Ice sheets have a complex interplay with the Earth's systems (atmosphere, ocean, lithosphere, and asthenosphere), exhibiting both positive and negative feedbacks that influence their growth and retreat. These interactions are responsible for the cyclical patterns observed in glacial-interglacial cycles. Freshwater discharge from melting ice sheets, potentially reaching magnitudes comparable to the total global river discharge, can significantly influence climate. The Atlantic Meridional Overturning Circulation (AMOC), a crucial heat transport mechanism, is particularly sensitive to freshwater input. Experiments simulating this freshwater input have shown its impact on ocean stratification, deep-water formation, and global climate. Future projections suggest substantial freshwater discharge from the Antarctic Ice Sheet (AIS), potentially exceeding 1 Sv in the next century, comparable to significant meltwater events of the past. However, predicting the net effect of these feedbacks on ice sheet retreat remains uncertain, as previous studies using different models have yielded conflicting conclusions, some suggesting a positive feedback leading to accelerated ice loss, and others indicating a negative feedback that could slow down ice sheet retreat. This study aims to address these uncertainties by using a coupled ice sheet-climate model to simulate the interactions between ice sheet freshwater flux (FWF) and climate under various future warming scenarios.

Previous research on ice sheet freshwater discharge and its climate feedbacks has produced mixed results. Some studies using offline-coupled ice sheet models suggested a positive feedback, where meltwater-induced subsurface warming accelerates ice loss and sea level rise. Conversely, other studies pointed to a negative feedback, with surface cooling caused by meltwater suppressing ice melting and delaying the consequences of anthropogenic warming. The inconsistencies arise from differences in model complexities, coupling strategies (offline vs. online), and the inclusion of certain key processes such as hydrofracturing and ice cliff failure, which are crucial for representing the Marine Ice Cliff Instability (MICI) mechanism. While several earth system models with integrated ice sheet components exist, their application to studying centennial-scale ice sheet-climate feedbacks has been limited. The offline coupling approach, using prescribed climate conditions, fails to adequately capture the dynamic interactions between the ice sheet, atmosphere, and ocean, leading to uncertainties about the net feedback effect.

This study employed a three-dimensional ice sheet model (ISM), PSUICE3D, quasi-synchronously coupled to a reduced-complexity Earth system model of intermediate complexity (EMIC), UVic-ESCM, with a coupling time step of 1 year. PSUICE3D simulates ice sheet surface and basal mass balance, including basal sliding and bedrock deformation, driven by bias-corrected climate fields from UVic-ESCM. UVic-ESCM includes a three-dimensional ocean model, a land model, and a two-dimensional atmospheric model. Using a reduced-complexity climate model enabled extensive simulations with varying configurations, climate sensitivities, emission scenarios, and initial conditions. Four coupling configurations were designed: (1) constant pre-industrial FWF for both ice sheets, (2) interactive FWF from the Greenland Ice Sheet (GIS) only, (3) interactive FWF from the AIS only, and (4) fully interactive FWF from both ice sheets. The model’s equilibrium climate sensitivity (ECS) was explored by scaling atmospheric CO2 concentrations to emulate different ECS values (3.0 °C, 4.0 °C, and 5.6 °C), encompassing a range of future warming scenarios represented by six Shared Socioeconomic Pathways (SSPs). Ensemble runs (10 members) were conducted for two future scenarios (moderate and intensive warming) to reduce internal variability noise. The feedback factor (γ) was calculated to quantify the influence of ice sheet FWF-climate interactions on AIS mass loss, defined as 1 minus the ratio of mass loss rates with and without interactive FWF. The model also considered the impact of icebergs (though not explicitly simulated, treated as imposed sea ice flux) on surface cooling and associated energy budgets.

The study revealed significant scenario-dependence of ice sheet freshwater-climate feedbacks. The peak ice sheet FWF varied by an order of magnitude across the scenarios. The warmest scenario showed peak FWF of 0.37 Sv for the GIS and -1.1 Sv for the AIS, leading to near-total GIS loss by 2500 and a significant AIS contribution to sea-level rise exceeding 10 m. WAIS collapse, initiated around 2100, contributed more than 4 m to sea level rise over three centuries in this scenario. Increased ice sheet FWF generally reduced future global warming, with the most substantial reductions associated with higher FWF. AMOC weakening was more significant in scenarios with more intensive ice sheet melt. The GIS and AIS exhibited contrasting responses to interactive FWF. The GIS retreat was slower in most scenarios with interactive FWF, reflecting an overall negative feedback, primarily due to the AMOC weakening and reduced northward heat transport. However, exceptions were observed in the warmest scenarios, potentially influenced by the Antarctic FWF. The AIS showed dominance of positive feedback in moderate warming scenarios, accelerating ice loss. In the intensive warming scenario, the negative feedback prevailed after 2100, significantly slowing WAIS collapse. This transition was attributed to the differing spatial locations of the positive (basal melting) and negative (surface cooling) feedbacks. In Antarctica, surface cooling from enhanced stratification and sea ice expansion counteracted subsurface warming, leading to a net negative feedback in the intense warming scenario after the initial phase of rapid ice shelf loss. A feedback factor analysis revealed a positive feedback at low ice loss rates and a negative feedback at higher rates, indicating a transition point around 0.2 Sv. Inter-hemispheric ice sheet interactions, mediated through the AMOC, were observed. Greenland meltwater weakened the AMOC, leading to SH warming and increased Antarctic ice loss, while Antarctic meltwater strengthened the AMOC, enhancing GIS ice loss, demonstrating the bipolar seesaw mechanism.

The findings reconcile seemingly contradictory conclusions from previous studies, showing that the net feedback's sign depends heavily on the intensity of warming and the extent of ice shelf loss. The negative feedback, primarily driven by surface cooling, is more prominent in scenarios with intense warming and extensive ice shelf loss, while the positive feedback from basal melting is dominant in scenarios with moderate warming and larger ice shelf areas. The study highlights the importance of using high-resolution, dynamically comprehensive models that can resolve crucial processes such as ice shelf cavity circulations and iceberg dynamics to accurately capture the complexities of ice sheet-climate interactions. The findings also underscore the trans-hemispheric impacts of ice sheet FWF, mediated through changes in the AMOC. This necessitates a more comprehensive approach that incorporates these interactions into climate models to improve projections of future ice sheet retreat and sea level rise.

This study demonstrates the complex interplay of competing climate feedbacks resulting from ice sheet freshwater discharge, emphasizing the strong dependence of the net feedback on the rate of warming. While acknowledging model limitations, the findings highlight the crucial role of ice sheet-climate interactions via freshwater flux in future ice sheet retreat and sea level rise projections. Further research using more sophisticated models with higher resolution and improved representation of key processes is needed to refine these findings and reduce uncertainties in future projections. Future models should specifically focus on improved spatial resolutions, explicit modeling of ice shelf cavity circulations, realistic representation of iceberg dynamics, and accurate representation of freshwater injection depth to more fully capture the complexities of these interactions.

The study acknowledges several limitations. The reduced-complexity climate model used has lower spatial resolution and simplified representations of certain processes compared to state-of-the-art climate models. Specifically, limitations include lower spatial resolutions in both ocean and atmosphere modules, lack of explicit modeling of ice shelf cavity circulations, simplified iceberg representation (treated as added sea ice), and top-layer injection of freshwater flux instead of representing the actual depth of injection. These limitations may influence the quantitative aspects of the results, particularly regarding the precise strength and timing of the feedback transitions. These limitations must be addressed in future studies using more comprehensive models.