The Mid-Pleistocene Transition and the Vostok Oldest Ice Challenge

Marine records indicate a dramatic change in the predominant periodicity of climate variability, from about 40 ka to about 100 ka around one million years ago. The reason for this major climatic shift, which is called the Mid-Pleistocene Transition or MPT, remains unknown – and is of great interest to the climate scientist. Could the core of the oldest meteoric ice bedded at Vostok between 3310 and 3539 m, which has experienced severe deformation, nevertheless be useful in deciphering some of the aspects of the MPT enigma? Reflecting upon this question and considering the available data from the disturbed section of the ice core, we feel impelled to propose a new project focused on the oldest Vostok meteoric ice, which could be named the Vostok Oldest Ice Challenge or VOICE.


Introduction
Over the last few decades, the deep ice cores drilled at Russia's Vostok Station have provided a wealth of information about past climate and envi ronmental changes . At this East Antarctica site, the ice thickness amounts to 3770 m and the snow ac cumulation rate is only 2 .1 cm of water equivalent per year . This gives us the unique opportunity to ob tain a long climatic record with a relatively high time resolution . In 1998, drilling operations in the deep 5G borehole, conducted at that time as a collabora tive project between Russia, France and the United States, reached a depth of 3623 m . At 3539 m below the surface, the drill penetrated into the ice refrozen from water from Lake Vostok, a deep subglacial water body which extends below the ice sheet over a large area . After a long break, the drilling was resumed in January 2006 . Finally, in February 2012, the drill reached the surface of Lake Vostok for the first time, at a depth of 3769 m .
The whole 5G ice core can be separated into three distinct sections (Fig . 1) . The upper 3310 m of the core are characterised by an undisturbed sequence of meteoric ice layers . Analysis of this section of the core has resulted in the first Antarctic ice record of atmo spheric composition and climate extending through four climate cycles back to 420 kyr BP [31] . Between 3310 and 3539 m there are indications of iceflow anomalies that could alter the original stratigraphy of meteoric ice . Finally, between 3539 m and the ice water interface at a depth of 3769 m, the core consists of ice accreted at the bottom of the Antarctic ice sheet from Lake Vostok's water .
The Vostok ice cores which represent the upper section of the meteoric ice (0-3310 m) and the 230 m thickness of accreted ice (3539-3769 m) have been comprehensively analysed and have received extensive coverage in the scientific literature, in cluding a number of top, widely cited papers related to past climate change and exploration of Antarc tic subglacial environments . Meanwhile, the oldest meteoric ice, bedded between 3310 and 3539 m, has seen much less attention . Earlier works have shown that complex ice deformation, which has occurred when the ice was still grounded upstream from Vostok Station, resulted in the folding and intermix ing of ice at a submetric scale in the stratum bedded below 3450-3460 m and at a larger scale between 3310 and 3450 m (see Fig . 1) [18,[39][40][41] . Using an appropriate correction of the ice stratigraphy for flow disturbance in the 3320-3345 m interval, it has only been possible to extend the Vostok ice record further back to 440 kyr BP, which implies full cov erage of the Marine Isotope Stage (MIS) 11 [36] . Because of the ice mixing and the diffusive smooth ing of the climatic signals, extracting useful paleo climatic information from the deeper section of the Vostok disturbed ice is a challenging problem . The aims of our paper are: (a) to explain our motivation to tackle this difficult task, (b) to discuss the old and new data relevant to this issue, and (c) to present a new project focused on the oldest Vostok meteoric ice, referred to from now on as the Vostok Oldest Ice Challenge or VOICE .

The enigma of the Mid-Pleistocene Transition
During the Pleistocene, roughly the last 2 .5 mil lion years, the Earth's surface has known a series of glacialinterglacial cycles stimulated by changes in the distribution of the solar radiation reaching the Earth's surface, themselves driven by the periodici ties of the Earth motion around the sun (about 100 and 400 ka) and of changes in the tilt (41 ka) and di rection (23 and 19 ka) of Earth's rotation axis . These oscillations are well imprinted in the long deepsea sediment records [20] but surprisingly, they indicate a systematic change in the predominant periodic ity from about 40 to about 100 ka around one mil lion years ago, although there is no marked change in the insolation signal driven by the Earth's orbital pa rameters (Fig . 2) . This is called the MidPleistocene Transition and is a major, if not the key Pleistocene enigma yet to be solved .
Conceptually, it is possible to shift from a 'rapid' to a slower oscillation by adding a long trend . This is what Paillard [27] and Paillard and Parrenin [29] showed by using a simple conceptual model with multiple equilibria (threshold model) in the climatic I -undisturbed section of the Vostok meteoric ice containing continuous climatic record (0-3310 m); II -disturbed section of the meteoric ice (3310-3539 m), in which two distinct strata can be discerned: in the stratum between 3310 and 3460 m a largescale folding of ice is presumed, in the stratum between 3460 and 3538 m the submetric scale ice interbedding is observed; III -accreted ice refrozen from Lake Vostok's water . The deuterium profile is composed from available published data [1,31,40] (axis is not shown) Рис. 1. Схематическое изображение вертикального строения антарктического ледника в районе станции Восток .
Most hypotheses for the origin of the MPT in voke a response to a longterm cooling possi bly induced by decreasing atmospheric CO 2 con centrations (see [6] and references herein) . Raymo et al . [33] argued that preMPT predominance of the obliquity frequency is due to the cancellation effect of the integrated precession signal associated with changes in ice sheet volume, which is out of phase between the northern and southern hemispheres . They speculated that during MPT, high latitudes be came cool enough to enable the growing Antarctic ice sheet to cover the terrestrial margin no longer vul nerable to ice melting . The consequence would have been that the Antarctic ice volume was controlled by sea level changes essentially driven by Northern Hemisphere ice sheet fluctuations, leading to inphase changes of northern and southern ice sheets . This would have strengthened the precession signal of the global δ 18 O marine record, and so induced a change in the observed predominant periodicity . The pro posal that at the MPT, marinebased ice sheet mar gins replaced terrestrial margins around Antarctica, has been borne out by drill records from the margins of Antarctica on the edge of the Ross Sea [25,26] and by icesheetice shelf model simulation of the dynamics of West Antarctica [32] . Further evidence for the major role of Antarctic ice volume during the MPT arises from a deepocean sediment core (ODP 1123) taken east from New Zealand . By separat ing the effects of ocean temperature and ice volume on the benthic oxygen isotopic record, Elderfield et al . [8] suggest that an abrupt increase in Antarc tic ice volume initiated the MPT around 900 ka ago .

MPT and atmospheric CO 2
Assuming that the Pliocene cooling played a major role in the establishment of the MPT, we have yet to assess the potential contribution of atmospher ic CO 2 to this cooling trend and to the MPT shift be tween obliquity and eccentricity as the dominant or bital signal in the paleorecord .
Ice core record. The most direct and reliable ar chives of longterm atmospheric CO 2 background level during the past are enclosed in Antarctic ice [34,35] . The Vostok [31] and EPICA DC [21] ice cores offer the best CO 2 record covering the last 800 ka . The record is, on the whole, remarkably cor related with the climate record imprinted in the same cores under the signature of the ice isotopic D/H re cord [13,31] and exhibits a dominant ~100 ka peri odicity in the spectrum of the Earth motion driven by its orbital and axial oscillations . Unfortunately no other long vertical ice core offers, up to date, a con tinuous chronological sequence reaching 1 Ma or more, and this is one of the major challenges in ice core science for the future [10] .

Fig. 2. Changes in insolation and benthic δ 18 Ο over the past two million years:
a -the LR04 stack of benthic δ 18 Ο records [20] over the last 2 Ma; b -time series of June 21 insolation at 65° N [14] filtered using a 100ka Gaussian filter with a bandwidth of 10 ka; bthe LR04 stack filtered using a 100ka Gaussian filter with a bandwidth of 10 ka . Calculations are performed with the Analyseries software [28] Рис. 2. Изменение инсоляции и изотопного (δ 18 Ο) со става бентосных фораминифер за последние 2 млн лет: a -сводный ряд LR04 данных по δ 18 Ο бентосных фора минифер [20] за последние 2 млн лет; b -ряд инсоляции 21 июня, рассчитанный для параллели 65° с .ш . [14] и сглаженный фильтром Гаусса 100 тыс . лет с полосой про пускания 10 тыс . лет; c -сводный ряд LR04, сглаженный фильтром Гаусса 100 тыс . лет с полосой пропускания 10 тыс . лет . Расчёты выполнялись с помощью программы Analyseries [28] Палеогляциология  98  There is another way to recover very ancient ice in Antarctica: by sampling a zone of blue ice at or near the surface . This ice, which has travelled a long way in depth, now appears at the surface in regions where no snow accumulates because of the winds and where ablation may occur due to radiative sub limation . Of course, the stratigraphy of this ice may have been disturbed because of its journey in depth and close to the bedrock, but it may be very old . Re cently, Higgins et al . [11] reported on atmospher ic composition, including CO 2 , 1 million years ago from blue ice in the Allan Hills in Antarctica . The estimated 1 Ma ice, dated by the 40 Ar method with an uncertainty of about ±200 ka, was found in a stratigraphically disturbed section at the base of a 126 m ice core . The CO 2 concentrations measured on the 1 Ma ice are in the range 221-277 ppm . They have to be compared with the 280 ppm preindustri al value and the glacialinterglacial oscillations over the last 800 ka, whose amplitudes are between about 170 and 300 ppm [21] .
Marine record. In 2009, motivated by the lack of an ice core record of atmospheric CO 2 covering the MPT period, Hönisch et al . [12] proposed using the boron isotopic composition in planktonic fora minifer, which is a proxy for past sea water pH, to estimate atmospheric partial pressure of CO 2 across the MPT . Based on this study, which presents a re cord back to 2 Ma, the preMPT atmospheric par tial pressure of CO 2 during interglacials were simi lar as during the recent postMPT (the last 500 ka), whereas atmospheric CO 2 during the preMPT gla cial periods was higher than during postMPT .
More recently [24], the boronisotope re cord has been extended to the late Pliocene (3 .3 to 2 .3 Ma ago), a period, which is supposed to include the warmest intervals of the Pliocene between 3 .3 and 3 Ma ago . This new record indicates atmo spheric CO 2 concentrations on the whole between 400 and 300 ppm .
In summary. The first attempt to get atmospher ic CO 2 levels around 1 Ma ago from Antarctic blue ice indicates values between 221 and 277 ppm, i .e . below the preindustrial level . This preliminary work reinforces the hope to get an accurate record from ice across the MPT in future . On the other hand, the marine Boronisotope record has the potential to extend the record back to the Miocene . Combining the 2 records should help to assess the role of atmo spheric CO 2 in driving the long trend cooling during the Pleistocene and the MPT shift in the dominant orbital signal, which is observed in the paleorecord .

The bottom section of the Vostok meteoric ice
The next question to be discussed in this section is: сould the oldest meteoric ice bedded at Vostok below 3345 m, which has experienced severe defor mation, folding, intermixing and diffusive smooth ing, nevertheless still be useful in deciphering some of the aspects of the MPT enigma? The precondi tion for a positive answer is that the disturbed ice should be much older than the undisturbed section of the Vostok ice core (420 ka) and older than the oldest continuous EPICA record obtained so far, which spans back to 800 ka [9] . In that case, despite the stratigraphic discontinuity, studies of such ice, in particular the measurements of the air content and the concentration of CO 2 in trapped air, can yield new important information on the potential role of the Antarctic ice sheet instability and the atmospher ic carbon dioxide in the genesis of MPT . Below we present and discuss some old and new data on the ice texture and fabric, the air content of ice and the geometrical properties of the airhydrate crystals ob tained from the Vostok ice core below 3310 m, which in our view should stimulate further comprehensive investigations of the 228 m thick stratum of disturbed meteoric ice at Vostok .
Ice texture and fabric. In the Vostok core of me teoric ice, two different types of microstructure, re ferred to as A and B, distinguish interglacial ice from glacial ice [2,18] . In the A layers, which have low impurity content of ice and are associated with inter glacial conditions, the ice grains are larger and their caxes exhibit verticalgirdletype orientation . Stress field leading to such fabric is characterised by uniaxi al tension in the direction of the ice flow . As a result, caxes rotate away from the tensile axis, which makes the ice harder to deform .
On the contrary, in the B layers, which have been formed under conditions of glacial maxima and are therefore characterised by high impurity content, the finegrained ice with vertical clustering of caxes is observed . This kind of fabric pattern corresponds to vertical compression in the upper part of ice sheet, or to simple shear in its bottom part .
Below 3460 m, a scale of the A and B layers in terbedding is reduced to the 10 −2 -10 −1 m . At the same time, the amplitude of the grain size vari ations increases by an order of magnitude, mostly due to the abnormal grain growth at annealing tem peratures, uninhibited by insoluble impurities in the A layers . The ice fabric pattern remains the same here as in the upper section of the disturbed ice, that is the girdletype caxis orientation is observed in the A layers with low concentration of atmospheric dust, and the singlemaximum fabric in the B layers with high dust concentration (see Fig . 5 in [39]) . How ever, the diffusive smoothing superimposed on the ice mixing at submetric scale would normally have drastically altered the isotopic signal . Consequent ly, highresolution isotope measurements, which have been performed continuously along the select ed depth intervals, do not reveal coherent variations with the dust and textural properties of A and B lay ers (A . Ekaykin, unpublished data) . Importantly, the airhydrate crystals exhibit uninterrupted gradual growth with depth (age) of ice in the bottom section of the Vostok core (see below) . This has been regard ed as a proof that only centimeter scale intermixing of ice, if any, could have taken place here [18] .
At the very bottom of meteoric ice, between 3522 and 3539 m, the amplitude of the grain size varia tions significantly decreases (see Fig . 3, c), appar ently due to the sudden disappearance of the B type layers . This could result from the uniformly low con centration of dust below 3522 m, although more dust measurements are needed to prove this assertion (see available data in Fig . 3, b) .
Folding and the ice flow disturbance in the bot tom ~250 m of glacier ice are well known phenom ena . It has been shown that the difference in the mechanical properties of layers with distinct micro structure, as those described above as layers A and B, may cause microfolding at a scale of a few cen timeters [7] . Assuming a longitudinal compression of ice would lead to a concept of tectonic thicken ing and progressive buildup of the deformed basal ice upwards with more advanced deformation close to the bed [41] . Such stress conditions indeed may occur in the vicinity of Vostok Station, at the south ern end of Lake Vostok . Here the ice sheet is sub stantially grounded, so that the lake edge, which is located less than 5 km downstream from the bore hole, can hinder the basal ice movement and favour ice blocking [18] . The existence of a «young» (pres entday) shear zone beneath Vostok appears to be the most plausible explanation for the highly deformed ice observed in the deepest section of the disturbed ice (3460-3539 m) . Indeed, if such a zone was gen erated upstream from Lake Vostok, all textural and fabric distinction between layers A and B would have been eliminated by the ice recrystallisation under annealing temperatures (4 .5-6° below the pressure melting point) during the long (~40 kyr [38]) journey of this ice over subglacial Lake Vostok, from the up stream grounding line to the drilling site .
On the contrary, the large scale folding, which is presumed to be characteristic of the upper, cool er section of the disturbed ice, between 3310 and 3460 m (though up to now it has only been con firmed on one occasion, in the depth interval of 3320-3345 m), is thought to have occurred when the ice was grounded upstream from Vostok, due to the interaction of the basal ice with the bedrock un dulations . Earlier studies have shown that the freez ing of lake water on the sole of the ice sheet in the area of Vostok Station occurred as soon as the ice crossed the grounding line [38,41] . This implies that the basal ice did not experience loss by melting after contact with the lake water, and that very old, though badly deformed, Antarctic ice may exist here .
Air content of ice. We have measured the air con tent of ice (V) along the disturbed section of the Vostok ice core using a barometric method imple mented with an experimental setup called STAN [17] . The absolute precision of the STAN measurements is estimated to be within ±0 .6% for typical for polar ice level of air content of an order of 0 . 1 cm 3 g −1 (here after the gas volume is given at standard conditions: T = 273 . 15 K and P = 0 . 1013 MPa) . The new data presented in Fig . 3, d extends the previously obtained Vostok V records [19,22] to the boundary between the meteoric ice and accreted ice at 3539 m below the surface . The whole combined V record (not shown) demonstrates a weak tendency of air content to de crease with increasing depth . The mean values of V were found to be 0 .0892±0 .0037 cm 3 g −1 (±10σ) be tween 114 and 3310 m (i .e . during the last ~420 ka), 0 .0884±0 .0021 cm 3 g −1 in the depth interval of 3310-3500 m, and 0 .0881±0 .0022 cm 3 g −1 below 3500 m .
The air content of polar ice averaged over the time spans covering several climatic cycles should be proportional to the mean atmospheric pressure at the site of the ice formation (see e .g . [23]) . For instance, the observed decrease of the V in the deeper sections of the Vostok ice core reflects advection of ice from the sites with higher elevation, located upstream from Vostok . Aside from this trend, the new measurements show that the mean air content of the oldest meteoric ice is almost the same as that measured in the upper section of the Vostok core (114-3310 m) [19,22], and in other words, typical for polar ice formed at the conditions (atmospheric pressure, temperature) prevailing at Vostok Station [23] . This implies that during a long time period covering the formation of the presently 3539 m thick stratum of meteoric ice at Vostok, the icesheet surface elevation in this part of Antarctica has been essentially stable and similar to that during the last 420 ka .
Growth of air-hydrate crystals and the maximum age of meteoric ice at Vostok. Dating the disturbed ice is the key challenge . Provided the basal ice flow has not been disturbed, extrapolation of the existing gla ciological timescales below the end of the continuous climatic record at Vostok can give a conditional esti mate of the agedepth relationship for meteoric ice bedded below a depth of 3310 m . With different as sumptions about temperature and accumulation rate prior to 420 ka BP, the boundary conditions along the Vostok flow line, and available independent con straints on the ice flow modeling and ice dating, the glaciological models estimate the age of ice at a depth of 3500 m to be between 716 and 930 ka [30,38] .
Another approach to dating very old glacier ice employs the postformation growth («Ostwald ripen ing») of mixed air clathratehydrate crystals (Fig . 4), which occurs in the polar ice sheet due to diffusion of air molecules through the ice matrix . The differ ence in size between crystals induces the gascon centration gradients in the ice matrix and creates a driving force for oxygen and nitrogen diffusion from smaller crystals towards the larger ones . Based on the theory of precipitation from supersaturated so lutions [15], A . Salamatin with coauthors [37] have developed a mathematical description of this process . Their model describes the time evolution of the hy dratesize distribution below the bubbletohydrate transition zone .
The size and number concentration, N, of air hydrates have been measured in thin section of ice under a binocular microscope, using experimental and calculation procedures elaborated for the Vostok ice core [16] . In accordance with the model predic tion, the data show a sharp increase in the mean ra dius of hydrates, , and corresponding decrease in N within the lowest 40 meters of the meteoric ice, in close proximity to its boundary with ice accreted from Lake Vostok's water (see Fig . 3, e, f) . It is worth noting that the volume concentration of hydrates (as well as the mean air content of ice) in the disturbed ice between 3310 and 3539 m remain at the same lev els as in the upper section of the Vostok core covering the last four climatic cycles .
Numerous computational experiments per formed with the model of hydrate ripening [37] and an improved thermomechanical model simulating the ice flow and the heat transfer along the fixed flow tube passing through Vostok Station [38], has shown that the climatically induced fluctuations in the air hydrate geometrical properties [16] become essen tially extinguished below 3400 m and that the mean radius of the inclusions increases linearly with the age of the ice between 3100 and 3500 m [3] . This in ference has been used to extend the extrapolated gla ciological timescale [38] from 3500 m down to the interface between meteoric and accreted ices using experimental data on the size of air hydrates . Using this procedure, the maximum age of the meteoric ice beneath Vostok was calculated to be 1 .85±0 .2 Ma [3] .

Summary and outlook
The ice flow disturbances, which alter the strati graphic continuity and mask the paleoclimatic record in the Vostok ice core below 3310 m, occur on a rel atively small scale and do not disguise the expected growth of airhydrate crystals with depth (age of ice) at the annealing conditions prevailing in the bottom part of the ice sheet . This has allowed the first assessment of the maximum age of the disturbed ice to be carried out on the basis of the hydrate growth experimental data and theory . The preliminary results seem to be encour aging, since they indicate that potentially, Antarctic ice older than 1 .5 Ma is present in the existing Vostok ice core . The air content measured in the deepest section of the core suggests that the surface elevation in the central part of the East Antarctic ice sheet has been sta ble over the time span corresponding to the period of meteoric ice accumulation (>1 .5 Ma?), thus support ing the current model results [32] .
The results obtained at this stage of the studies impel us to propose further comprehensive investi gations in order to meet the Vostok Oldest Ice Chal lenge more closely . We therefore suggest the follow ing next steps within the VOICE project .
1 . Ice dating and chronology reconstruction . This theme includes further refinement of the ice dat ing method based on the airhydrate crystal growth data and theory . The uncertainties of this approach should be more carefully assessed . A number of new absolute methods for old ice dating are currently under development . Taking into account the limited amount of ice available for analysis, it is most fea sible that only 40 Ar/ 38 Ar [5] and 26 Al/ 10 Be [4] dat ing techniques can be applied to obtain independent age estimates for the oldest meteoric ice at Vostok . Reconciliation of the results would help to clarify the uncertainties of the different dating methods used, and would be crucial for the overall progress of the VOICE project . One can also envisage that the age of the Vostok disturbed ice younger than 800 ka may be reconstructed through the matching of globally homogeneous atmospheric parameters (δ 18 O, CO 2 , CH 4 ) measured in the Vostok core to those in the dated EPICA ice core records . New, additional gas measurements on the Vostok ice core are necessary to allow such matching .
2 . CO 2 measurements . If the very old age of Vostok meteoric ice, as inferred from the airhydrate growth below 3500 m, is confirmed, the measure ments of concentration of carbon dioxide in the air extracted from this ice will allow us to extend the Antarctic ice record of CO 2 beyond 1 Ma BP .
3 . Highresolution stable isotope measurements . The existing isotopic record for the deepest section of the Vostok meteoric ice (see Fig . 3, a) was mea sured continuously on 1m long samples . It is advis able to perform new measurements with a resolution of 0 . 1 m or better in order to obtain data which could be useful for studying relative contributions of the smallscale ice mixing and diffusive smoothing of the isotopic signal to the observed dumping of the iso tope record below 3350 m . Depending on the prog ress of this study, the Vostok isotopic record can be deconvolved and matched to the appropriately scaled EPICA record in the age interval between 800 and 400 ka BP . This would help to reconstruct the ice chronology in the upper section of the disturbed ice, between 3345 and 3460 m .
4 . Highresolution studies of ice microstructure (texture, fabric and imperfection of ice crystals) . Such studies, if performed on a continuous basis in the depths interval of 3460-3539 m, may yield valu able information related to the formation of the shear zone and of the submetricscale ice mixing at the base of meteoric ice at Vostok . Highresolution mea surements of dust concentration in selected depth intervals would be of use for interpreting the micro structural data . Aside from this, the microstructural properties may help to distinguish between intergla cial ice and glacial ice layers in the conditions when even a highresolution isotopic profile becomes non informative in this respect . This may be in demand when interpreting the data from gas analyses, which are planned for this section of the core .
The additional measurements proposed above will require a considerable amount of ice, especial ly in the case of the 40 Ar/ 38 A and 26 Al/ 10 Be analyses . The replicate ice core from borehole 5G3 recently obtained at Vostok Station, which duplicates the old 5G1 core between 3458 and 3538 m, will allow im plementation of the planned measurement, and re validation of obtained results, if needed, as well .
5 . Reconnaissance studies along the Vostok flow line, and in the vicinity of Ridge B . Although the continuous paleoclimatic record is hardly available from the Vostok ice below 3310 m, there is potential for recovering a longer, continuous paleoclimatic re cord at a site located upstream of Vostok, in the vicin ity of Ridge B . Detailed geophysical and glaciologi cal surveys in this region, accompanied by modeling efforts, as recommended by the IPICS communi ty [10], should be part of the VOICE project .
In our view, implementing the VOICE project in accordance with the proposed plan will significantly boost our understanding as to how and to what extent the oldest Antarctic ice may help in deciphering the enigma of the MidPleistocene climatic transition . The methodological developments and the expertise associated with progress in VOICE could be useful for future studies of the new oldest ice core, which is anticipated from the new deep drilling project in East Antarctica recently proposed by the international ice core community, represented by the International Partnership for Ice Core Science -IPICS (http:// www .pagesigbp .org/ini/endaff/ipics/intro) .