Main

Understanding the hydrologic cycle during past times of global warmth and reduced ice cover can provide constraints on the fundamental laws governing the climate system. The middle Miocene represents an intriguing period of Earth history during which high temperatures and a substantial retreat of the West Antarctic Ice Sheet and shrinkage of the East Antarctic Ice Sheet1,2, occurred under CO2 levels in the range of 190–850 ppmv (refs 7, 8). At the height of the middle Miocene Climatic Optimum, global annual mean surface temperatures were ∼ 3 °C higher than present, equivalent to projections for the coming century, with 6 °C warming in mid latitudes and 10 °C warming for high latitudes, based on an evaluation of proxy data and model simulations9. As the present warming is already causing melting of Antarctic ice10, the apparent similarities with the middle Miocene make this a climatic period of profound interest. Here we report multiproxy evidence (pollen concentrations, leaf-wax concentrations and leaf-wax isotopes) from the Antarctic geologic drilling (ANDRILL) programme’s AND-2A core in the Ross Sea (77° 45.488′ S, 165° 16.613′ E) recording local warmth and the presence of vegetation around the continental margins (Supplementary Fig. S1). In particular, we use the hydrogen isotopic composition of plant leaf waxes as a tracer of the hydrological cycle during middle Miocene warmth.

It has been suggested that Antarctica was progressively glaciated after 34 Myr ago and vegetation became increasingly restricted to the margins of the continent11. However, recent discoveries from the AND-2A core indicate dynamic conditions in the Antarctic cryosphere between 20 and 10 Myr ago (ref. 12). A sharp increase in Nothofagidites (southern beech) and Podocarpidites (podocarp conifer) pollen was linked to the growth of vegetation on Antarctica (Fig. 1a,b), at the same time as evidence for ice-free oceans and fresh water on land (Fig. 1c,d) during a rapid warming event 15.7 Myr ago (ref. 3). Here, we reanalysed an earlier palynomorph peak at finer resolution and identify the largest vegetation expansion of the past 34 Myr, 16.4 Myr ago (Fig. 1a,b and Supplementary Table S1). The temporal distribution of pollen and spores indicates that local vegetation was tundra with small trees growing in protected areas throughout the middle Miocene. This type of vegetation indicates that the climate was somewhat cooler than that found at the modern austral polar–alpine tree limit (<10 °C in the warmest month), which today is best represented in the southerly Chilean Andes. Peak pollen abundance is composed of species present throughout the middle Miocene in lower quantities indicating that these species had expanded from local refugia. Low species diversity and poor dispersal characteristics of many of these taxa confirm that the vegetation expansion during warm intervals was achieved by recolonization from local sources of cold-tolerant, tundra species and not long-range transport3.

Figure 1: Geochemical and palynological results from AND-2A core.
figure 1

a,b, Pollen relative abundance and c, relative relative abundance of dinocysts produced by dinoflagellate species tolerant of low salinity and d, freshwater algae. Data in a–d from ref. 3 and this study. e, Leaf-wax abundance. f, δD, mean (dark green), one standard deviation (light green). These and additional data are provided in table format in the Supplementary Information. Pink shading highlights two peaks identified in abundance data.

Full size image

Plant leaf waxes identified for the first time in Antarctic margin sediments provide independent evidence for vegetation during the middle Miocene. Peak wax concentrations up to 2.5 μg per gram of dry weight (gdw−1; Fig. 1e) coincide with peak pollen indicating a coherent signal of vegetation optima. Although plant leaf waxes can be transported thousands of kilometres, they are not ubiquitous in these sediments suggesting that long-range transport is unlikely here. Consideration of both archives suggests that refugia of tundra vegetation survived on the margins of Antarctica throughout the middle Miocene, with increases in primary productivity, pollen production and possibly stature (to more tree-like form) during two optima 15.7 and 16.4 Myr ago.

Hydrogen isotope analysis of these plant leaf waxes offers new insights into climate during the middle Miocene. Stable isotopes in precipitation act as a tracer of the hydrological cycle, with the most depleted values on the planet falling over the isolated, high-elevation East Antarctic Ice Sheet13. Isotopic composition is strongly associated with spatial gradients in temperatures in mid and high latitudes14. On Antarctica the δD values of precipitation and snow covary spatially with temperature, elevation and distance inland13. Using observed spatial relationships, temporal variations in temperature have been inferred from Pleistocene ice-core records from the continental interior at Vostok15 and Dome C extending back to 0.8 Myr ago (ref. 16). There is, unfortunately, no earlier glacial stratigraphy surviving on Antarctica, although subterranean ice has been reported from the dry valleys17. Plant leaf wax is therefore our best proxy for the hydrogen isotopic composition of precipitation for deep time applications18. Previous application of this technique to Arctic Palaeocene–Eocene Thermal Maximum (PETM) age deposits has demonstrated striking differences in high northern latitude hydrological conditions during extreme greenhouse conditions 55 Myr ago relative to modern19.

We identify C20 to C30 n-alkanoic acids (hereafter acid) from the marine sediments of the AND-2A core between 144 and 1,100 m below sea floor, equivalent to 20–10 Myr (ref. 20). n-acids with chain lengths longer than C24 with a strong even-over-odd chain length predominance are derived from the waxy coating on terrestrial plant leaves, presumably forming during austral summer (December–February). Leaf waxes are transported to marine sediments by fluvial or aeolian transport. Glacial erosion is also a candidate during the re-advance of the East Antarctic Ice Sheet after 14 Myr (ref. 21). The presence of n-acids with isotopic variability within the homologous series and between closely spaced samples, indicates that leaf waxes are well preserved and that isotopic composition is unaltered (Supplementary Information).

We select the C28-acid for plant-leaf-wax-based palaeoprecipitation reconstructions as it is sufficiently abundant and most probably derived from terrestrial leaf waxes. δD values of the C28-acid range from −165 to −121‰ and average −146±2‰(1 standard error of the mean (s.e.m.), n=37, Fig. 1f and Supplementary Table S2). To calculate palaeoprecipitation, the measured δD values must be adjusted for the net fractionation (ɛacid/precip) between precipitation (precip) and leaf-wax n-acids. Modern calibration studies, typically on the C29n-alkane, indicate ɛalkane/precip−121±2‰ (1 s.e.m., n=316) and must be adjusted for ɛalkane/acid=25‰ (ref. 18). Although smaller fractionations have been reported for some species in greenhouse experiments under continuous light22, high-latitude fractionations23 have not been established as significantly different from the global data set. We therefore assume ɛacid/precip values of −100±2‰ (1 s.e.m.) in Miocene palaeoprecipitation predictions, similar to fractionations applied in Eocene reconstructions (ɛalkane/precip −140‰ (ref. 24) and −130 to −100‰; ref. 19) after correction for compound class. Using these constraints, we calculate that Antarctic coastal palaeoprecipitation averages −51±3‰ (compounded 1 s.e.m., Fig. 2 and Supplementary Information).

Figure 2: T– δD relationship for Antarctica comparing modern versus Miocene.
figure 2

Modern snow across Antarctica (black; ref. 13) and locally (red) with spatial slope. Model results (blue) for coast at 77° S using iGRAM by varying atmospheric opacity (Supplementary Fig. S2). Leaf-wax-based Miocene δDprecip (green), box and whisker plot showing median (line), 25th and 75th percentiles (rectangle) and range (error bars). Temperature conversion of median value, using extrapolated modern slope (black dashed line), model slope (dark blue) and 25th–75th percentile using model slope (shaded area). Microfossil temperature estimates3 and Pleistocene range of δDice values from Vostok15 and Dome C (ref. 16; labelled bars).

Full size image

Modern snow samples across the continent range from −150 to −450‰ depending primarily on temperature and distance inland; with coastal, sea level precipitation at 77° S close to the AND-2A site having an approximate δD value of −150‰ (ref. 13). In contrast, our Miocene δDprecip estimates for coastal Antarctica are dramatically heavier than modern (Fig. 2).

This raises questions as to how such isotopically heavy precipitation can fall at the margins of the Antarctic continent during the middle Miocene and what this may mean for the regional and global climate system. Clues originate from microfossil analyses in the same sediments that indicate ice-free conditions in the Ross Sea in two brief pulses. Local optima 16.4 and 15.7 Myr ago (Fig. 3d) correspond to pulses of global warmth each <30,000 yr in duration identified in deep-sea archives within the generally warm middle Miocene4 (Fig. 3a). In the Ross Sea, temperate dinoflagellate species replace sea-ice indicator species and indicate sea surface temperatures in the range of 0–11.5 °C (Fig. 1c; ref. 3). Deep-ocean temperature reconstructions also show increased temperatures and reduced ice volumes throughout the middle Miocene1 (Fig. 3) and model experiments suggest that global warmth was intensified at high latitudes9.

Figure 3: Miocene climate records.
figure 3

a, Benthic stacked δ18O, five-point smooth4 (black); labelled warm events (Middle Miocene Climatic Optimum, MMCO, peak 1 and 2), and glacial events at marine isotope (Mi) stages 1–3 (ref. 1); Southern Ocean site 747 (ref. 4; blue). b, Southern Ocean site 747 Mg/Ca (ref. 1; blue). c, Benthic stacked δ13C with five-point smooth4 (grey); δ13C maxima (CM) 1–5 linked to productivity increases5; Southern Ocean site 747 (ref. 4; blue). d, Leaf-wax abundance. e, δDshown downcore (green) and δD summarized in box and whisker plot showing median (line), 75th and 25th percentiles (rectangle) and range (error bars). Pink shading as in Fig. 1; grey shading disrupted sedimentation.

Full size image

Our δDprecip estimate of ∼−50‰ supports the multiproxy evidence for warmer temperatures at the Antarctic coast during the middle Miocene. However, other factors beyond temperature are known to influence water isotopes13,14 and the temperature versus δD slope (T– δD) in the Miocene could be different from today. Pairing proxy isotope data with isotope-enabled modelling is one effective way to identify competing dynamical influences on proxy isotopes25. As there are no isotope-enabled general circulation models with realistic boundary conditions for the middle Miocene at present, we tested our results with an idealized grey-radiation aquaplanet moist general circulation model26,27 with the addition of isotopic tracers (iGRAM). Aquaplanet experiments provide reasonable constraints for the coastal site because precipitation isotopes at the margin of a continent are similar to the adjacent ocean6. Furthermore, iGRAM results match annual mean temperatures and precipitation isotopes by latitude compared with more complex realizations in the National Center for Atmospheric Research Community Atmosphere Model version 2 (ref. 6; Supplementary Fig. S2). Experiments with iGRAM include warm, ice-free conditions that provide physical and dynamical support for heavy polar precipitation isotopes (Supplementary Fig. S3). The model response of −51.4‰ for +7.7 °C (at 77° S) is quantitatively consistent with the leaf-wax result (−51±3‰). Critically, these model experiments allow us to attribute our proxy-based observations to not only a temperature increase, but also to a less D-depleted atmospheric vapour source sustained by evaporation over the adjacent warmer oceans (Supplementary Table S3). The model–proxy combination allows us to infer that Miocene isotopes in precipitation reflect not just warmer temperatures, but also important changes in hydroclimate over the warmer Southern Ocean.

Both the modern Antarctic spatial T– δD relationship and a shallower slope based on our model experiments predict temperatures >0 °C (Fig. 2). δDprecip reconstructions are therefore consistent with pollen assemblage estimates of summer land surface temperatures of 0–10 °C (ref. 3). Our best estimate of summer (December–February) temperatures is ∼7 °C over Miocene vegetation versus −4 °C over the modern glaciated surface, although it remains an open question whether the warmth1,4 was sufficiently persistent throughout the middle Miocene to explain the long duration of increased δDprecip values. Temperature inferences have uncertainties associated with topography, land cover and seasonality, not considered here; however, our results demonstrate that increased δDprecip over Antarctica requires a high-latitude moisture source and warmer temperatures (Supplementary Table S3).

With the dynamical nature of the water isotope tracer in mind, we note that our leaf-wax-based δDprecip reconstructions are similar to or higher than Arctic values reported during extreme greenhouse conditions19. The contrast in flora is dramatic between the tundra vegetation of the Miocene Antarctic and the tropical-type flora of the PETM Arctic28; clearly temperatures alone cannot explain matching δDprecip values. Instead, dynamics of water-vapour transport and rainout must explain the difference. For the Arctic, surrounded by land, surface ocean water became fresher during the PETM with lower δD (ref. 19) and convective rainfall may have generated D-depleted values at the extreme greenhouse temperatures of the PETM (ref. 28). The similarity of δD values belies very different temperatures and hydrological processes.

For the middle Miocene, we find leaf-wax evidence for vegetation refugia on Antarctica with brief pulses of high pollen production. δDwax values form the basis of palaeoprecipitation estimates that indicate warmer temperatures. D-enrichment also implies that the water-vapour source is close to the precipitation site6 and model experiments demonstrate that a warmer ocean would have supplied moisture to the atmosphere locally (42% sourced polewards of 75° S at 7.7 °C), and increased rainfall on Antarctica. Relative to modern polar-desert conditions (at McMurdo station average temperature −16.9 °C, precipitation 203 mm yr−1) our findings suggest an intensified hydrological cycle consistent with evidence for tundra in the dry valleys of Antarctica29. These exceptional reversals of past icehouse conditions may offer valuable insights into hydrological processes under future warming scenarios.

Methods

Data.

Accompanying data can be found at: ftp://ftp.ncdc.noaa.gov/pub/data/paleo/contributions_by_author/feakins2012/feakins2012.txt, ftp://ftp.ncdc.noaa.gov/pub/data/paleo/contributions_by_author/feakins2012/feakins2012.xls.

Leaf-wax δD.

Sediment samples (20 g) were processed to extract and purify leaf-wax n-alkanoic acids as described in detail elsewhere30. δD values were analysed using a Thermo Scientific Trace gas chromatograph (ZB-5 ms column, 30 m×0.25 mm×1 μm) with a programmable temperature vapourization injector operated in solvent-split mode connected through an Isolink pyrolysis furnace (at 1,420 °C) to a Delta V Plus mass spectrometer. Peaks of hydrogen reference gas bracketed analytes, two of which were used for standardization of the isotopic analyses. Data were then normalized to the standard mean ocean water/standard light Antarctic precipitation isotopic scale by comparison with an external standard containing eight fatty acid methyl esters (i-C14 to i-C20; obtained from A. Schimmelmann, Indiana University, Bloomington) with δD values ranging from −167 to −231‰. The accuracy of replicate analyses of the external standard was 4.4‰ (root mean square error, n=32). The H3 factor was determined daily averaging 6.056±0.125 ppm mV−1. The results are reported using conventional delta notation (δD‰).

Palynology.

As in ref. 3.

Climate model experiment.

See Supplementary Information.