The mid-Pliocene warm period between ca. 3 Ma and 3.3 Ma was the last time the atmospheric CO2 concentration was as high as it is today. Global sea level was higher than it is now, because the Greenland ice sheet and some portion of the Antarctic ice sheets were not present. If atmospheric CO2, global ice volume, and sea level are linked as closely as the geologic record indicates, we care a lot about the middle Pliocene, and it’s truly important to figure out exactly what the ice sheets looked like at that time.

This task is not easy. Although it is easy to identify past ice sheet expansions from the distinctive geologic record around the margins of the present-day ice sheets, it is not at all easy to unambiguously determine if and when a particular ice sheet was smaller than it is now. Mainly, this is because any direct geologic evidence of this condition—if it exists at all—is beneath the ice sheet. Lacking access to the ice sheet bed, we’re left with indirect and often circumstantial evidence. Extracting the correct answer from this evidence is the subject of two papers in this issue of Geology: Winnick and Caves (2015, p. 879) revise the estimate of mid-Pliocene global ice volume inferred from the marine oxygen isotope record, and Austermann et al. (2015, p. 927) account for slow changes in continental elevation driven by mantle dynamics in establishing the correct boundary conditions for model simulations of Pliocene ice sheets. These papers suggest that the Pliocene East Antarctic Ice Sheet (EAIS) was either bigger (Winnick and Caves, 2015) or smaller (Austermann et al., 2015) than previously believed, and highlight the fundamental difficulty of determining when an ice sheet wasn’t there.

So far, investigating Pliocene ice sheet changes in Antarctica has proceeded along two tracks. First, asking “what happened” using geologic evidence; second, asking “what could have happened” using ice sheet models driven by Pliocene boundary conditions.

The first approach to finding out what happened was to look at the direct evidence: the terrestrial stratigraphic record of Pliocene events in Antarctica that is exposed above the ice sheet at present. This record is not extensive and was thoroughly explored in the 1990’s in an effort to resolve a controversy about the age of the Sirius Formation, a sedimentary deposit containing a temperate-climate forest flora. This led to the overall conclusion that the terrestrial geologic record provided no evidence of significant Pliocene deglaciation in Antarctica, and in fact was most consistent with a Pliocene configuration of the EAIS similar to present.

Indirect evidence for Antarctic ice volume change, on the other hand, is abundant and tells a different story. It’s been recognized for some time that (1) the oxygen isotope composition of mid-Pliocene seawater requires reduced global ice volume relative to the present, and (2) there exist extensive shoreline deposits of mid-Pliocene age well above present sea level on stable continental margins worldwide. Estimates of mid-Pliocene global mean sea level based on this evidence range quite widely between ∼10 m and 40 m above present, with consensus estimates near +25 m.

The +25 m scenario requires the total absence of the Greenland Ice Sheet (∼7 m), complete deglaciation of marine-based portions of the West Antarctic Ice Sheet (WAIS) (∼5 m), and, in addition, a large contribution—as large as or larger than Greenland and West Antarctica together—from East Antarctica.

The present EAIS contains ∼50 m sea-level equivalent, so this mass balance implies that it looked very different in the Pliocene, and led to ice sheet model experiments with Pliocene boundary conditions aimed at determining whether extensive EAIS deglaciation could have happened. Several such efforts, for example, that of Pollard and DeConto (2009), found that it couldn’t have happened: their model simulation generated a maximum of 7 m from East and West Antarctica combined.

This mismatch implied that either the geologic evidence was wrong or the model simulation was wrong, and subsequent research has explored both possibilities. Pollard et al. (2015) revamped their model to include additional processes related to the so-called marine ice margin instability, a positive feedback incurred when marine ice margins, whose iceberg calving rate is proportional to water depth, begin retreating into a subglacial basin. This revised model predicts mid-Pliocene deglaciation of nearly all marginal basins in East Antarctica where ice is currently grounded below sea level, with significant ice-volume changes from the largest of these, the Wilkes and Aurora Basins. It ups the Antarctic contribution to global sea level to 17 m, which is approximately consistent with the +25m scenario. Austermann et al. also highlight the importance of the Wilkes Basin by showing that simulations of mantle dynamics over the past 3 m.y. require that the elevation of the Wilkes Basin has increased by ∼100 m during that time. The Wilkes Basin was deeper in the Pliocene than it is now, implying a larger potential deglaciated area than one would infer from present topography, and increasing the model sea-level contribution of this sector of Antarctica by a further ∼2 m.

This work—adding processes to the model simulations and improving the boundary conditions—appears to resolve the model-data mismatch for Pliocene ice sheet change in East Antarctica in favor of the +25m scenario. It shows that this scenario is not, in fact, inconsistent with a physically reasonable ice sheet model. At the same time, however, new analyses of the geologic evidence have shown that the +25m estimate from geologic data, not the initial ice sheet model predictions, might be wrong.

A group of researchers led by Maureen Raymo (the “PLIOMAX” project; www.pliomax.org) took a closer look at Pliocene shorelines on passive continental margins and showed that a significant fraction of the present elevation of these shorelines is not due to higher eustatic sea level in the Pliocene, but instead to subsequent changes in continental elevation due to (1) glacioisostatic adjustments caused by differences in the modern and Pliocene ice sheet configuration, and (2) dynamic topography, that is, transient departures from isostatic equilibrium between continental crust and underlying mantle, caused by mantle flow on long time scales (see Raymo et al., 2011). They have argued that mid-Pliocene sea level was no more than +20 m, possibly much less, and that no Pliocene reduction in EAIS volume relative to present would be consistent with the shoreline data (Rovere et al., 2014).

Winnick and Caves come to a similar conclusion with respect to the Pliocene ice volume estimate derived from marine oxygen isotope records. They point out that existing such estimates rely on an assumed isotope composition for Antarctic ice that is appropriate for the late Pleistocene. However, it is nearly certain that Antarctic climate was systematically warmer in the Pliocene, which implies a different isotope composition for ice. When this is taken into account, Pliocene oxygen isotope records imply significantly less than 25 m of sea-level rise and, in fact, do not require a reduction in EAIS volume relative to present.

So to summarize, updated ice sheet models now suggest that substantial mid-Pliocene deglaciation in East Antarctica, enough to cause 25 m of sea-level rise, could have happened. This conclusion, frankly, would imply dire possibilities for the future of East Antarctica. But new analysis of the geologic record of Pliocene sea level change—the actual, albeit indirect, evidence for changes in the size of ice sheets—suggests that it didn’t happen. Although this conclusion leads to less concern about the future of the EAIS, it also highlights the importance of finding and exploiting any geologic evidence from Antarctica itself that could resolve the overall issue by directly telling us how big the EAIS was in the Pliocene. The problem is that it is surprisingly difficult to find any such evidence.

Some recent studies have argued that marine sedimentary records from the Antarctic continental shelf may provide such evidence. The ANDRILL drilling program at the McMurdo Ice Shelf in the western Ross Sea recovered a stratigraphic record that provides some constraints on Pliocene changes in the size of the WAIS, but the core site is on the wrong side of the EAIS to provide any information about the behavior of its marine margins. Cook et al. (2013) hypothesized that the provenance of mid-Pliocene ice-rafted sediment offshore of Wilkes Land implied complete deglaciation of the Wilkes Basin; however, this hypothesis relies on hard-to-evaluate assumptions about both basal ice sheet conditions and the geochemistry of subglacial regions that have never been sampled. Fundamentally, although marine sedimentary records offshore of East Antarctica clearly record the mid-Pliocene warm period, their connection to actual ice margin positions is speculative.

The terrestrial record of EAIS change primarily consists of glacial deposits that are preserved above the level of the present ice sheet. If the EAIS was, in fact, thicker than present in the mid-Pliocene—a potentially viable hypothesis based on the general idea that a warmer climate implies higher precipitation and hence steeper ice surface slopes and a thicker ice sheet—we should observe mid-Pliocene glacial deposits above the present ice sheet. There do exist glacially transported erratic boulders with apparent cosmogenic-nuclide exposure ages in the 2–4 Ma range, but even the very slow rock weathering rates prevalent in Antarctica have a significant effect on rock surfaces of this age, and so far no one has addressed this issue at the level of precision needed to unambiguously identify deposits emplaced at 3–3.3 Ma. This is possible in principle, however, and one straightforward, feasible, and potentially important line of research into the Pliocene extent of the EAIS is to systematically apply exposure-dating methods to known glacial deposits in East Antarctica that could be Pliocene in age.

Glacial deposits above the present ice sheet surface can by definition only tell us about times when the EAIS was larger than present. To prove that it was smaller, we would need evidence that lies beneath the present ice surface. This evidence might exist. Specifically, if the EAIS was thinner in the past, bedrock surfaces that are now shielded by ice would have been exposed to the surface cosmic-ray flux. If we detected significant concentrations of cosmic-ray–produced nuclides in a rock sample collected by drilling through the ice sheet, it would be truly unambiguous evidence that the ice sheet had been thinner in the past. Using this information to determine when that happened is more challenging, but theoretically feasible. A similar argument applies to sections of the ice sheet that are now grounded below sea level: marine sediments, deposited at a time when the EAIS was much smaller than present, might remain below the ice sheet and be recoverable by drilling.

Overall, the papers by Winnick and Caves and Austermann et al. both represent significant progress in answering the questions of “what happened” and “what could have happened.” But they also show how the best available answers to these questions have evolved as the interpretation of the existing evidence has become more complete. In 2009, it appeared that significant EAIS deglaciation must have happened even though it couldn’t have. Now subsequent research seems to show that although it could have happened, it is not at all clear that it did. This, in turn, highlights the importance of looking for direct geological evidence for Pliocene ice sheet change. It is very possible that there is clear and direct evidence of absence in East Antarctica waiting to be found. We just have to figure out how to find it beneath the present ice sheet.