A new global ice sheet reconstruction for the past 80 000 years

Reconstructing the history of global ice sheets is crucial for understanding paleoclimate, present-day Earth deformation, and future sea-level change. Glacial isostatic adjustment (GIA)-based reconstructions are commonly used, but uncertainties remain, particularly regarding the "missing ice" problem during the LGM (26,000–19,000 years before present). This discrepancy arises from a difference between far-field sea-level indicators and modeled sea levels from existing ice sheet reconstructions, representing an apparent 8–28 m of unaccounted-for ice volume. Before the LGM, marine δ¹⁸O proxy records are frequently used to infer ice volume due to the scarcity of paleo sea-level data. This study aims to create a new ice sheet reconstruction independent of these proxy records and far-field sea-level data to address these uncertainties and assess the validity of existing methods. The study leverages geological and geophysical constraints to create a more physically realistic model and overcome previous assumptions in other reconstructions.

Previous ice sheet reconstructions have employed various methods, often relying on GIA constraints without considering ice physics, using regional thickness scaling parameters, or averaging low-resolution model simulations. All these approaches incorporate assumptions about far-field sea level in their tuning strategies. Prior to the LGM, the absence of extensive paleo sea-level data has led researchers to rely on oceanic δ¹⁸O fluctuations as proxies to tune their reconstructions. Recent studies in the Hudson Bay region challenge the assumptions of consistently large ice sheets during MIS 3, suggesting potentially ice-free conditions and a climate analogous to the present. These findings indicate a smaller Laurentide Ice Sheet extent than previously assumed and raise questions about the accuracy of proxy-based sea-level reconstructions. The reliability of dating methods used to infer reduced ice sheet extent during mid-MIS 3 is also debated. This paper aims to address inconsistencies and biases present in the existing literature by proposing an independent reconstruction.

The authors developed PaleoMIST 1.0 (Paleo Margins, Ice Sheets, and Topography), a global ice sheet reconstruction independent of indirect proxy records and far-field sea-level records. The reconstruction uses the software ICESHEET, which considers ice sheet physics, geological constraints on ice margin location and flow direction, and near-field sea-level data. The process started at 80,000 years BP with a relatively high global sea level, using a 2500-year time step. New margin reconstructions were developed for North American and Antarctic ice sheets based on geological data, incorporating previously published Eurasian ice sheet reconstructions. For the North American ice sheets, the reconstruction aimed to maximize ice sheet extent, resulting in a MIS 3 margin extent that is generally larger than recent assessments. Two scenarios were created for MIS 3: a minimal scenario with a temporary Hudson Bay ice-free period, and a maximal scenario with continuous ice cover. GIA modeling, using the SELEN program, accounted for sea-level equation, shoreline migration, and Earth deformation. The Earth model employed a three-layered structure with specific lithosphere thickness, upper and lower mantle viscosity values. The reconstruction iteratively refined ice sheet configuration based on GIA-calculated sea level and geological constraints, aiming for consistency between ice load and observed sea level and deformation. The Laurentide Ice Sheet reconstruction was primarily tuned to achieve the observed LGM lowstand, while the Eurasian ice sheets were tuned to fit postglacial sea-level indicators and chosen mantle viscosity.

PaleoMIST 1.0 successfully achieves the LGM sea-level lowstand observed at many far-field locations without requiring additional ice volume, resolving the "missing ice" problem. The model suggests that the origin of the missing ice problem in previous studies likely stemmed from initial assumptions about ice distribution and Earth rheology. The ice volume in PaleoMIST 1.0 does not match the pre-LGM δ¹⁸O values based on empirical relationships between ice volume and δ¹⁸O but aligns with some sea-level indicators and GIA studies. This suggests that the relationship between δ¹⁸O proxy records and sea level/ice volume might be more complex than previously thought. The Laurentide Ice Sheet's geometry dominates the global sea level contributions in the reconstruction. The model produced two MIS 3 scenarios for the Laurentide Ice Sheet: minimal (with temporary Hudson Bay deglaciation) and maximal (with continuous ice cover). The difference in globally averaged sea level between the two scenarios is about 14 meters of sea level equivalent (SLE). The reconstruction's calculated sea level at the LGM is compatible with far-field constraints at various locations (Barbados, Sunda Shelf, Bonaparte Gulf, and Great Barrier Reef). The calculated maximum ESL fall and ice volume are substantially lower than previous estimates, further supporting the resolution of the missing ice problem. The Southern Hemisphere ocean, along with coastal regions of Asia, show calculated sea levels below the global average. Some regional discrepancies in calculated versus observed sea level are attributed to the choice of a higher lower mantle viscosity. The model highlights the model-dependency of defining global sea level. In contrast to the LGM results, the reconstruction's pre-LGM sea level (during MIS 3) does not align with δ¹⁸O-based reconstructions, presenting a new "missing ice" problem that requires further investigation.

PaleoMIST 1.0's success in resolving the LGM "missing ice" problem demonstrates the importance of using a reconstruction method independent of far-field sea-level data and adhering to ice physics and geological constraints. The discrepancy between PaleoMIST 1.0 and δ¹⁸O-based reconstructions during MIS 3 underscores the complexities of interpreting marine oxygen isotope records and the need for integrating multiple lines of evidence. The model highlights the need to consider each location's relative sea level independently, rather than targeting a specific global sea-level lowstand. The model-dependent nature of defining global sea level also requires further clarification. The higher lower mantle viscosity used in PaleoMIST 1.0, although compatible with far-field indicators, might contribute to the regional sea level differences. Future studies should explore refining the relationship between δ¹⁸O and sea level, using techniques like water isotope modeling, and reassessing current ice sheet extent indicators for MIS 3.

PaleoMIST 1.0 provides a novel, data-driven global ice sheet reconstruction that addresses the long-standing "missing ice" problem at the LGM. The reconstruction is consistent with far-field sea-level data and offers a new perspective on ice sheet dynamics during this period. The study also highlights limitations in the existing relationship between δ¹⁸O proxies and sea level, particularly during MIS 3, suggesting a need for further research integrating various proxies and improving model fidelity. Future work should focus on refining the model's temporal resolution, exploring different Earth rheology assumptions, and investigating the complex relationship between δ¹⁸O and sea level to further constrain the pre-LGM ice sheet history.

The 2500-year time step used in PaleoMIST 1.0 limits the model's ability to capture rapid near-field Holocene sea-level changes. The reliance on equilibrium ice sheet dynamics in ICESHEET might not perfectly capture the actual non-equilibrium behavior of ice sheets. Uncertainty exists in the chronological constraints on ice margin fluctuations, particularly during MIS 3. The reconstruction does not include contributions from smaller glaciers and ice caps, thermal expansion, or groundwater storage, which could slightly affect the estimated ice volume. Regional variations in sea level, particularly the lower sea levels in some far-field regions, might be influenced by the relatively high lower mantle viscosity used in the model.