Exposure-age data from across Antarctica reveal mid-Miocene establishment of polar desert climate

High-elevation rock surfaces in Antarctica have some of the oldest cosmogenic-nuclide exposure ages on Earth, dating back to the Miocene. A compilation of all available 3 He, 10 Be, and 21 Ne exposure-age data from the Antarctic continent shows that exposure histories recorded by these surfaces extend back to, but not before, the mid-Miocene cooling at 14–15 Ma. At high elevation, this cooling entailed a transition between a climate in which liquid water and biota were present and could contribute to surface weathering and erosion, and a polar desert climate in which virtually all weathering and erosion processes had been shut off. This climate appears to have continued uninterrupted between the mid-Miocene and the present.


INTRODUCTION
From the perspective of Earth's geomorphology, Antarctica is notable because in highelevation, inland regions, liquid water is almost entirely absent, which leads to a near-total lack of biota and an almost complete standstill of most of the erosional and weathering processes that are active elsewhere on Earth. Marine sedimentary records suggest that this polar desert climate was established during the mid-Miocene cooling ca. 14-15 Ma (e.g., Zachos et al., 2001;Shevenell et al., 2004), and this hypothesis is supported by several lines of evidence from Antarctica, which we summarize below.
Fossil organisms and pollen in terrestrial and marine sediments in Antarctica show that a tundra environment, with mean summer temperatures up to ∼5-10 °C, existed during the mid-Miocene and appears to have gone extinct at ca. 14 Ma, replaced by polar conditions that have persisted to the present (Lewis et al., 2008;Ashworth and Cantrill, 2004;Ashworth and Erwin, 2016;Wei et al., 2014;Warny et al., 2009;Anderson et al., 2011). Inland volcanoes in Marie Byrd Land (Fig. 1) show alpine glacial morphology only on pre-mid-Miocene volcanic edifices (Rocchi et al., 2006), and stratigraphic evidence for a shift from wet-to frozen-based glaciation in the Transantarctic Mountains is dated to ca. 14 Ma (Lewis et al., 2008). The evidence from the Transantarctic Mountains implies a cooling of ∼8 °C, which is similar to the 6-7 °C cooling estimated for surface waters in the southwestern Pacific Ocean at this time (Shevenell et al., 2004).
Cosmogenic-nuclide exposure dating in Antarctica has established that many rock surfaces currently exposed above the ice sheet have experienced extremely low erosion rates, and therefore that these surfaces record exposure histories that extend millions of years into the past (e.g., Nishiizumi et al., 1991;Van der Wateren et al., 1999;Margerison et al., 2005). Here we exploit a compilation of all available exposure-age data from bedrock and boulder surfaces from Antarctica to show that the exposure history recorded by Antarctic rock surfaces extends back to, but not past, the mid-Miocene cooling at ca. 14-15 Ma. We conclude that this is the result of a transition between a climate in which liquid water and biota were present and could facilitate surface weathering and erosion, and one in which they were absent and could not, and that this condition has been nearly or fully uninterrupted between the mid-Miocene and the present.

CALCULATION OF EXPOSURE AGES
We computed apparent exposure ages from measurements of the stable or long-lived cosmogenic nuclides 3 He on pyroxene or olivine, 10 Be on quartz, and 21 Ne on quartz that are compiled in the ICE-D:ANTARCTICA database (http://antarctica.ice-d.org; Balco, 2020). Apparent exposure ages assume a single, continuous period of exposure during which zero erosion occurs. Hence, for surfaces that have been eroded and/or covered by ice, till, or other material for a portion of their history, apparent exposure ages underestimate true cumulative exposure durations, and thus they are strictly lower limits.
The ICE-D:ANTARCTICA database documents ∼3700 samples that were collected for exposure-dating purposes, ∼2600 of which have measurements of 3 He, 10 Be, and/or 21 Ne. Of these, 84% are described in at least one of 120 journal articles published between 1986 and 2020 (source references are tabulated at http:// antarctica.ice-d.org). The remaining 16% of the samples were nearly all contributed to ICE-D:ANTARCTICA or other public data repositories to satisfy the data dissemination requirements of funding agencies. Samples span the Antarctic continent ( Fig. 1), but their distribution is neither random nor systematic, instead reflecting a range of research objectives (from long-term landscape evolution to recent icesheet change), the accessibility of field sites, and the interests of individual researchers. Nevertheless, most regions in Antarctica with rock outcrops are represented (Fig. 1).
Production rates and exposure ages for 3 He, 10 Be, and 21 Ne were computed using the production-rate scaling method of Lifton et al. (2014), as implemented in version 3 of the online exposure age calculator described by Balco et al. (2008) and subsequently updated. We used the relationship between Antarctic air pressure and elevation of Stone (2000). Production rates are based on the global calibration data set of Borchers et al. (2016) and 21 Ne/ 10 Be production-rate ratios centered on 4.03 (Balco et al., 2019). sites in Antarctica ( Fig. 2; see the Supplemental Material 1 ). While 3 He and 21 Ne are stable nuclides, 10 Be is radioactive with a 1.4 m.y. half-life. After ∼7 m.y. (5 half-lives) of continuous exposure and zero erosion, 10 Be concentrations asymptotically approach a "saturated" steady state such that nuclide production is balanced by radioactive decay. Thus, samples with apparent 10 Be exposure ages of ∼7-8 m.y. may, in fact, have experienced much longer exposure histories.
The oldest samples in our compilation were collected from high-elevation sites in the Transantarctic Mountains or from isolated nunataks in East and West Antarctica (Figs. 1 and 3C). These sites experience a climate that is perennially extremely cold and arid and thus are nearly completely devoid of liquid water and biota (Graly et al., 2018;Dragone and Fierer, 2019). The high-elevation regions where these sites are located are also where glacier and icesheet velocities are lowest, ice (where present) is frozen to the bed, and ice thickening during 1 Supplemental Material. Table S1 (sample and exposure-age information for all samples in the ICE-D:ANTARCTICA database that have Miocene apparent exposure ages). Please visit https://doi .org/10.1130/GEOL.S.12869681 to access the supplemental material, and contact editing@geosociety. org with any questions.  past glacial periods was limited (Spector et al., 2019). As a result of these meteorological and glaciological factors, nearly all weathering, erosion, and sediment transport processes, both subaerial and subglacial, do not operate in these regions at present, and have likely been minimal throughout the Quaternary. In contrast, samples located below ∼1000 m elevation have young exposure ages, ranging from Holocene on deposits emplaced during the most recent deglaciation to early and middle Pleistocene on various older deposits and bedrock surfaces (Fig. 2). Due to the parabolic profile of the ice sheet, low-elevation ice-free areas in Antarctica are found only near the coastal perimeter. Low elevation and proximity to the ocean imply the highest temperature and precipitation in the continent, so rock surface weathering rates are relatively high (Fig. 3A). Low-elevation ice-marginal sites are also both adjacent to fast-flowing outlet glaciers and subject to the largest glacial-interglacial changes in ice thickness, factors that increase the likelihood of subglacial erosion by wet-based ice during glacial periods (e.g., Sugden et al., 2005; Fig. 3B). For example, two of the warmest and wettest regions in Antarctica are the Antarctic Peninsula and Marie Byrd Land (Fig. 1). These regions are represented by ∼10% of the samples in our analysis, but none of these samples have an exposure age exceeding 1 m.y.
Apparent exposure ages alone show that the oldest sample in our compilation has been exposed for at least 13.5 m.y., and that several samples have been exposed for at least 11 m.y. We hypothesize that preservation and exposure of these high-elevation rock surfaces at nearzero erosion rates began during the mid-Miocene cooling at 14-15 Ma. Prior to that time, the Antarctic climate was warmer (Lewis and Ashworth, 2016), which, by analogy to Arctic mountain ranges where water and biota are present (see discussion below), resulted in much higher weathering and erosion rates.

EFFECT OF SLOW SURFACE EROSION ON EXPOSURE AGES
Erosion reduces apparent exposure ages by removing the outer layer of rock, which has received the highest cosmic-ray dose (Lal, 1988). Our objective here is to determine whether the oldest samples from Antarctica are consistent with continuous exposure for the past 14-15 m.y. at very low erosion rates, or whether much older exposure durations are permitted, which would falsify our hypothesis. To quantify this, we rely on samples on which both 10 Be and 21 Ne have been measured. Measurements of two nuclides with different decay rates can constrain two unknown values related to the geomorphic history of a sample, and in this case, we solve simultaneously for the exposure age and the steady erosion rate (Lal, 1991 ; Fig. 4).
The paired-nuclide data (Fig. 4) show that although the oldest samples have apparent 21 Ne exposure ages (assuming zero erosion) up to 13.5 m.y., three samples are consistent Although unrecognized past ice cover of these samples would cause us to slightly overestimate the true erosion rates and exposure ages, these samples lie near the simple-exposure line and thus require near-continuous exposure. Our hypothesis predicts that no sample should have an exposure age >14-15 m.y., and this is, in fact, what we observed. All old samples that lie within the simple-exposure region are consistent with exposure ages indistinguishable from, or postdating, the mid-Miocene climate transition at 14-15 Ma. In addition to surface erosion, surface uplift can cause the apparent exposure age of a rock surface to underestimate the true age. Peaks in some areas of Antarctica are hypothesized to experience uplift on million-year time scales due to isostatic compensation of bedrock erosion beneath nearby outlet glaciers (e.g., Stern et al., 2005;Sugden et al., 2014) or time-varying dynamic topography (Austermann et al., 2015). From the perspective of a rock surface, uplift and erosion are the same: both cause the production rate to increase over time by removing overlying mass-atmosphere for uplift, rock for erosion-that previously shielded the surface from the cosmic-ray flux. An uplift constraint can therefore be substituted for an erosion constraint (Balter et al., 2020). If we assume a limiting case of zero erosion for the oldest surfaces, their nuclide concentrations are consistent with uplift rates of ∼35-40 m m.y. −1 .
Although Figure 4 is drawn under the assumption that surfaces are experiencing erosion and not uplift, the uplift-erosion equivalency means that this assumption does not affect the exposure-age constraints inferred from the pairednuclide data.

EROSION RATES REMAINED LOW DURING WARM CLIMATES
Marine sediment records and climate-model experiments suggest that portions of Antarctica were 5-10 °C warmer during the late Miocene and Pliocene than during the pre-industrial period (Zachos et al., 2001;DeConto and Pollard, 2016;Herbert et al., 2016). Liquid water and biota may have been more abundant at these times, so rock weathering and erosion rates may have been elevated. The closest analogue to the late Miocene and Pliocene climates of Antarctica may be the Pleistocene Arctic, where the highest measured cosmogenic-nuclide concentrations in rock surfaces record no more than ~1 m.y. of exposure and imply erosion rates >50 cm m.y. −1 (Margreth et al., 2016;Strunk et al., 2017). Although the oldest Antarctic surfaces indicate much lower erosion rates (Fig. 4), some are consistent with a scenario in which erosion rates were somewhat higher during past warm periods.
To quantify this, we calculated the maximum duration that Antarctic rock surfaces could have eroded at 50 cm m.y. −1 -the lowest rate implied by Arctic rock surfaces-during the late Miocene and Pliocene. The most restrictive constraints come from samples that are at or near saturation with respect to 10 Be. An upper limit on the duration of accelerated erosion is obtained by assuming a scenario in which 10 Be concentrations were initially at saturation, decreased during a period in which rock surfaces eroded at 50 cm m.y. −1 , and subsequently increased to present-day values under zero erosion. If we assume that erosion occurred at ca. 3 Ma during the Pliocene, then samples with present-day 10 Be concentrations near saturation permit erosion for no more than ∼80 k.y. Longer erosion durations are permissible if they occurred earlier during the late Miocene, for example, ∼250 k.y. of erosion at ca. 5 Ma.
Even with the conservative assumption of initial 10 Be saturation, these results indicate that Antarctic erosion rates during warm climates of the late Miocene and Pliocene rarely, if ever, approached even the lowest Pleistocene erosion rates implied by Arctic rock surfaces. Therefore, we surmise that at high-elevation sites in Antarctica, polar desert conditions have persisted with zero or minimal interruption since the mid-Miocene cooling at 14-15 Ma.

SUMMARY
The oldest rock surfaces in Antarctica have been continuously exposed for 14-15 m.y. at extraordinarily slow erosion rates of 2 cm m.y. −1 or less. Of the ∼2600 samples in our compilation, none indicate exposure for more than 14-15 m.y. Samples that are at or near saturation with respect to 10 Be require that erosion rates remained very low for the vast majority of the late Miocene  261). Asterisks indicate that nuclide concentrations are normalized to the surface production rates for each sample. The abscissa is equivalent to the apparent 21 Ne exposure age, and thus is a lower bound on true cumulative exposure. Ellipses represent 68% confidence regions. Nuclide concentrations of continuously exposed, non-eroding samples would lie along the upper blue line ("simpleexposure line"). Nuclide concentrations of samples that have reached production-decay-erosion equilibrium would plot along the lower black line ("steady-erosion line"). These lines bound the "simple-exposure region" . Main panel expands the area in the rectangle in the upper-right panel. The simple-exposure region in the main panel is contoured with dashed blue isolines of erosion rate (in cm m.y. −1 ) and solid black isolines of exposure age (in m.y.). Red band represents the 14-15 m.y. exposure-age contour. All Miocene-age samples that lie in the simple-exposure region plot along or above the 14-15 m.y. contour band (i.e., they are younger than ∼15 m.y.) and are thus consistent with the hypothesis that no rock surface in Antarctica has an exposure age that pre-dates the mid-Miocene climate transition. Miocene-age samples that lie below the simple-exposure region reflect past ice-sheet cover and/or non-steady erosion. and Pliocene, periods when it is presumed that Antarctica experienced a warmer-than-present climate. Hence all exposure-age data from highelevation, inland sites support the hypothesis that the mid-Miocene cooling at 14-15 Ma resulted in a polar desert climate that shut down virtually all weathering and erosion processes, and that this climate appears to have persisted with zero or minimal interruption to the present.