Progressive unanchoring of Antarctic ice shelves since 1973

The Antarctic Ice Sheet’s contribution to global sea-level rise has been accelerating, largely due to warm ocean currents weakening the buttressing provided by ice shelves, especially in West Antarctica and along East Antarctica’s Wilkes Land. Existing satellite altimetry records of ice-shelf thickness change span about three decades and indicate substantial thinning in parts of West Antarctica, the western Antarctic Peninsula, and Wilkes Land, with limited change elsewhere. However, this record is short relative to multidecadal ice-shelf response times, leaving uncertainty over how widespread thinning was before 1992. Extending observational records over longer periods is crucial for improving understanding of Antarctic change and constraining models of future sea-level contribution. This study aims to extend the record of ice-shelf thickness change back to 1973 by using changes in the surface expression of pinning points as a proxy for ice-shelf thickness evolution, thereby providing a pan-Antarctic, observation-based characterization of changes over the past five decades.

Previous work attributes much of Antarctic Ice Sheet mass loss to basal melting of ice shelves driven by incursions of warm modified Circumpolar Deep Water (MCDW), particularly in West Antarctica and the Antarctic Peninsula. Satellite altimetry since 1992 has documented major thinning in the Amundsen and Bellingshausen Sea sectors and parts of East Antarctica’s Wilkes Land, with relatively stable conditions elsewhere. Studies highlight decadal ocean variability and potential anthropogenic influences (e.g., wind-driven changes at the shelf break) as drivers of warm water access to ice-shelf cavities. Numerical modeling emphasizes the critical buttressing role of pinning points and ice rises in modulating upstream ice discharge, but observational constraints on multi-decadal pinning-point evolution prior to the altimetry era have been lacking. Geological and historical evidence indicates that retreat at Pine Island and Thwaites glaciers began prior to the satellite era, suggesting earlier onset of dynamic change than captured by modern altimetry alone.

The study assembled a five-decade observational record of pinning-point evolution using the Landsat archive. Two new near-cloud-free mosaics were constructed for 1973 (60 m resolution) and 1989 (30 m), which, together with the Landsat-7 LIMA mosaic (2000) and Landsat-8/9 imagery (2022), enabled tracking of changes in the surface expression of pinning points across three epochs: 1973–1989, 1989–2000, and 2000–2022. Pinning points—local bathymetric highs where ice shelves locally ground—produce distinct bumps or ice rises/rumples in optical imagery. Changes in their visible extent are interpreted as a proxy for ice-shelf thickness change: increases in pinning-point area indicate thickening (greater contact with the bedrock high), whereas decreases indicate thinning (reduced contact). Each mapped pinning point was categorized as smaller in extent, no detectable change, or larger in extent over each epoch. Validation was performed by comparing 2000–2022 pinning-point change with ICESat/ICESat-2-derived ice-shelf thickness change (2003–2019). The comparison showed high spatial agreement: 86% of pinning points that grew aligned with regions of thickening (>0 m yr⁻¹), 85% of no-change points matched regions of limited thickness change (−1 to 1 m yr⁻¹), and 66% of shrinking pinning points coincided with regions of thinning (<0 m yr⁻¹). Lower correlation for shrinking cases is attributed to factors such as ungrounding-induced localized downstream thickening and conservative classification (ambiguous cases labeled no detectable change). Conflicting localized signals between altimetry products were noted in a few cases. Regional summaries were compiled using pie charts of change proportions over delineated sectors, overlaid on the REMA mosaic. In areas lacking cloud-free 1973 imagery, changes were reported for 1989–2022 instead.

Pan-Antarctic trends: Pinning-point loss has accelerated over five decades. The share of pinning points that reduced in area rose from 15% (1973–1989) to 25% (1989–2000) and 37% (2000–2022). Validation against ICESat/ICESat-2 (2003–2019) shows 86% agreement for growth with thickening, 85% for no change with limited thickness change, and 66% for reduction with thinning. Early hotspots and spread: Even in 1973–1989, thinning hotspots existed in the Amundsen Sea Embayment (West Antarctica) and East Antarctica’s Wilkes Land (Holmes, Moscow University, and Totten ice shelves), indicating thinning began at least 50 years ago in these sectors. Antarctic Peninsula: Following collapses of Prince Gustav, Larsen A, Larsen B, and Wordie ice shelves, all associated pinning points were lost. Larsen C and D showed little change overall, except shrinkage at Bawden Ice Rise since 1989, implying localized basal melt increases. Bellingshausen sector: The thinner Wilkins and Abbot ice shelves thickened from 1973–1989 and 1989–2000; from 2000–2022 they showed mixed/neutral patterns as the thermocline reached mean draft depths. Thicker George VI, Stange, and Venable experienced increasingly widespread thinning, with nearly all pinning points shrinking since 2000, consistent with a shallowing/strengthening warm layer on the continental shelf. Amundsen Sea sector: 1973–1989 saw 35% of pinning points shrink and 15% grow; Pine Island, Thwaites, Dotson, and Crosson were already unanchoring and thinning before altimetry records. Pinning-point loss intensified with 83% shrinking in 1989–2000 and 94% in 2000–2022, aligning with pervasive altimetry-diagnosed thinning. Marie Byrd Land and Ross sector: Hull Glacier showed notable pinning loss 1973–1989; Sulzberger largely stable, though some loss near the grounding line suggests warm water access. Swinburne Ice Shelf exhibited substantial thickening (estimated >30 m in parts). On Ross Ice Shelf, most pinning points were stable, but Steershead Ice Rise and two others downstream of Kamb Ice Stream consistently shrank, supporting model-predicted thinning following Kamb’s shutdown. Engelhardt Ice Ridge retreated ~5 km from 2000–2022. East Antarctica (Wilkes Land and beyond): Holmes showed steady pinning reduction across all epochs; at Moscow University Ice Shelf an elongated ice rise eroded by ~6 km between 1973 and 2000, enabling a new tributary to form; Totten showed subtle losses pre-2000, with more widespread loss after 2000. In Victoria/George V Lands, a major pinning point was lost on Campbell Glacier tongue and a pinning point shrank on Rennick Ice Shelf in the 2000s. Enderby Land saw growth of pinning points in 2000–2022 (Wilma–Robert–Downer Embayment; Lützow–Holm Bay), consistent with thickening. Filchner–Ronne region: Most pinning points remained stable, but prominent ice rises underwent rapid changes. Rapidly changing ice rises: Borchgrevink Ice Rise ungrounded in the late 1970s, drifted downstream, then regrounded near the front forming a new rumple; Hemmen Ice Rise (Ronne front) gradually shrank over three decades before breaking apart in the mid-2000s, implying calving-regime changes; Korff Ice Rise grounded area expanded by ~20 km on its northern flank, with surface elevations ~25 m above surrounding shelf, suggesting possible initiation of broader ice-rise expansion.

By using pinning-point surface changes as a proxy, the study extends observational evidence of Antarctic ice-shelf thickness changes back to 1973, revealing that significant thinning commenced earlier than captured by satellite altimetry. The progressive and accelerating unanchoring—especially across the western Antarctic Peninsula and Amundsen Sea—indicates declining buttressing, which is known to increase upstream ice discharge to the ocean. Regional contrasts are consistent with oceanographic controls: decadal variability and a shoaling/thickening of warm modified Circumpolar Deep Water on continental shelves since 2000 enhanced basal melt beneath both thick and thin shelves in different ways. Potential drivers include anthropogenic trends in winds over the shelf edge (in the Amundsen Sea) and a poleward shift of MCDW linked to westerly wind shifts in East Antarctica. Freshwater feedbacks from ice-shelf melt may further modulate local delivery of warm water to cavities. The hysteretic nature of pinning means that once ungrounded, re-establishing prior buttressing requires greater thickening than the prior loss, implying some changes may be effectively irreversible on multidecadal timescales. Rapid structural changes in key ice rises demonstrate decadal-scale reorganization of calving style and buttressing not yet represented in most ice-sheet models. Collectively, the findings underscore heightened vulnerability and a likely strengthening contribution to sea-level rise if current trends persist.

This work provides the first pan-Antarctic, observation-based reconstruction of ice-shelf thickness change since 1973 by tracking the evolution of pinning points. It shows that thinning began at least five decades ago in the Amundsen Sea and Wilkes Land and has since spread and accelerated, with the fraction of shrinking pinning points increasing in each successive epoch. Many shelves in West Antarctica and parts of East Antarctica are progressively unanchoring, diminishing buttressing and priming further acceleration of ice discharge. Particularly concerning are large shelves that remain substantially pinned yet exhibit rapid recent pinning-point loss (e.g., George VI, Getz, Holmes, Moscow University, Totten). Future research should focus on quantifying the mechanisms driving the intensified access of warm water to ice-shelf cavities (including wind-driven and anthropogenic influences), integrating rapid ice-rise dynamics and hysteresis into numerical models, refining proxy-to-thickness calibrations, and maintaining sustained, high-resolution observations to monitor evolving buttressing and calving regimes.

The approach infers thickness change direction from a proxy (pinning-point area changes), not direct thickness measurements, and thus cannot uniformly quantify magnitudes. Correlation with altimetry is strong for growth and no-change classes but lower (66%) for shrinking cases, potentially due to ungrounding-induced localized downstream thickening and conservative classification of ambiguous cases as no detectable change. In some localized regions, different altimetry products provide conflicting signals. Spatial coverage is limited where cloud-free 1973 imagery was unavailable; for these areas, only 1989–2022 changes could be assessed. Local oceanographic and geometric complexities (e.g., cavity circulation, bed topography) may decouple pinning-point evolution from broader shelf-mean thickness trends in limited cases. The main text provides limited methodological detail on classification thresholds and uncertainty quantification; fuller details likely reside in Methods and Supplementary materials.