East Antarctic warming forced by ice loss during the Last Interglacial

The Last Interglacial (129–116 ka) featured strong high-latitude warming driven mainly by orbital forcing, while greenhouse gases were slightly lower than pre-industrial. Sea level high stands imply substantial cryospheric change, with the Antarctic Ice Sheet likely contributing 1–7 m sea level equivalent (SLE), potentially peaking near 127 ka. Proxy reconstructions indicate Southern Ocean summer SSTs 1.8 ± 0.8 °C higher and up to 3 °C annually at ~127 ka, yet coordinated PMIP4 LIG simulations underestimate Southern Hemisphere warming and Antarctic surface air temperature changes, in part because they prescribe modern Antarctic ice sheets without LIG ice loss. Past studies suggest subsurface Southern Ocean warming is necessary for AIS retreat, with possible WAIS collapse above ~2–3 °C subsurface warming, and mechanisms including AMOC weakening or meltwater-induced stratification. However, dynamic coupling of an Antarctic ice sheet with a fully coupled climate model for LIG remains technically challenging. This study investigates how partial AIS reduction and associated meltwater fluxes during the LIG affect Southern Ocean circulation, subsurface temperatures, and East Antarctic surface climate, testing feedbacks that influence AIS stability and explaining proxy–model discrepancies.

Proxy data indicate enhanced Northern Hemisphere summer warmth and Southern Ocean SST anomalies during the LIG, while Antarctic temperature proxies suggest 0.9–3.3 °C higher Antarctic SAT than PI at ~127 ka. PMIP4 lig127k simulations generally underrepresent Southern Ocean and Antarctic warming, likely due in part to static AIS boundary conditions. Studies identify subsurface Southern Ocean warming as a prerequisite for AIS retreat and potential WAIS collapse, with mechanisms including AMOC weakening and increased surface stratification from meltwater input that suppresses deep convection and warms subsurface waters. High-end LIG sea-level contributions in ice-sheet models often invoke marine ice cliff instability (MICI), though its necessity and parameterization are debated and not included in ISMIP6 projections. Prior coarse-resolution studies of WAIS removal examined isotopic impacts without detailed ocean circulation analysis, and iterative ice–climate coupling has revealed positive feedbacks between stratification and ice melt in future scenarios but has not been widely applied to the LIG. These gaps motivate targeted experiments that separately and jointly evaluate ice geometry changes and meltwater forcing on the coupled climate system.

Model: The coupled climate model GFDL CM2.1 is used, with MOM5.1.0 ocean (1° × 1° horizontal resolution with equatorial refinement to 0.33°, 50 vertical levels; tripolar grid), SIS sea ice (dynamical EVP rheology; three vertical layers, five thickness categories), AM2 atmosphere (2° × 2.5°, 24 levels), and LM2 land surface. A bottom roughness mixing scheme is applied in the ocean. Experimental design: Two baseline simulations—Pre-Industrial (PI) and Last Interglacial (LIG)—follow PMIP4 protocols for orbital parameters, greenhouse gases, and solar constant. Vegetation, soils, and runoff schemes are unchanged in LIG and perturbations; river runoff relocation is fixed. Simulations are initialized from observed T–S (PHC) and integrated for 1500 years to quasi-equilibrium. The LIG run shows AABW formation of 9.8 Sv (vs. 11.1 Sv in PI) and serves as the control for perturbations. Partial AIS removal and freshwater experiments begin from LIG year 1000. Ice-sheet perturbations: Ice geometry changes are derived from Golledge et al. (2015) Antarctic ice-sheet simulations (RCP4.5 year 5000, high scenario). Two AIS mass-loss configurations are constructed by mapping to the CM2.1 grid and ensuring the perturbed ice elevation does not exceed the control elevation and does not fall below sea level: (1) SL4.1: anomaly-based perturbation (RCP4.5 minus 1900 CE), yielding 4.1 m SLE loss; (2) SL7.1: absolute RCP4.5 elevation, yielding 7.1 m SLE loss. Where marine-based ice is removed, land points are converted to ocean at bedrock depth (minimum depth 100 m for numerical stability), extending the ocean grid from 82°S to 85°S. Where terrestrial ice is removed, land elevations are lowered to bedrock while retaining ice-type land-surface properties. Freshwater forcing: Freshwater is added around Antarctica proportional to the longitudinal distribution of ice loss in 60° sectors. The total freshwater volume equals the ice loss volume and is applied to the upper ocean in the two coastal grid cells for 500 years at constant rates: 0.09 Sv (FW4.1) and 0.16 Sv (FW7.1). Combined experiments apply both geometry change and freshwater forcing simultaneously (COMB4.1, COMB7.1). Diagnostics: Meridional overturning circulation (MOC), mixed layer depth, SST, subsurface (100–500 m) temperatures, sea-ice extent, winds, and neutral density (approximately neutral, ω) surfaces are analyzed. Neutral density surfaces are calculated from 50-year averages and zonal means using the Stanley et al. (2021) algorithm, referenced at 49.5°S, 20.5°W at specified depths. Drift correction uses LIG-minus-perturbation differences at matched simulation years.

• Ice geometry change alone (SL4.1, SL7.1) warms the Antarctic coastal ocean and interior: Weddell and Ross Sea annual mean SST increase by ~0.5–1.5 °C (up to 2 °C in summer) with summer sea-ice retreat; East Antarctic interior SAT increases by 2–4 °C (strongest in SL7.1), matching proxy constraints. Summer ice–albedo feedbacks and reduced surface elevation (lapse-rate effect) drive warming, aided by altered winds that reduce southerly offshore flow and allow warmer marine air inland. Subsurface warming of 0.25–0.5 °C along the East Antarctic shelf (0–180°E) occurs at 100–500 m. Mixed layers deepen strongly in the Ross Sea (to 400–500 m in SL4.1), increasing AABW formation from 9.8 Sv (LIG control) to 13.0 Sv (SL4.1) and 11.2 Sv (SL7.1). ACC fronts shift, with regional cooling in the Amundsen–Bellingshausen sector and subsurface pattern changes linked to winds and bathymetry. • Freshwater forcing alone (FW4.1, FW7.1) cools the surface but warms subsurface waters: Coastal SSTs decrease by ~1–2 °C (FW4.1) and 2–3 °C (FW7.1), with large sea-ice expansion and reduced deep convection, particularly in the Weddell Sea. Southern Ocean deep-water formation declines from 9.8 Sv (LIG) to 7.8 Sv (FW4.1) and 5.8 Sv (FW7.1). Subsurface temperatures (100–500 m) increase by up to 2 °C in the Ross Sea and modest warming elsewhere around East Antarctica, due to enhanced stratification that isolates the cold surface from warmer circumpolar deep water and flattens isoneutral slopes. Antarctic SAT decreases, opposite to proxy-indicated warming. • Combined perturbations (COMB4.1, COMB7.1) yield mixed surface signals and amplified subsurface and deep warming: Surface shows Antarctic warming from ice loss but Southern Ocean cooling dominated by meltwater, with sea-ice extent larger than LIG by 4.7% (COMB4.1) and 12.8% (COMB7.1). Weddell Sea SSTs decrease by ~1–1.5 °C (COMB4.1) and ~1.5–2 °C (COMB7.1); Amundsen–Bellingshausen cooling is ~1 °C and 2–3 °C, respectively. Subsurface warming exceeds meltwater-only cases in some regions, notably >2 °C in the Ross Sea and 0.5–1 °C more broadly around East Antarctica in COMB7.1. Deep Southern Ocean exhibits non-linear warming stronger than the sum of parts: despite deep cooling in SL cases and modest warming in FW cases, COMB shows larger, more extensive warming (e.g., 0.2–0.4 °C to 50–60°S at intermediate depths in COMB4.1; 0.5–1.0 °C in COMB7.1). AABW formation is further reduced to 6.0 Sv (COMB4.1) and 4.7 Sv (COMB7.1). • Wind and circulation changes: Ice-loss experiments weaken midlatitude westerlies (reduced baroclinicity from polar surface warming) and southerly offshore winds over East Antarctica, allowing inland advection of marine air and strengthening the Antarctic Slope Current. Meltwater experiments strengthen westerlies (enhanced equator-to-pole gradient). A new ocean gateway at 73°S, 86–80°W permits small inter-basin throughflows with directions differing between ice-loss-only and combined cases. • Proxy comparison: Ice-loss-only experiments match East Antarctic SAT proxies (0.9–3.3 °C warming) better than meltwater or combined cases. Southern Ocean SST proxies (mostly 40–50°S) generally indicate warming larger than simulated; models tend to be cold relative to proxies, and peak surface cooling from meltwater suggests peak melt may have lagged peak SST warming during the LIG.

Separating ice geometry and meltwater perturbations clarifies their contrasting roles: lowering the AIS surface and removing marine-based ice warms East Antarctica via lapse-rate and albedo effects, modifies winds to reduce offshore flow, and enhances AABW in some cases; in contrast, freshwater input strengthens stratification, suppresses deep convection, cools surface waters, and warms subsurface layers that can destabilize marine-based ice margins. The combined perturbations produce a non-linear deep ocean response, with reduced ventilation and flatter isoneutral slopes advecting relatively warmer deep waters poleward, amplifying deep Southern Ocean warming beyond meltwater alone despite cooling in the ice-loss-only deep ocean. These mechanisms help reconcile aspects of LIG proxy–model discrepancies: East Antarctic SAT warming is consistent with partial AIS loss, while widespread Southern Ocean surface cooling from meltwater implies that maximum meltwater discharge likely did not coincide with peak SST warmth. The simulated subsurface warming around the Ross Sea and East Antarctic shelves highlights a feedback pathway that could promote further AIS retreat during the LIG and is relevant to future projections. Spatial limitations of Southern Ocean proxies (mostly 40–50°S) complicate direct validation of simulated high-latitude changes, but the mechanistic insights underscore the importance of interactive ice–ocean–atmosphere coupling in paleoclimate and future climate assessments.

Partial removal of the Antarctic Ice Sheet during the Last Interglacial can explain observed East Antarctic surface warming through surface-elevation reduction, albedo effects, and wind changes, while enhanced meltwater fluxes primarily cool the surface ocean, increase stratification, and warm subsurface waters. When combined, ice loss and meltwater inputs produce stronger and more extensive subsurface and deep Southern Ocean warming than either perturbation alone, alongside reduced AABW formation, indicating non-linear feedbacks that could further destabilize marine-based sectors of the AIS. These findings suggest that models lacking dynamic ice–climate interactions may underestimate East Antarctic warming and misrepresent subsurface warming feedbacks during warm periods. Future research should develop fully coupled climate–ice sheet simulations for the LIG and beyond, refine the timing and magnitude of LIG meltwater pulses relative to SST peaks, and expand proxy coverage at higher southern latitudes to better constrain surface and subsurface ocean changes.

• The Antarctic ice sheet is prescribed rather than dynamically coupled to the climate model; interactive ice–ocean coupling and feedbacks are not fully represented. • Atmospheric resolution (~2°) under-resolves katabatic winds and fine-scale topographic influences over Antarctica. • The chosen AIS perturbations are based on a specific future-forced ice-sheet scenario mapped to LIG conditions to elicit a strong signal; they are not a vetted LIG reconstruction. • Freshwater forcing is idealized (constant over 500 years, coastal distribution by 60° sectors), and the timing relative to peak SST anomalies is uncertain. • River runoff relocation remains fixed, and newly created shelf bathymetry imposes a minimum 100 m depth for numerical stability, potentially affecting shelf processes. • The model exhibits small residual drift; although accounted for by anomaly calculations, it adds uncertainty. • Oxygen isotopes are not simulated, precluding direct comparison to δ18O records. • Southern Ocean SST proxies are mostly at 40–50°S, whereas many modeled anomalies occur poleward of 50°S, limiting direct validation.