2.1. Satellite imagery

We used freely available satellite visible imagery from the European Space Agency’s (ESA) Copernicus Sentinel-2 Earth Observation mission to observe and quantify ice-front variability and associated surface deformation on PIG’s south shelf from 2017 to 2023, a period marked by substantial morphological change. We used the Sentinel-2A/B red (Band 4; ∼649.6–679.6 nm), green (Band 3; ∼542.5–577.3 nm), blue (Band 2; ∼460.2–525.2 nm) and visible and near-infrared (VNIR; Band 8; ∼780.3–885.3 nm) bands, each of which has a spatial resolution of 10 m. The constellation of identical satellites (Sentinel-2A and Sentinel-2B) has a revisit frequency of 5 days (reduced to 2–3 days at the poles), allowing for detailed geospatial analysis of surface feature evolution.

2.2. Ice velocity

We used surface velocities from The Inter-Mission Time Series of Land Ice Velocity and Elevation (ITS_LIVE) project (Gardner and others, Reference Gardner, Fahnestock and Scambos2019)—part of NASA’s Making Earth System Data Records for Use in Research Environments (MEaSUREs) program—to observe velocity changes at locations of apparent surface deformation on PIG’s south shelf from 2014 to 2023.

2.3. Surface deformation

2.3.1. Manual surface deformation tracing

We monitored the temporal changes (2017–23) in surface deformation on PIG’s south shelf by manually tracing the surface expressions of ice deformation, consisting of well-defined ripples in the ice and fractures that ruptured the surface (e.g. Fig. 2a), similar to the method employed by Alley and others (Reference Alley2024) for digitizing sub-ice-shelf channels in West Antarctica. We limited our analysis to a region of substantial surface deformation (spatial extent indicated on Fig. 1b), excluding the complex south margin and the ice-ocean interface, to provide a consistent evaluation of deformation that is not biased by calved ice loss or intricate margin damage. We used a consistent bounding box (85 km² in area, with a mean velocity of 735 m a⁻¹) for all years (i.e. an Eulerian reference frame), which does not account for seaward ice advection on the positions of recognizable surface features. This approach simplifies tracking and, although it does not follow individual features, is reasonable given the relatively slow velocities within the selected region.

Figure 2.

Increase in surface deformation from (a) 7 March 2018 to (b) 13 December 2013 on PIG’s south shelf, observed from Sentinel-2 visible imagery. The loss of a substantial area of ice at the PIGIS front (visualized in (c) by overlaying (a) and (b)) coincides with a change of $34\pm8$ m a⁻² in ice speed (see Figure S5 for uncertainty quantification details) in the region of highly deformed ice on PIG’s south shelf, following the 2018 calving event. The dotted bounding box in (c) delineates the bounds of Figure 3. (d) Time series of ice flow speed at discrete points of the south shelf. The cyan, teal and purple colors represent surface velocity time series at the corresponding locations marked in (c). These sites show relatively stable surface velocities early in the record, with seaward acceleration around 2018.

For seven cloud-free scenes (one per year), we digitized deformation on scenes taken roughly at the same time of year (here between October and December) to minimize spectral variability from external controls (i.e. solar zenith angle) that impact surface reflectance and may bias the appearance of surface features (e.g. shadows) without necessarily reflecting physical changes in the surface itself. We then quantified the digitized surface deformation by calculating the total length of traced deformation (summing the length in kilometers of all traced segments) and the mean length of traced segments per scene, providing quantifiable measures of how the surface is evolving with time. Although manual tracing generally involves some level of subjectivity—where different investigators may interpret features differently (e.g. as noted by Alley and others, Reference Alley2024)—our focus on relative changes and overall trends minimizes the influence of these variations on the study’s broader conclusions. Nonetheless, we provided another measure of surface deformation that limits subjective bias to complement the manual tracing method described here.

2.4. Satellite-derived strain rates

To assess the time-evolving surface deformation near the south margin, we used effective strain rate fields derived from satellite-based velocity data presented in Joughin and others (Reference Joughin, Shapero, Smith, Dutrieux and Barham2021). Their study applied speckle-tracking techniques (after Joughin, Reference Joughin2002) to Sentinel-1A/B data, creating gridded (250 m resolution) velocity measurements for the PIG basin from January 2015 to September 2020, at regular 6-day intervals (Rosen and others, Reference Rosen, Gurrola, Sacco and Zebker2012; Lei and others, Reference Lei, Gardner and Agram2021). Time-dependent surface velocity fields are then used to calculate time series of horizontal strain rates $\dot{\epsilon}_{ij}$ for i,j = x,y (the two horizontal components), defined by:

(1)\begin{equation} \dot{\epsilon}_{ij} = \frac{1}{2} \left( \frac{\partial u_i}{\partial x_j} + \frac{\partial u_j}{\partial x_i} \right) \end{equation}

where $u_i$ and $u_j$ are the horizontal velocity components and $x_i$ and $x_j$ represent the horizontal coordinates of the velocity data in south polar stereographic projection. To calculate the components of the velocity gradient, we applied a two-dimensional Savitzky-Golay filter with a 3 km window size (Savitzky and Golay, Reference Savitzky and Golay1964). Taking the square root of the second invariant of the horizontal strain rate tensor ( $\dot{\epsilon}_{ij}$) then gives the effective horizontal strain rate, $\dot{\epsilon}_{e}$:

(2)\begin{equation} \dot{\epsilon}_{e} = \sqrt{\dot{\varepsilon}_{xx}^{2} + \dot{\varepsilon}_{yy}^{2} + \dot{\varepsilon}_{xx}\dot{\varepsilon}_{yy} + \dot{\varepsilon}_{xy}^{2}} \end{equation}

Equation (2) defines a scalar quantity that accounts for both the normal ( $\dot{\varepsilon}_{xx}$ and $\dot{\varepsilon}_{yy}$) and shear ( $\dot{\varepsilon}_{xy}$) components of strain on a Cartesian (x,y) grid, resulting in a value that showcases the intensity of surface deformation. Therefore, the effective strain rate, which we use here, provides an integrative measure of the rate at which deformation occurs, encompassing all forms of deformation such as extension, compression and shear.

We examined the evolution of surface effective strain rate at specific locations near the glacier’s southern margin to identify where deformation rates changed most significantly. To pinpoint the most notable changes in strain-rate time series, we first identified hinge points (Fig. S1)—moments of steep curvature change—by calculating the second time derivative of the strain rate and highlighting points where this deviated >1.5 standard deviations from the mean. A larger threshold (e.g. 2 standard deviations) tightens the criterion, whereas a smaller one relaxes it, capturing more hinge points but still focusing on significant changes. Next, we applied a KMeans clustering algorithm to group these hinge points based on their timing. The central cluster identified the principal moment of maximum curvature change, highlighting when the time-evolving strain rates collectively changed across multiple locations. We repeated this process in four regions of interest (color-coded in Fig. 4) and with four clusters to match the number of calving events during this time period (Fig. S2). This approach systematically identifies key changes in the structural evolution of the south margin of PIG by identifying time-varying deformation change.

3.1. A structurally changing south shelf

We observed substantial morphological changes on PIG’s south shelf between 2017 and 2023. Surface deformation near a large ice-shelf basal channel became increasingly pronounced (Fig. 2), and both of our deformation metrics confirmed a sustained increase in ice deformation and roughness over time. In a section of extensively deformed ice (extent shown in Fig. 1b), the total length of visible deformation features increased from 33.4 to $107.4~\,\mathrm{km}$ between 2017 and 2023 (Fig. 3a), indicating a mean increase in linearly deformed ice of ∼12 km a⁻¹. Image pixel standard deviation values (in image digital numbers) increased by 12.0, 14.1, 16.3 and 18.6 from initial values of ∼2 for the blue, green, red and VNIR bands, respectively, during this time (Fig. 3b). This trend signifies a shift toward greater surface roughness, with the ice surface evolving from relatively spectrally homogeneous to increasingly heterogeneous. The emergence of undulations and crevasses is reflected in spectral variations across each Sentinel band, driven by surface factors such as snow cover fluctuations, melt patterns, and notably, crack formation that alters optical properties. The escalating deformation over time, confirmed by both deformation and roughness metrics, signals substantial dynamic change on the south shelf.

Figure 3.

Surface deformation evolution on a section of the south shelf from 2017 to 2023 shown spatially (top) and temporally via the (a) manual tracing method and (b) systematic spectral analysis. All imagery is from Sentinel-2. Expanded inset (right) is contrast stretched and gamma corrected to reveal finer detail. Inset in (a) shows an example of manually traced deformation.

This transition toward more extensive surface deformation coincides with a 29 October 2018 calving event, during which >60 km² of ice was abruptly lost from the south shelf (Fig. 2c). This local loss was part of a larger calving event totaling >375 km². The event triggered an ice-flow acceleration of $34 \pm 8~\mathrm{m~a}^{-2}$ immediately inland of the >60 km² removal (Fig. 2). The lost ice had provided important resistance to seaward flow, and its removal left the south shelf less constrained, though it remains laterally supported by grounded ice and the PIGIS trunk.

Between 2017 and 2023, PIGIS experienced two other major calving events (Fig. S3): one on 23 September 2017 (>295 km²) and another on 11 February 2020 (>315 km²), resulting in frontal retreats of ∼7 and ∼19 km, respectively (Sun & Gudmundsson, Reference Sun and Gudmundsson2023). Together with the 2018 event, these three calving events caused a total area loss of ∼20% for PIGIS between 2017 and 2023 (Joughin and others, Reference Joughin, Shapero, Smith, Dutrieux and Barham2021). However, the 2018 event had the most distinct impact on the south shelf: it caused the largest ice loss near the southern margin (Fig. 2c), coincided with increased surface deformation and led to a marked velocity increase in this region (Fig. S4). The shelf’s response to the 2018 event, compared to those of 2017 and 2020, underscores the importance of localized buttressing points in controlling ice-shelf behavior (Dupont and Alley, Reference Dupont and Alley2006; Miele and others, Reference Miele, Bartholomaus and Enderlin2023).

3.2. Changing dynamics: weakening margins

Surface strain-rate fields provide clear indication of a glacier’s surface stress state and are indicative of mechanical weakening in glaciers or ice shelves prior to rift propagation (e.g. Jansen and others, Reference Jansen, Luckman, Kulessa, Holland and King2013). Our satellite-derived strain-rate fields from 2015 to 2020 reveal a progressive increase in strain rates near PIGIS’s south margin (Fig. 4). This increase aligns with previously observed glacier acceleration (e.g. Mouginot and others, Reference Mouginot, Rignot and Scheuchl2014; De Rydt and others, Reference De Rydt, Reese, Paolo and Gudmundsson2021), intensifying velocity gradients across PIG’s boundaries and dynamic stress within the ice, which results in measurable deformation changes (i.e. strain-rate changes). Notably, since 2018, these elevated strain rates have extended beyond the south margin—where velocity gradients (and associated shear stresses) are most pronounced—and have started to propagate southwestward toward the south shelf (Fig. 4b). This pattern of increasing strain rates, particularly southwest of the southern margin and notably pronounced post-2018 (Fig. 4), aligns with recent observations of margin damage (e.g. Lhermitte and others, Reference Lhermitte2020). Our results provide quantifiable evidence of continued ice fragmentation (e.g. Fig. 4a, b) and a widening, deteriorating and migrating margin.

Figure 4.

Effective strain-rate evolution on the south shelf from 2015 to 2020 with Sentinel-2 imagery (a) before and (b) after the 2018 calving event, overlaid with the respective effective strain rates. (c) Strain-rate time series at specific locations (inset) of concentrated elevated strain rates. Calving events are marked as vertical orange lines. Dashed and dotted black lines denote when images (a) and (b) were taken, respectively. Dashed purple lines indicate the principal curvature hinge points.

The 2018 calving event altered the forces exerted on this section of the southern shelf. Not only did the ice in this region deform and accelerate post-calving (Fig. 2c), but we also observed a distinctive shift in the strain-rate evolution near the south margin shortly before the loss of this ice (Fig. 4). The principal curvature change points (purple dashed lines in Fig. 4), independently identified across the four regions of interest (represented by four colors in Fig. 4c), occurred in the season just before the 29 October 2018 calving event (between 42 and 104 days before this event; Fig. 4c). Rather than implying direct causality, we interpret these points as quantitative markers of trend shifts in strain rate evolution that coincide with key structural transitions. This consistency in time suggests a systemic change in stress state across the south shelf that ultimately triggered this large calving event, which strongly impacted the ice-flow dynamics of the ice shelf. Although it is possible that the identified principal curvature change points are a delayed response to the 2017 calving event, which happened almost a year before the change points, the proximity in time and the larger consequences for ice-shelf dynamics suggest that this change is more likely linked to the 2018 event. Moreover, the strain-rate gradient at sites southwest of the southern margin (dark blue, light blue and light pink in Fig. 4c) appear to increase over time, whereas the strain rate gradient at the south margin (dark pink in Fig. 4c) tends to decrease toward zero during this period. These observations suggest a potential ‘jump’ in the shear margin (Fig. 4b), where the zone of active deformation is shifting outward from its previous boundary, indicative of important structural changes across this region.

Four other major calving events occurred between 2015 and 2020 (vertical orange lines in Fig. 4). Although we show the single maximum point of strain-rate change correlated to the 2018 calving event (Fig. 4), a multiple hinge point analysis demonstrated that the 2017 and 2020 calving events also correlate with smaller, but identifiable changes in effective strain (Fig. S2). These hinge points may signify structural transitions where the ice shelf has undergone a change in its mechanical behavior. Although the presence of fractures, crevasses and rifts itself implies that a critical stress was attained to result in irreversible failure, it is the spatiotemporal pattern of observed strain-rate evolution that is particularly crucial here: elevated strain rates extending toward PIG’s south shelf suggest that the ice-shelf’s southern shear margin is structurally compromised, increasing its susceptibility to further fracturing and weakening. Taken together, these evolving strain rate patterns likely reflect the cumulative effects of multiple perturbations over time, with the 2018 calving event generating a notable step change in the evolving stress state of the south shelf.

3.3. Additional processes potentially influencing PIGIS stability

In addition to the observed changes across the trunk and southern shelf, two dynamic features may significantly impact the future evolution of the PIGIS: (1) the existence and development of a prominent basal channel across the south shelf and (2) extensive lateral detachment along the northern shear margin.