D28 shape were discarded. All the altimetry data used are presented in Figure 2a-c. The ground tracks of three altimeters can supplement each other and make up some data gaps indicated by the distribution.
Seabed elevation models of the study area were from Bedmap2 and BedMachine datasets. Bedmap2 is a suite of gridded products comprising the surface elevation, the ice thickness, the sea floor, and the subglacial elevation of Antarctica south of 60°S [32]. Bed-Machine contains a bed topography map of Antarctica based on mass conservation [43]. The seabed elevation model from BedMap2 data has a spatial resolution of 1 km, and that from BedMachine data has a finer resolution of 500 m.

Iceberg Draft from Satellite Altimetry
The thickness of iceberg D28 can be estimated from the knowledge of its origin, assuming that there was no significant change in the thickness during its first-year drift. The same approximation is used by previous studies on iceberg keel and its impacts on the seabed [40]. Thus, if the freeboard and density model of the front part on Amery Ice Shelf are available, thickness can be calculated through the hypothesis of hydrostatic equilibrium.
The draft of iceberg D28 during the grounding event cannot be directly calculated using the assumption of hydrostatic equilibrium in the same way as for a floating iceberg, since part of the bottom was supported by the seabed. However, slightly grounded icebergs show a similar relationship between thickness and draft to floating ones [18]. Moreover, according to the principles of buoyancy, the water depth at the grounding points of an iceberg in touch with the seabed cannot be larger than the draft estimated with the presumption of floating. Thus, we calculated the draft of iceberg D28 through the hydrostatic equilibrium theory and treated the results as maximum water depths at grounding points.
Altimetry profiles collected from Cryosat-2, Sentinel-3, and ICESat-2 level 2 products provided surface elevations over Amery Ice Shelf from September 2018 to September 2019, in the format of time, height, longitude, and latitude, together with essential geophysical correction parameters. Ocean tides of all the altimetry data were corrected with the CATS2008 tidal model [44]. Considering the rapid ice flow of the front part of the Amery Ice Shelf, the altimetry ground points were corrected to simulated locations on 20 September 2019 with the ice flow velocity from MEaSUREs InSAR-based Antarctica Ice Velocity Map, Version 2 [45,46]. To reduce the uncertainties induced by the rapid propagation of the rift system, data within 1 km away from the edge of the D28 shape were discarded. All the altimetry data used are presented in Figure 2a-c. The ground tracks of three altimeters can supplement each other and make up some data gaps indicated by the distribution. Prior to ice-flow velocity correction and calculating the freeboard of Amery Front, cross validations of the three altimeters were carried out. Since the average ice flow velocity of Amery Front was about 3 m/day, we regarded data obtained within 11 days as nearcoincident observations. The restriction of 11 days can ensure that the distance brought by ice flow was around 40 m, which is the along-track spacing of ICESat-2 ATL06 products. We compared the synchronized observations from different altimeters with a distance less than 350 m and 300 m, with and without Cryosat-2 data, respectively. Then, the average of their differences (Diff) and standard deviations of differences (Std) were calculated. The results are shown in Table 1. The average differences of data from different altimeters were less than 1 m, allowing for the assimilation of data from three altimeters without additional corrections. The largest standard deviation of differences between altimeters, which was 1.59 m, was regarded as the uncertainty in surface elevation estimation. Table 1. Cross validation of the three altimeters on Cryosat-2, Sentinel-3, and ICESat-2 over Amery Front before the calving event.

Number of Points Diff (m) Std (m)
The surface topographic map was interpolated by merging data from three altimeters. Freeboard was then calculated by Equation (1) with the EIGEN-6C4 model [47] and the DTU15MDT model [48] to represent mean sea level [49].
where H sur f ace is the surface elevation; H f is the Freeboard; Geoid is from the EIGEN-6C4 model; and H MDT is the mean dynamic topography from the DTU15MDT model. The draft of the Amery Front was estimated through the assumption of hydrostatic equilibrium with a freeboard map and a density model. According to the in situ survey on Amery Ice Shelf, ice density varies linearly from 890.5 kg/m 3 (at calving front) to 903.5 kg/m 3 (315 km away from the calving front along the direction of ice flow) [50]. The density of the snow layer was set to 355 kg/m 3 as Weinhart put forward in the study of the snow layer properties on East Antarctic Plateau [51]. The density of water around Amery Ice Shelf was 1028 kg/m 3 [52]. Thus, ice thickness and draft can be calculated by the buoyancy equation: where H s is the thickness of the snow layer; H is the iceberg thickness; H f is the freeboard; H d is the iceberg draft; ρ s is the snow density; ρ i is the ice density; and ρ w is the water density. The surface snow depth was obtained from the Modern-Era Retrospective analysis for Research and Applications, Version 2 (MERRA-2). To simplify calculations, the thickness of the snow layer was set to a constant value of 3.97 m over the entire front part of Amery Ice Shelf.

Determination of Grounding Points
Iceberg D28 did pinwheel-like movement during its grounding off Kemp Coast from 6 August to 15 September 2020, shown by its boundaries delineated from Sentinel-1 and MODIS images (Figure 1b). Similar motion of iceberg C16 grounded in Ross Sea was recorded by a GPS receiver set on its surface in 2002 [2]. If an iceberg is grounded and has only one grounding point with the seabed, then the iceberg will move like a pinwheel under environmental forces. That means some parts of the iceberg should keep nearly stationary, and at the farther end, the drift speed is relatively large. Thus, according to the features of such a movement, we put up a new method utilizing remote sensing images of multiple sources to precisely locate the grounding points of iceberg D28.
Considering the availability of satellite images, we divided the study period into ten four-day intervals to describe the iceberg's movement in detail. To calculate the moving speed of different parts of iceberg D28, we made a fishnet with 1 km grid cells over its surface and carried out the geo-registrations of the iceberg shapes obtained every four days to locate the identical cell on different dates. Then, we measured the distance of grid cell centers and calculated the moving speed of different parts. The method diagram is shown in Figure 3. The red cells are the same cell on different dates. The uncertainty depends on the delineation of iceberg shapes and geometric registration of the shapes.
Iceberg D28 did pinwheel-like movement during its grounding off Kemp Coast from 6 August to 15 September 2020, shown by its boundaries delineated from Sentinel-1 and MODIS images (Figure 1b). Similar motion of iceberg C16 grounded in Ross Sea was recorded by a GPS receiver set on its surface in 2002 [2]. If an iceberg is grounded and has only one grounding point with the seabed, then the iceberg will move like a pinwheel under environmental forces. That means some parts of the iceberg should keep nearly stationary, and at the farther end, the drift speed is relatively large. Thus, according to the features of such a movement, we put up a new method utilizing remote sensing images of multiple sources to precisely locate the grounding points of iceberg D28.
Considering the availability of satellite images, we divided the study period into ten four-day intervals to describe the iceberg's movement in detail. To calculate the moving speed of different parts of iceberg D28, we made a fishnet with 1 km grid cells over its surface and carried out the geo-registrations of the iceberg shapes obtained every four days to locate the identical cell on different dates. Then, we measured the distance of grid cell centers and calculated the moving speed of different parts. The method diagram is shown in Figure 3. The red cells are the same cell on different dates. The uncertainty depends on the delineation of iceberg shapes and geometric registration of the shapes.

Wind and Sea Water Velocities
The mean wind field near Kemp Coast was derived from the ERA5 hourly 10m-u/v wind velocity components from 6 August to 15 September 2020. Sea water velocities at different depths in the north and east directions were from output variables of CMIP6 Hires models. The output variables consisted of 75 vertical layers presenting sea water velocities from the surface to a depth of 6000 m. Considering an average iceberg draft of about 250 m (Figure 4b), level 1(depth = 0.5 m) to level 33 (depth = 271.4 m) data were used to generate the mean velocity of the water bulk affecting the motion of iceberg D28 during the study period.

Iceberg Thickness and Draft
The thickness and draft of iceberg D28 in the grounding area are shown in Figure 4. Ice volume does not distribute evenly over the entire iceberg. The ocean-terminated part was thinner than the part adjacent to the ice shelf. The highest and lowest points had a thickness difference of about 120 m. The topographic relieves illustrated by the thickness

Wind and Sea Water Velocities
The mean wind field near Kemp Coast was derived from the ERA5 hourly 10m-u/v wind velocity components from 6 August to 15 September 2020. Sea water velocities at different depths in the north and east directions were from output variables of CMIP6 Hi-res models. The output variables consisted of 75 vertical layers presenting sea water velocities from the surface to a depth of 6000 m. Considering an average iceberg draft of about 250 m (Figure 4b), level 1(depth = 0.5 m) to level 33 (depth = 271.4 m) data were used to generate the mean velocity of the water bulk affecting the motion of iceberg D28 during the study period.

Location of Grounding Points
With the method mentioned in Section 3.2, we can get the moving speed of different parts of the iceberg every four days during the study period ( Table 2). The uncertainty was induced by outline delineation from satellite images. The images obtained on 14 August-26 August and 7 September were from MODIS Band 1, which have a spatial resolu-

Iceberg Thickness and Draft
The thickness and draft of iceberg D28 in the grounding area are shown in Figure 4. Ice volume does not distribute evenly over the entire iceberg. The ocean-terminated part was thinner than the part adjacent to the ice shelf. The highest and lowest points had a thickness difference of about 120 m. The topographic relieves illustrated by the thickness map can also be found on the Sentinel-1 SAR image on 28 December 2019 (Figure 4c). The iceberg draft map shows a similar ice volume distribution to the thickness map.

Location of Grounding Points
With the method mentioned in Section 3.2, we can get the moving speed of different parts of the iceberg every four days during the study period ( Table 2). The uncertainty was induced by outline delineation from satellite images. The images obtained on 14 August-26 August and 7 September were from MODIS Band 1, which have a spatial resolution of 250 m, and the others were Sentinel-1 SAR images with a resolution of 40 m. We took the uncertainty in the delineation of the iceberg boundaries as one pixel. Then, the uncertainty of the moving speed can be determined by adding the uncertainties in retrieving both boundaries on day 1 and day 4 and dividing this between four days. The results are shown in Table 2. During two periods, 10 August to 18 August and 26 August to 15 September, the minimum moving speeds were almost 0 m/day considering the calculation uncertainty. Meanwhile, the maximum speeds were over 500 m/day. The results were in accord with the characteristics of pinwheel-like movement, indicating that iceberg D28 had only one grounding part with the seabed. Seven grounding points were inferred from the locations of grid cells in these two periods, which contained seven four-day intervals ( Figure 5). In the other four-day intervals, 6 August to 10 August and 18 August to 26 August, the speed of different parts of the iceberg did not vary by much. Meanwhile, the drifting speed kept below 1 km/day, which is rather small compared with the free-drifting speed of a giant iceberg, implying the iceberg scoured on the seabed. With the locations of the grounding points, the maximum water depths were determined through the iceberg draft map (Figure 4b). The uncertainties of estimating iceberg draft are discussed in Section 5.1.
The water depths from seabed elevation models were calculated by subtracting the seabed elevation from mean surface height (H mss ). Gaps as large as hundreds of meters were found through the water depths estimated by the grounding event, BedMap2 data, and BedMachine data ( Figure 5, Table 3). Only the water depth at grounding point 3 from BedMap2 data established a shallower water than the iceberg draft. The differences implied the slope of the continental shelf off Kemp Coast is not as smooth as it seemed from the seabed maps. The diameter of the circumcircle of the grounding points was 20.56 km, indicating that the inaccuracies of existing seabed elevation data are common phenomena in the study area. There are supposed to be undetected banks or ridges that can impact the mesoscale characteristics of the slope current.
Remote Sens. 2022, 14, x FOR PEER REVIEW 9 of 17 points, the maximum water depths were determined through the iceberg draft map (Figure 4b). The uncertainties of estimating iceberg draft are discussed in Section 5.1. The water depths from seabed elevation models were calculated by subtracting the seabed elevation from mean surface height ( ). Gaps as large as hundreds of meters were found through the water depths estimated by the grounding event, BedMap2 data, and BedMachine data ( Figure 5, Table 3). Only the water depth at grounding point 3 from BedMap2 data established a shallower water than the iceberg draft. The differences implied the slope of the continental shelf off Kemp Coast is not as smooth as it seemed from the seabed maps. The diameter of the circumcircle of the grounding points was 20.56 km, indicating that the inaccuracies of existing seabed elevation data are common phenomena in the study area. There are supposed to be undetected banks or ridges that can impact the mesoscale characteristics of the slope current.

Wind and Sea Water Velocities of the Study Area
The mean wind velocity of the study area from 6 August to 15 September and its hourly trend is established in Figure 6a,b. The wind blew from the north in the study area, nearly perpendicular to the linear velocity direction of the iceberg. The wind speed peaked on August 9 at over 30 m/s for an instant, while the average speed was about 3 m/s at the iceberg location. The mean sea water speed map is shown in Figure 7a,b, presenting a water speed near to 0 at the grounding points. Meanwhile, its farther end was surrounded by fast running water. With a larger proportion of the iceberg volume proceeding into the area with fast sea water velocity, it moved a longer distance in the first 15 days of September than nearly a whole month of August.
peaked on August 9 at over 30 m/s for an instant, while the average speed was ab m/s at the iceberg location. The mean sea water speed map is shown in Figure 7a,b senting a water speed near to 0 at the grounding points. Meanwhile, its farther end surrounded by fast running water. With a larger proportion of the iceberg volume ceeding into the area with fast sea water velocity, it moved a longer distance in the 15 days of September than nearly a whole month of August.
In most situations, the water force is considered to be more dominant than the by wind for icebergs longer than 12 km [13]. During the study period, the wind field Kemp Coast illustrated a force with a direction perpendicular to the linear velocity d tion of the iceberg. Meanwhile, the sea water velocities did not change much with during the study period and provided a stable propelling force due to the differen sea water velocities in the grounding area and the farther end of the iceberg. Thus, w summarize that the driving force from the sea water was a more important factor tha wind dominating the iceberg motion during the grounding event. This inference is i with a previous study on iceberg dynamics [13,22,53].

Iceberg Draft Estimation
Several assumptions were made in estimating the iceberg draft in this study. First, the iceberg thickness did not change significantly throughout the first year after the calving event. Second, the density model of iceberg D28 was the same as that of its origin on Amery Ice Shelf. Finally, Iceberg D28 was cylindrically shaped with the same areas at its surface and the bottom.
The uncertainty from each assumption is discussed separately. 1. According to previous studies on iceberg evolution, processes affecting the iceberg thickness include snow accumulation and basal melting [54]. Snow accumulation can cause an increase in iceberg thickness. Meanwhile, basal melting accounts for 95% of the decrease, and surface melting, together with strain thinning, contributes the other 5% of the thickness change as simulations on large tabular icebergs [55]. Snow accumulation can be estimated with ERA5 reanalysis products by subtracting snowmelt and snow evapora- In most situations, the water force is considered to be more dominant than the force by wind for icebergs longer than 12 km [13]. During the study period, the wind field near Kemp Coast illustrated a force with a direction perpendicular to the linear velocity direction of the iceberg. Meanwhile, the sea water velocities did not change much with time during the study period and provided a stable propelling force due to the difference in sea water velocities in the grounding area and the farther end of the iceberg. Thus, we can summarize that the driving force from the sea water was a more important factor than the wind dominating the iceberg motion during the grounding event. This inference is in line with a previous study on iceberg dynamics [13,22,53].

Iceberg Draft Estimation
Several assumptions were made in estimating the iceberg draft in this study. First, the iceberg thickness did not change significantly throughout the first year after the calving event. Second, the density model of iceberg D28 was the same as that of its origin on Amery Ice Shelf. Finally, Iceberg D28 was cylindrically shaped with the same areas at its surface and the bottom.
The uncertainty from each assumption is discussed separately. 1. According to previous studies on iceberg evolution, processes affecting the iceberg thickness include snow accumulation and basal melting [54]. Snow accumulation can cause an increase in iceberg thickness. Meanwhile, basal melting accounts for 95% of the decrease, and surface melting, together with strain thinning, contributes the other 5% of the thickness change as simulations on large tabular icebergs [55]. Snow accumulation can be estimated with ERA5 reanalysis products by subtracting snowmelt and snow evaporation from snowfall. Hourly snow accumulation was aggregated from October 2019 to September 2020, resulting in 102 mm of snow water equivalent over the iceberg, which equals a snow layer of about 2.87 m thick taking the snow density as 355 kg/m 3 (Figure 8). It was insignificant compared to the total thickness of iceberg D28 (Figure 4a), hence the neglect of its impact on thickness change was acceptable.

Iceberg Draft Estimation
Several assumptions were made in estimating the iceberg draft in this study. First, the iceberg thickness did not change significantly throughout the first year after the calving event. Second, the density model of iceberg D28 was the same as that of its origin on Amery Ice Shelf. Finally, Iceberg D28 was cylindrically shaped with the same areas at its surface and the bottom.
The uncertainty from each assumption is discussed separately. 1. According to previous studies on iceberg evolution, processes affecting the iceberg thickness include snow accumulation and basal melting [54]. Snow accumulation can cause an increase in iceberg thickness. Meanwhile, basal melting accounts for 95% of the decrease, and surface melting, together with strain thinning, contributes the other 5% of the thickness change as simulations on large tabular icebergs [55]. Snow accumulation can be estimated with ERA5 reanalysis products by subtracting snowmelt and snow evaporation from snowfall. Hourly snow accumulation was aggregated from October 2019 to September 2020, resulting in 102 mm of snow water equivalent over the iceberg, which equals a snow layer of about 2.87 m thick taking the snow density as 355 kg/m 3 (Figure 8). It was insignificant compared to the total thickness of iceberg D28 (Figure 4a), hence the neglect of its impact on thickness change was acceptable.  Influences of basal melting on iceberg thickness have been evaluated by many researchers. The thickness of iceberg A68 decreased by 12.89 ± 3.34 m per year in its first 1.5 years of drifting [22]. The melting rate of iceberg A38B varied from 1.2 m per year to 10.8 m per year with its trajectory from Ronne Ice Shelf to the warm sea water near the Antarctic Peninsula [56]. Iceberg B30 had an average thinning rate of 17.3 ± 1.8 m per year estimated with satellite altimetry data in 6.5 years [23]. Melting rates of icebergs between 60 and 150 • E were 11-18 m per year from ship-based observations over 15 years [57]. Though the iceberg melting rate varies dramatically in different regions in the Southern Ocean with water temperature and latitude, we inferred that the thickness of iceberg D28 decreased less than 20 m during the first year after its appearance, since it drifted at a high latitude (Figure 1a). A deviation of 20 m would not change our findings about seabed topography, considering that the maximum water depth estimated through a grounding event has an average disagreement of hundreds of meters with the existing seabed elevation models.
The assumption can be verified with altimeter data. The altimeter ground tracks were mapped onto the iceberg surfaces from near-coincident SAR images (Figure 9a,c,e). We compared the freeboard derived from altimeter ground tracks over the iceberg with its initial surface topography from August 2020 to September 2020 and calculated the differences. The results are shown in Figure 9. The freeboard on 6 August was calculated from ICESat-2 ATL03 product downscaled to 350 m by the algorithm introduced in Algorithm Theoretical Basis Document for ATL06 [58]. Freeboards for 10 August and 25 September were derived from Sentinel-3 SRAL level 2 product. The mean freeboard differences were 0.36 m, 0.96 m, and 0.68 m, respectively. Regardless of the mismatch of surface terrain at some points, the results indicated little freeboard and thickness change within the first year of D28 s drifting. event has an average disagreement of hundreds of meters with the existing seabed elevation models.
The assumption can be verified with altimeter data. The altimeter ground tracks were mapped onto the iceberg surfaces from near-coincident SAR images (Figure 9a,c,e). We compared the freeboard derived from altimeter ground tracks over the iceberg with its initial surface topography from August 2020 to September 2020 and calculated the differences. The results are shown in Figure 9. The freeboard on 6 August was calculated from ICESat-2 ATL03 product downscaled to 350 m by the algorithm introduced in Algorithm Theoretical Basis Document for ATL06 [58]. Freeboards for 10 August and 25 September were derived from Sentinel-3 SRAL level 2 product. The mean freeboard differences were 0.36 m, 0.96 m, and 0.68 m, respectively. Regardless of the mismatch of surface terrain at some points, the results indicated little freeboard and thickness change within the first year of D28′s drifting. 2. We set the uncertainties of densities of the sea water, snow, and ice to 2 kg/m 3 , 5.3 kg/m 3 (1.5% of 355 kg/m 3 ) [51], and 10 kg/m 3 , respectively. The freeboard uncertainty was set to 1.59 m, which is the maximum standard deviation in cross validations of different altimeters (Table 1), leaving out the uncertainty in estimating the mean sea surface level. Then, the uncertainty at each grounding point was estimated through Monte Carlo simulations ( Table 3). The uncertainty of retrieving the iceberg draft was about 20-25 m on 2. We set the uncertainties of densities of the sea water, snow, and ice to 2 kg/m 3 , 5.3 kg/m 3 (1.5% of 355 kg/m 3 ) [51], and 10 kg/m 3 , respectively. The freeboard uncertainty was set to 1.59 m, which is the maximum standard deviation in cross validations of different altimeters (Table 1), leaving out the uncertainty in estimating the mean sea surface level. Then, the uncertainty at each grounding point was estimated through Monte Carlo simulations ( Table 3). The uncertainty of retrieving the iceberg draft was about 20-25 m on average, suggesting that a field investigation into densities is necessary in order to precisely evaluate iceberg geometric properties.
3. The iceberg draft estimation was carried out with a hypothesis that the iceberg had a cylindrical shape under the sea surface, and areas of the surface and the bottom were exactly the same. However, according to bathymetric surveys, the underwater parts of icebergs tend to be irregularly shaped [54]. Here, we considered the ideal case that iceberg D28 maintained a conical shape underwater (Figure 10a). To ensure the center of mass is vertically consistent with the surface center and to simplify the calculation, we used the average freeboard and average ice density to estimate its mass. Thus, according to the buoyancy principle and the relative positions of grounding points to the iceberg center (Figure 10b), the iceberg draft can be calculated with Equation (3): where H f is the average freeboard of iceberg D28, which was 40.63 m according to the freeboard map derived from altimetry data; S is the surface area; S h is the cross-sectional area at the draft h; H d max is the maximum iceberg draft; ρ i is the average ice density (891 kg/m 3 ); D gp is the horizontal distance from the center to the grounding point; and D is the distance from the edge to the center, passing the projection of grounding points on the surface. The maximum iceberg draft (H d max ) was 746.91 m as a consequence. The iceberg draft at each grounding point is shown in Table 4, and the uncertainties caused by the density model and freeboard estimation were also estimated using Monte Carlo simulations.

= − •
where is the average freeboard of iceberg D28, which was 40.63 m according to the freeboard map derived from altimetry data; is the surface area; is the cross-sectional area at the draft ℎ; is the maximum iceberg draft; is the average ice density (891 kg/m 3 ); is the horizontal distance from the center to the grounding point; and is the distance from the edge to the center, passing the projection of grounding points on the surface. The maximum iceberg draft ( ) was 746.91 m as a consequence. The iceberg draft at each grounding point is shown in Table 4, and the uncertainties caused by the density model and freeboard estimation were also estimated using Monte Carlo simulations.   Draft at grounding points ranged from 13.40 ± 1.99 m to 326.90 ± 31.58 m through the assumption of iceberg D28 maintaining conical shape under the water line. The results display a tremendous disagreement with the draft of a cylinder-shaped iceberg. Few studies on icebergs consider cone-shaped icebergs. Enderlin et al. explored the melting rate of icebergs in Sermilik Fjord with high-resolution digital elevation models by estimating the submerged area with ideal cylindrical shape and conical shape [59]. The melting rates of both situations were in line with each other, while the drafts differed significantly. Evaluating the iceberg properties requires more specific underwater bathymetric surveys to obtain a knowledge of its immersion topography.

Seabed Topography Revealed by the Maximum Water Depths at Grounding Points
Although a lack of the underwater geometric parameters of iceberg D28 may lead to a misestimation of its draft, the results in "cylindric-shaped" and "cone-shaped" situations remained consistent with findings about the seabed topography. Water depths estimated by iceberg draft indicated a more complicated topography on the continental land shelf off Kemp Coast than the documented steep slope extending smoothly to the deep ocean ( Figure 1b). Most of the results demonstrated a much shallower water area than the existing elevation models, implying the existence of ridges or hills on the continental shelf. Moreover, seabed elevation models from BedMap2 and BedMachine data are not in agreement with each other in this area, presenting elevation gaps from tens to hundreds of meters ( Figure 5). The lack of bathymetry data may cause inconsistencies. As a previous study concluded, the area off Kemp Coast is an area where icebergs tend to become grounded [33]. Further studies on grounded icebergs will provide more details about the seabed topography in this area.

Conclusions
The main purpose of this study was to estimate the seabed elevation by characterizing the grounded iceberg off Kemp Coast with multiple remote sensing data and to explore the driving force of iceberg motion through reanalysis products during the grounding event. The results indicate that the existence of banks or ridges on the continental shelf near Kemp Coast is not presented by seabed elevation models of BedMap2 and BedMachine data. The terrain features presented by the grounded iceberg extend for at least 20 km in the study area, which can affect the characteristics of the slope current. The force by sea water is a more dominant factor than that of wind in determining the motion of iceberg D28 during the grounding event. The estimation of iceberg draft with different assumptions was also discussed. This research demonstrated the possibility of investigating the seabed topography in the Southern Ocean through grounded icebergs as a supplement to traditional submarine bathymetry. Further studies on grounded icebergs will provide various perspectives on the seabed topography.