In PNAS, two companion studies by Levy et al. (1) and Gasson et al. (2) underscore the importance of ice-proximal geologic data for improving computer models of Antarctic ice sheet response to oceanic and atmospheric warming. Current knowledge of Antarctic ice sheet evolution (∼40–0 Ma) is based on deep-sea records of global ice volume, deep ocean temperature, and carbon cycling preserved in the calcium carbonate shells of benthic foraminifera (3, 4). Shackleton and Kennett (5) hypothesized, from moderate-resolution southwest Pacific Ocean benthic foraminifer stable oxygen (δ¹⁸O) and carbon (δ¹³C) isotope compilations, that the deep ocean cooled ∼15 °C through the Cenozoic and that Antarctic ice sheets expanded significantly at the Eocene–Oligocene boundary, varied dynamically until the middle Miocene climate transition (MMCT; 14.2–13.8 Ma), and then rapidly expanded and stabilized. Over the last 41 y, paleoceanographers have used deep-sea sediments recovered by scientific ocean drilling programs, including the International Ocean Discovery Program (2013–2023), to increase the resolution of the global deep-sea stable isotope record (3⇓–5), develop geochemical methods to separate ice volume and temperature signals contained in the δ¹⁸O signal (4), and resolve climate forcings and feedbacks involved in Antarctic ice growth and global climate evolution (4, 6).

Because of a lack of high-quality ice-proximal drill sites, only a handful of studies (e.g., refs. 1 and 7) directly link Antarctic ice sheet variability to far-field deep-sea δ¹⁸O (3⇓–5) and passive margin sea-level records (8). Such linkages are required to improve understanding of the timing and magnitude of ice volume fluctuations, relative contributions of Antarctica’s ice sheets to global eustacy, individual histories of the East and West Antarctic ice sheets, and climatic boundary conditions favorable for ice growth and decay (3). In PNAS, Levy et al. (1) present well-dated ice-proximal evidence for dynamic Antarctic ice sheets and climate in the Ross Sea during the warm Miocene Climatic Optimum (MCO; ∼17–14.5 Ma) and subsequent cooling of the MMCT (Fig. 1) (3⇓⇓–6, 9, 10). In a companion paper, Gasson et al. (2) describe major advances in coupled climate and ice sheet model ability to simulate observed Antarctic ice sheet variability in a narrow range of atmospheric pCO₂ (partial pressure of CO₂) and temperature. Taken together, these two studies provide insights into climate forcings, feedbacks, and sensitivity during the enigmatic middle Miocene.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Early to middle Miocene section of AND-2A correlated to selected datasets. (A) Southern Ocean and Antarctica with present day locations of AND-2A and Ocean Drilling Program (ODP) Site 1171. (B) AND-2A lithologic log tied to age with gray shading indicating time missing from the AND-2A core; purple dashed lines: maximum ice sheet advance; blue shading: cold polar intervals; green dashed lines: peak warmth (1). Southern Ocean ice volume (δ¹⁸O seawater; blue; black arrows indicate intervals of glacial expansion), bottom water temperatures (purple), and sea surface temperatures (black), derived from foraminifer stable isotope and trace metal proxies from ODP Site 1171 (9, 13). Proxy atmospheric pCO₂ data include boron isotopes (blue circles), alkenones (black triangles), stomata (green diamonds), and paleosols (orange squares) (11). Thick gray line represents 21 point weighted average (1). (C) Cold (gray box) and warm (green box) ice sheet simulations after Gasson et al. (2). (D) Schematic of high-latitude drilling transects proposed to capture ice-proximal to ice-distal glacial-marine to marine sedimentary sequences.

The MCO is of particular interest because geologic records indicate that atmospheric pCO₂ fluctuated within a narrow range [280–500 parts per million by volume (ppmv)] (11), encompassing both present day concentrations and predictions for the next half-century (12). Average global MCO temperatures were 3–4 °C warmer than present with reduced equator-to-pole thermal gradients (10, 13, 14). Deep-sea stable isotopes suggest substantial Antarctic ice sheet variability and a shift in global carbon cycling (3⇓⇓–6, 9, 10). Thus, Miocene geologic data suggest that either Earth’s climate system is so sensitive to pCO₂ that slight increases above modern levels may result in substantial warming through positive feedbacks, or that MCO warmth, MMCT cooling, and the associated Antarctic ice sheet response was decoupled from carbon cycling (9⇓⇓⇓⇓–14). To date, MCO warmth and MMCT cooling cannot be explained by simply changing atmospheric pCO₂, indicating that a combination of influences (e.g., tectonically driven ocean circulation changes and orbital forcing) worked to increase middle Miocene climate sensitivity (9⇓–11, 13, 14).

Levy et al. (1) present evidence of middle Miocene Antarctic ice sheet fluctuations from an ice-proximal sedimentary sequence recovered in Southern McMurdo Sound (AND-2A) by the multinational Antarctic Geologic Drilling Project (Fig. 1). The recovery of AND-2A was technologically remarkable. It was drilled from a floating sea-ice platform using riser drilling technology in shallow water, resulting in high (>80%) core recovery. Using an integrated age model, Levy et al. (1) reconstruct MCO glacial dynamism long-suspected from both near-field Antarctic terrestrial and continental shelf studies, and far-field ice volume and sea-level records (3, 8⇓–10, 13⇓⇓–16). Because of reworking inherent in Antarctic margin sediments, the multiproxy approach strengthens Levy et al.’s middle Miocene paleoenvironmental interpretations. Ocean temperatures from two recently developed geochemical paleothermometers (17, 18) are integrated with pollen and geochemical data to provide insight into the evolution of the regional environment. Although these paleothermometers are not yet calibrated adequately to assess absolute ocean temperatures at high latitudes (17, 18), intervals of relative ocean warmth coincide with independent marine and terrestrial pollen, spore, and geomorphologic evidence for regional warmth, meltwater influx, and ice retreat into the Transantarctic Mountains between 16.4 and 15.8 Ma (1, 10, 16). Additionally, a series of three chronologically constrained missing and condensed sections in the AND-2A drill core mark the MCO onset and MMCT progression of ice advance across the Ross Sea continental shelf (15).

Levy et al. (1) integrate the shallow ice-proximal AND-2A record with far-field deep-sea records. This integration reveals that dynamic Antarctic ice sheets existed when the southern high latitudes were relatively warm and pCO₂ was 300–500 ppmv, confirming the MCO’s reduced equator-to-pole thermal gradient (Fig. 1) (1, 4, 9⇓–11, 13). Models cannot yet replicate this result without very high (>1,000 ppmv) atmospheric pCO₂ concentrations not indicated by the majority of paleo-pCO₂ proxy records (2, 6, 11, 14). The timing of missing sequences in the AND-2A drill core coincide with deep-sea geochemical evidence for ice growth, further supporting observational and model interpretations of Antarctic ice sheet expansion (1, 2, 8). Although Levy et al. (1) discuss missing sections at 14.7–14.6 Ma and 14.1–13.7 Ma and highlight their association with low (<300 ppmv) pCO₂ (11), they do not interpret the 15.5–14.8 Ma gap in the AND-2A record. This gap is also coincident with deep-sea geochemical changes, confirming previous interpretations that Antarctic ice growth began at ∼15.5 Ma when high-latitude temperatures were relatively warm (8, 13) and atmospheric pCO₂ was slightly higher than present (∼400–500 ppmv) (Fig. 1) (8, 11). This evidence, coupled with model consensus that MCO warmth remains difficult to explain without invoking significant but poorly quantified climate feedbacks (14), suggests the modern relationship between atmospheric pCO₂ and temperature may not exist during the middle Miocene.

In a companion paper, Gasson et al. (2) present new climate-ice sheet modeling advances that enable dynamic Antarctic ice sheets within a narrow range of atmospheric pCO₂ and temperature (11). These results are a dramatic step forward because they solve a model hysteresis problem, which limits ice retreat from a fully glaciated state unless unrealistically large atmospheric pCO₂ or temperature increases are imposed (2, 6, 14). Isotope-enabled model reconstructions indicate that large-scale Antarctic ice sheet expansion caused the gaps in the AND-2A sequence, rather than localized Transantarctic Mountain outlet glacier variability. These results are consistent with deep-sea geochemical and Ross Sea continental shelf sequence stratigraphic data (1, 9, 15). Although the long-standing mismatch between data and models for middle Miocene ice sheet behavior may be resolved (6), a major uncertainty remains: the reconstruction of past Antarctic bedrock topography, critical for determining individual Antarctic ice sheet histories, global eustacy contributions, and the role of ocean thermal forcing on past ice sheets (2, 8, 9, 19). Depending on the choice of subglacial bedrock topography, Gasson et al. (2) reveal that the middle Miocene West Antarctic Ice Sheet may not have been marine-based. As such, its Miocene response may not be a useful analog for its future response in a warming world. Furthermore, East Antarctica’s subglacial basins may be more important contributors to the past global eustatic signal (2, 19). Today, the marine-based East Antarctic Ice Sheet contains 19 m of sea-level–equivalent (19) and –associated outlet glaciers are experiencing instability because of warm ocean waters interacting with deep grounding lines (20); however, the East Antarctic Ice Sheet has traditionally been considered relatively stable (5, 10).

The PNAS papers of Levy et al. (1) and Gasson et al. (2) highlight recent successes of integrated drilling and modeling studies. These studies indicate that a synchronous continent-wide Antarctic response to climate perturbations during past climate transitions is unlikely. Thus, there is a critical need for additional scientific drilling in multiple ice-proximal Antarctic margin locations and in the ice-distal Southern Ocean, particularly at paleolatitudes >50°S and paleodepths <2.5 km. Additionally, to address Miocene climate sensitivity and relationships between tectonics, ocean circulation, Earth’s orbits, and carbon cycling (3⇓⇓–6, 8⇓⇓–11, 13, 14), a network of high-resolution Miocene sedimentary sequences from a range of depths and latitudes is required. Strategic drilling targets are also needed to study the tectonic evolution of oceanic gateways, as large uncertainties exist surrounding the timing of significant Miocene tectonic changes [e.g., the Drake Passage (21), Central American Seaway (22), and Tethys gateways (23)]. Resulting data will enable development of a consistent Miocene modeling framework, as has been accomplished for the warm Pliocene (24).