BREAK OUT OF A-76 ICEBERG AND RECENT DYNAMIC CHANGES OF ITS ENCOLSURE RIFTS IN RONNE ICE SHELF, ANTARCTICA

Monitoring the stability of the Ronne Ice Shelf, particularly the calving event, is an integral and important part of the study of Antarctic ice sheet mass balance and sea-level rise. This paper presents the results of the analysis of the world’s largest iceberg (A-76 iceberg), which was formed during a calving event in the Ronne Ice Shelf (RIS) on May 13, 2020, and subsequently broke into three fragments (A-76A, A-76B, and A-76C icebergs). The iceberg development cycle, up to this point, including its formation, separation and drift, was observed and analyzed. The detailed development process of rifts associated with the detachment of the A-76 iceberg in front of RIS before calving was analyzed using remote sensing data from multiple sources (ERS, RADARSAT-1, ALOS PALSAR, Landsat7, and Landsat-8). In addition, based on a total of 66 Sentinel-1 A/B Synthetic Aperture Radar (SAR) images acquired between May 13, 2021 and March 11, 2022, a multi-scale segmentation approach was applied to continuously track the drift path of A-76 icebergs, A-76A, A-76B, and A-76C. We calculated the average drift velocity of these icebergs and found that A-76C iceberg drifted the fastest, followed by A-76A, and A-76B, from May 30, 2021, to March 11, 2022. Future tracking of other iceberg parameters, such as area, thickness, freeboard, and volume, could help assess the melting rates of the icebergs.


INTRODUCTION
The Antarctic Ice Sheet (AIS) is one of the most vulnerable and critical regions under the influence of global climate change, and its mass balance will significantly affect global sea level rise. Complete melting of the AIS would result in global sea level rise of ~57.9m (Morlighem et al., 2020). Two pathways by which lost Antarctic mass enters the Southern Ocean are calving at ice shelves and basal melting, each accounting for about half of the total mass loss (Paolo et al., 2015;Rignot et al., 2013). Monitoring accelerated ice shelf thinning and calving processes is essential to quantify ice shelf instability and assess the contribution of AIS to global sea level rise. In addition, an increasing number of calving events or accelerated ice shelf thinning could destabilize ice shelves and lead to a reduction in ice shelf support for the entire ice sheet (Gudmundsson, 2013).
In recent decades, many studies have been conducted on the calving processes of ice shelves at AIS, including the Amery Ice Shelf (Fricker et al., 2005), the Larsen C Ice Shelf (Larour et al., 2021), the Brunt Ice Shelf (Cheng et al., 2021), and others. However, research on calving in the Ronne Ice Shelf (RIS) is not as extensive. The Filchner-Ronne Ice Shelf (FRIS) in the southern Weddell Sea accounts for 30% (~430 000 km 2 ) of the total area of AIS, and the amount of ice stored in its upstream drainage basins has the potential to raise sea level by ~14 m (Morlighem et al., 2020). Using remote sensing datasets from multiple sources (Landsat-7 ETM+、Landsat-8 OLI、Envisat ASAR、ZY-3、WV-2、ICESat), Li et al. (2017) indicated that there are two rapidly developing transverse rifts T1 and T2 on the Filchner Ice Shelf front, which could cause a major calving event * Corresponding author similar to the1980s. On the other hand, the rifts in the RIS are particularly active in the shelf front region, especially in the two marginal areas and in the frontal calving area. The transverse rifts on the front of the RIS originate from the upstream grounding zone and propagate during their advection process (Hulbe et al., 2010;Walker et al., 2013Walker et al., , 2019. These transverse rifts, together with margin rifts, eventually form plains that are calved into tabular icebergs (Hulbe et al., 2010;Walker et al., 2015;Li et al., 2017). Therefore, studying the evolution of these icebergforming rifts can help understand the relationship between shelf stability and calving events in RIS.
Icebergs have a life expectancy of several years or more. Their size ranges from a few square kilometers to several thousand square kilometers. Because icebergs drift with ocean currents, their tracks are a good indicator of ocean circulation (Collares et al., 2018). In addition, drifting icebergs can endanger shipping and affect the marine environment around them (Lasserre, 2015). Therefore, it is important to detect and track icebergs. The Synthetic Aperture Radar (SAR) data can monitor icebergs during polar night or cloudy conditions (Mazur et al., 2017;Wesche and Dierking, 2015) and are valuable datasets for detecting iceberg movements and area changes. Earlier SAR satellite data sets, such as ERS (Young et al., 1998), Envisat (Li et al., 2018;Mazur et al., 2017), and Radarsat-1 (Wesche & Dierking, 2015) have been used to study ice shelf activity. Nowadays, Sentinel-1 A/B SAR GRD data are widely used for iceberg detection due to their high temporal and spatial resolution (Torres et al., 2012;Barbat et al., 2021;Braakmann-Folgmann et al., 2021,2022Koo et al., 2021).
In this paper, we examine rift evolution and iceberg displacement processes associated with a major calving event in the RIS using remote sensing data from multiple sources. First, we combined multi-satellite data from 2000 to 2021 to examine the dynamic evolution of rifts at the RIS front prior to the A-76 iceberg calving event (ESA, 2021). The multi-sensor data used include optical satellite and SAR images (e.g., Landsat, Radarsat-1, ALOS, ERS, and Sentinel-1) (Table 1). Second, the Sentinel-1 A/B GRD image dataset was used to analyze the motion trajectory of the A-76 iceberg (including A-76A 、 A-76B and A-76C). In the following paper, we first describe the data, the study area and the subjects, then the methods for rift measurement and iceberg extraction. Finally, we present our results and draw our conclusion.

DATA AND STUDY SITE
Satellite data were combined to measure the rift evolution and track monitoring related to the A-76 iceberg. We obtained the RES-1, RADARSAT-1, ALOS PALSAR, and Sentinel-1 GRD satellite imagery from the Alaska Satellite Facility (https://search.asf.alaska.edu/) covering the RIS front between May 9, 2000, and March 11, 2022, and Landsat-7/8 imagery between February 9, 2012, and October 11, 2014 by using USGS EarthExplorer (https://earthexplorer.usgs.gov/). The details of the data are shown in Table 1. The study area for this paper is located in the shelf front of RIS.

METHODS
This paper examines the life cycles of the A-76 iceberg, before and after its separation from RIS to investigate the relationship between the rift evolution and iceberg calving in RIS. We observed a 21-year process from the first rift appearance, to tabular plane formation, and to final iceberg disintegration, based on images acquired from May 5, 2000 to May 13, 2021.Specifically, we analyze changes of the rifts in the shelf front of the RIS after the last major calving event in 2000 (Lazzara et al., 1999) to investigate the propagation mechanism of rifts. Then, we examined the formation of the A-76 iceberg. Finally, the drift trajectory and dynamic changes of the A-76 iceberg are analyzed ( Figure 1). Figure 1. RIS rift measurements before calving and iceberg tracking after calving.

Rift Measurements and Tracking of the A-76 Iceberg
Both transverse and marginal rift propagation is involved in calving, ice front retreat, and. ice shelf instability. To understand the changes in RIS rifts and its separation mechanism, we measured the lengths of transverse and marginal rifts using multisensor data to investigate the relationship between the rifts' changes and ice shelf calving.
To track the A-76 (A-76A, A-76B, and A-76C) iceberg, we use Sentinel-1 GRD images for edge extraction and iceberg centroid detection. First, the Sentinel images are accurately aligned, radiometrically processed, and speckle filtered; and they are then used to measure the multi-temporal parameters. Second, we segment the processed images using a multiscale segmentation method to obtain the iceberg image objects. Next, we merge these extracted segments to obtain the iceberg boundaries of the A-76 (A-76A, A-76B, and A-76C) iceberg. Then we derive the centroid coordinates from the boundaries. Finally, we use the determined centroids of the A-76 (A-76A, A-76B, and A-76C) iceberg to trace the drift trajectories. The above procedure is shown in Figure 2.

SAR Image Pre-processing and Multi-scale Segmentation
We obtained the Sentinel-1 GRD data from Alaska Satellite Facility (ASF) and pre-processed the data using the workflow in Figure 2. We resampled all the data to a spatial resolution of 30 m. The purpose of image segmentation in this paper is to obtain homogeneous image objects. The multi-scale segmentation method developed by Baatz & Schäpe (2000), which has also been applied to land-fast sea ice extraction in Antarctica (Kim et al., 2020), is applied to extract boundaries of the A-76 iceberg (including A-76A, A-76B, and A-76C) from the multi-temporal SAR images. To achieve the best segmentation results, the multiscale segmentation algorithm is a bottom-up segmentation method that minimizes the heterogeneity and maximizes the homogeneity of image objects to obtain the best segmentation results (Amani et al., 2017). The segmentation process in the multi-resolution segmentation algorithm, the segmentation process is performed according to user-defined criteria with three parameters, including the scale parameter, the weight of the shape parameter, and the weight of the compactness parameter, which are all critical to the segmentation results (Baatz & Schäpe, 2000;Belgiu & Drǎguţ, 2014).
The following equations (1)-(5) provide a more detailed explanation of the principles and process of segmentation. We use equation (1) to calculate the heterogeneity of the image object region (Baatz & Schäpe, 2000).
where refers to the heterogeneity of the total area of image object. ℎ refers to the spectral heterogeneity ranging from 0 to 1, ℎ refers to the shape heterogeneity ranging from 0 to 1, and represents the weight of spectral heterogeneity.
In performing the segmentation, our goal is to obtain the best segmentation results by minimizing the heterogeneity f of image objects and maximizing the homogeneity of image objects where the spectral and shape heterogeneity are calculated using equations (2) and (3).
where ℎ refers to the smoothness, ℎ refers to the compactness, refers to the size of the image object, 1 and 2 represent the two smaller image objects for merging and represent the larger image object after connecting, and is the standard deviation of the spectral values within the objects. Furthermore, the shape heterogeneity, which includes the criterion of shape smoothness and the co-determination of shape density, is calculated using (4) and (5).
where refers to the perimeter of the image object, and refers to the outer tangential rectangle of image object. Throughout the segmentation process, the whole image is segmented. The scale parameters affect the average size of the image objects, which are generated according to several adjustable homogeneities or heterogeneity criteria for ℎ and ℎ .
This study tested various combinations of weights for the scale, shape, and compactness parameter fields, and an optimized combination was determined based on a visual inspection of the resulting objects. We employed the control variable method for parameter selection to get better segmentation parameters. To begin the experiment, we set the compactness parameter to 0.5, the shape parameters to 0.1, 0.3, and 0,5, and the segmentation scales to 100, 300, 500, 800, and 1000, for a total of 15-parameter combinations. In the segmentation experiments, these three parameters interact with one another. To control the segmentation quality, over-segmentation is checked in Figure 3   We found that a scale threshold of 500 produces a clear and reasonable segmentation of the iceberg's boundary, and that using smaller shape values produced better segmentation results. The A-76 iceberg (containing A-76A, A-76B, and A-76C) can be extracted well when the segmentation scale is 500, the shape parameter is 0.1, and the compactness parameter is 0.5. Figure 4 shows the segmentation results of the Sentinel-1 GRD image collected on July 3, 2021.

Iceberg's Centroid Computation
The iceberg's boundary is determined by the result of the preceding segmentation stage. Using equations (6) and (7), we extract the object centroid from the retrieved A-76, A-76A, A-76B, and A-76C icebergs, and then derive the coordinates of the iceberg's centroid points ( ̅ , ). The following are the details of equations (6) and (7) ( , ) refers to the object containing the icebergs, refers to different icebergs, refers to the iceberg area, j and k refer to rows and columns, and M and N represent the area of the object icebergs in the above equations.  20, 1973, May 9, 2000, November 26, 2002, and December 19, 2021, are shown in (a), (b), (c) and (d) respectively.

Rifts Changes
In the shear margin of RIS has always been occupied by marginal rifts at difference evolution stages (Figure 5a and 5b). Marginal rifts may increase their lengths in the lateral directions to connect to transverse rifts and form calving blocks (Lazzara et al., 1999). At same time they also break to fragments and produce mélanges. The transverse rifts, marginal rifts, and longitudinal rifts each accounted for 46.5%, 39.9%, and 13.6%, respectively, of the detachment boundary of the A-76 iceberg to complete the calving process. The long transverse rift (270 km) is one of the rifts defined as dormant or intermittently active by Walker and Gardner (2019) and Hulbe et al. (2010).
The RIS marginal rifts formed rapidly as warm Southern Ocean interacted with the shelf front, forming dynamic regions in the shear margins filled with mélanges. The major marginal faults have evolved dramatically over the last decade and are now poised to join the transverse rifts. We used multi-sensors data to analyze the pre-calving rift development process and iceberg characteristics, including optical satellite and Synthetic Aperture Radar (SAR) images (e.g., Landsat, Radarsat-1, ALOS, ERS, and Sentinel-1) ( Table 1). The transverse rift grew slowly, expanding 16.1 km over last 20 years, with 17% of the growth occurring in 2000. It is worth noting that the growth of the marginal rift has occurred in two stages. It grew by 16.5 km between May 2000 and January 2018, at a rate of 0.9 km/y. The rifts suddenly expanded to 72.6 km in the second period, from January to December 2018, at a rate of 43.7 km/y ( Figure 6). Finally, in December 2018, the active margin rift reached the transverse rift.

A-76 Iceberg Tracking from Space
A-76 iceberg entered the Weddell Sea from RIS on May 13, 2021, as observed by the British Antarctic Survey and confirmed by the United States National Ice Centre using Copernicus (Figure 7, ESA, 2021); the A-76 iceberg has a total area of 4320 km 2 . Thirteen days later, the iceberg was further divided into three sections, A-76A, A-76B, and A-76C (USNIC, 2021), each with areas of 3442 km 2 , 479 km 2 , and 400 km 2 . The entire calving process of A-76 iceberg was captured by Sentinel-1 GRD images from May 11, two days before calving, to May 26 (Figure 7).    Table 2. As the icebergs drift through the ocean, their position and orientation change over time.
We hypothesize that the varying trajectories and velocities of A-76A, A-76B, and A-76C icebergs are caused by a variety of complex environmental variables (e.g., ocean currents, winds, seafloor topography, atmosphere, and sea ice) (Li et al., 2018;Koo et al., 2021).

CONCLUSIONS
The final evolution process of the iceberg producing rifts of A-76 in RIS is investigated in this study, specifically the variation in rift length before calving and the iceberg's drift trajectory after calving. First, we tracked the changes in the rifts near the RIS's shelf front. The findings show that transverse rifts at the RIS shelf front determine iceberg size and develop relatively slowly, whereas marginal rifts develop at a faster rate before calving. The interconnection of transverse and marginal rifts may provided a high probability of a short term iceberg disintegration. Second, we developed a semi-automatic method for calculating the trajectories of A-76, A-76A, A-76B, and A-76C icebergs from the Sentinel-1 A/B GRD image data. Our findings can also be used to model the break-up of other large tabular icebergs that follow similar paths, and their effects can be incorporate into ocean models. We will also optimize our methodology to monitor additional parameters such as iceberg area, thickness, freeboard, and volume to assess the rate of iceberg melting. Additionally, we may also study the effect of air temperature, ocean temperature, wind speed, and seafloor topography on the drift, area change, and ice volume change of A-76A, A-76B, and A-76C icebergs.

Disclosure Statement
No potential conflict of interest was reported by the authors.

Author contributions
RL led the study and created the research concept. AZ developed the method and processed the data, YC analyzed the data and LA interpreted results and edited the entire manuscript. Everyone is involved in background research for this project.