The Soil Geochemistry in the Beardmore Glacier Region, Antarctica: Implications for Terrestrial Ecosystem History

Although most models suggest continental Antarctica was covered by ice during the Last Glacial Maximum (LGM) it has been speculated that endemic species of soil invertebrates could have survived the Pleistocene at high elevation habitats protruding above the ice sheets. We analyzed a series of soil samples from different elevations at three locations along the Beardmore Glacier in the Transantarctic Mountains (in order of increasing elevation): Ebony Ridge (ER), Cloudmaker (CM), and Meyer Desert (MD). Geochemical analyses show the MD soils, which were exposed during the LGM, were the least weathered compared to lower elevations, and also had the highest total dissolved solids (TDS). MD soils are dominated by nitrate salts (NO3/Cl ratios >10) that can be observed in SEM images. High δ17O and δ18O values of the nitrate indicate that its source is solely of atmospheric origin. It is suggested that nitrate concentrations in the soil may be utilized to determine a relative “wetting age” to better assess invertebrate habitat suitability. The highest elevation sites at MD have been exposed and accumulating salts for the longest times, and because of the salt accumulations, they were not suitable as invertebrate refugia during the LGM.

and N concentrations are within the same range as previously reported 13,15 , and are very low; < 0.15% for N and < 0.5% for C. The water:soil leachate data for both the cations and the anions are shown as ternary plots in Fig. 2. The two soils from CM and the four from MD show a range of cation distributions but the ER samples are dominated by Ca 2+ . The anion plot shows a clear separation between the three sites with ER dominated by HCO 3 − , CM by Cl − + SO 4 2− and MD by NO 3 − or NO 3 − and Cl − + SO 4 2− . The predominance of NO 3 − in the soluble salts of the MD soils can be seen in Fig. 3   anion concentrations 11,15,16 . The nature of the NO 3 − in these soils can be demonstrated in back scattered electron (BSE) SEM images, as NaNO 3 is clearly observable as a dark phase in the MD soils (Fig. 4). Composition of this phase was confirmed by energy dispersive spectroscopy (EDS) spot analysis. The stable isotopic signature of the soluble NO 3 − is similar to values observed from soils from the McMurdo Dry Valleys region ( Table 2). The elevated δ 18 O and Δ 17 O values indicate that the nitrate is derived 50% from tropospheric transport of photochemically produced HNO 3 , and ~50% from HNO 3 formed in the stratosphere 17 . The highly enriched oxygen values are attributed to oxidation reactions in the polar atmosphere involving ozone, and indicate that the nitrate was not produced by biological processes. It is not surprising that the nitrate is not derived from biological processes due to the very low organic carbon in the soil.
Previous work in the TAM has indicated that altitude, not latitude, may be far more important in soil development 13,16 . The amount and frequency of moisture addition to these soils have significant effects on the total salt accumulation as well as the elemental distribution of the salts. The high concentrations of nitrate in the MD samples imply there has been little to no moisture input into these soils for a significant amount of the time they have been exposed. Because salts such as NaNO 3 are extremely soluble, we can calculate the potential age of the last significant moisture flush, "wetting age", through the NO 3 -rich soil using the following approach. We can take the total water-leachable NO 3 − in the top 10 cm of this soil and divide this by the NO 3 − flux in the snow from the nearby Dominion Range ice core 17 . A wetting age of between 750,000-850,000 years BP is computed for the Meyer Desert.
If we use the same NO 3 − flux for the Dominion range determined by Lyons et al. 17 for the CM and ER sites, the accumulated NO 3 − yields wetting ages between 20-23,000 years at CM and between 56 and 63 years for ER. These findings suggest that the lower ice-free locations along the Beardmore receive more precipitation and/or melt. These calculations include a number of important assumptions, including that the NO 3 − input into the snow has remained constant over this time interval, and that there has been no post-depositional loss of NO 3 − either from the snow (which would impact the original nitrate flux calculation of Lyons et al. 17 ) or from the soil. Data suggest that there can be NO x loss from the Antarctic snow pack 18 , and others have argued for the possibility that NO 3 − can be added to TAM soils through a downslope transfer from the plateau 15 . It has been demonstrated that the lost NO 3 − from Antarctic snow is the lighter isotope 19 , which is what is accumulating in the soils, so we cannot rule out that a portion of the NO 3 − in these soils is from recycling of NO 3 − from nearby snow. In any effect, this "wetting age" is a best estimate based on the available information. It is interesting to point out that the Plunkett and Beardmore drifts were estimated to have a maximum age of 23,800 years 11 , but as noted above, recent cosmogenic exposure ages 13 from moraines in these drifts yield much older ages (300,000-500,000 years).  The "wetting age" of our sample locations and their elevation during the LGM is important for determining their potential to serve as refugia. Our MD samples came from the area adjacent to Brown's Butte near the upper bench of the Oliver Platform by the Koski Fault at the elevation of above 2200 m (our highest elevation sample is at 2551 m; Figs 1 and 5, Table 1). Ackert and Kurz 20 reported a number of cosmogenic exposure age dates from this area but none above 1900 m. Their ages range from ~2 to 5.37 Myr at elevations of 1805-1820 m, while the Koski Fault age data range from 0.37 to 2.11 Myr (at elevations of 1875-1950 m). So our highest elevation samples probably represent Sirius Group sediments that sit on an erosional surface above the local bedrock and are significantly older than lower elevation soils 21 . The soil at 2551 m could have been a candidate for a potential refugium location not only during the LGM, but for all of the glaciations during the Pleistocene, as a moraine between the lower Oliver and upper Oliver Platform dates at ~5 Myr 20 (Fig. 5).
Recent evidence indicates that both the West and the East Antarctic Ice Sheets (WAIS and EAIS, respectively) were at lower elevations in their interiors during the LGM than previously thought 22,23 . Denton and Hughes 24 have argued that the surface elevations of the EAIS inland of the TAM were lower than present during the LGM except near the outlet glaciers, which were dammed by grounded ice in the Ross Sea. This thickening of the outlet glaciers led to slightly increased surface elevations at the heads of the glaciers in the TAM but substantial rises in ice surfaces at the mouths of the glaciers near Ross Ice Shelf 24 . However, exposure age dating in the Darwin Glacier area indicates that even the increases near the ice shelf were probably lower than previous estimates 14 . Denton and Hughes 24 suggest a maximum inland ice-rise of the Beardmore Glacier of 35-40 m, but increases of as much as 1300 m in the Mt. Kyffin (Ebony Ridge) region. Taking all the evidence into account it has become clear that even during glacial maxima times, high-elevation ice-free areas existed even at these very high latitudes. The high elevation areas in the TAM could have served as refugia during the times when the lower elevation regions were covered with ice or in southern Victoria Land, when large lakes covered the valley floors 25 .
We have computed the Chemical Index of Alteration, CIA, using the major oxide data for each soil sample. This parameter has been utilized to determine the extent of chemical weathering in sedimentary rocks, sediments and soils 26 , and is determined by this equation: where Ca*O is the non-carbonate calcium present in the sample. As aluminosilicate minerals become more intensely weathered, the alkaline earths and alkali metals are lost and the CIA approaches 100%. Fresh granites/ granodiorites have CIA values between 45-55 26 , and mean continental crust is 58 27 . The CIA values from these soils are plotted vs relative latitude in Fig. 6. Clearly the MD data show much lower values, more representative of unweathered material, than the ER samples, which appear highly weathered. The CM samples lie in between. The previous compilation of data indicates that nitrogen concentrations increase with elevation, and with age in most Antarctic soils. Inorganic carbon (IC) increases with precipitation, and in general, organic carbon concentrations do not change with age of the soils 15 . The highest IC we observed is at ER, where we also see the highest water-leachable HCO 3 − and Ca 2+ in the soil (Table 1). IC was only observed in one of the four MD samples. The low solid OC/P ratios in these soils are due in very large part to the low OC values. The high water-leachable N:P ratios in the CM and MD sites are due to the very high inputs of atmospherically derived nitrate as described above, and as indicated by the isotopic data of the nitrate. These data indicate that with the exception of ER, these soil ecosystems are very P limited. This is also the situation for pre-LGM tills in southern Victoria Land 28 . Hence both the leachate data, with its low concentrations of soluble salt and predominance of Ca 2+ and HCO 3 − , and the bulk soil analysis, with a much higher CIA, suggest that ER has been more impacted by liquid water than the other sites, and that water has leached the soluble salts and caused extensive chemical weathering. In contrast, the MD samples have high concentrations of the most soluble salts, such as NaNO 3 , and little chemical weathering has taken place. Our ER samples are very close to, and at the same elevations, as a hotspot of lichen diversity described by Green et al. 29 . This location yielded 28 different lichen species and their occurrence has been attributed to "relicts" that have survived on-site through the glacier maxima periods. These authors suggested that these relict populations at the ER site could be as young as 200 Kyr, but they could be much older as well. Nonetheless, this site is clearly a more suitable habitat for life due to the abundance and/or frequency of liquid water as evidenced by the geochemical information.
When is a Refugia Not a Refugia? As noted in the introduction of this paper, there are significant observational and molecular/genetic data to suggest that much of the invertebrate fauna found in the continental soils of Antarctica today have survived through the glacial maxima periods of the Pleistocene 4,6 . In order for these species to persist over this time period, there would be a need for habitat continuity for each species 6 , and these habitats would probably have to be at higher elevations that the WAIS and the EAIS were unable to overtop as they advanced oceanward. However, different soil factors and climate have been found to define suitable habitats for particular species, and hence community composition in Antarctic soils 8,10 . Younger soils with less salt have a more complex community structure 9 . High salt concentrations, especially nitrate, can strongly inhibit the survival of the endemic nematodes including, Scottnema lindsayae and Plectus antarcticus 30 . Scottnema lindsayae previously collected from the Beardmore region show no discernable morphological or ITS1 rDNA sequence variation from those of the McMurdo Dry Valleys, suggesting relatively frequent and widespread dispersal within and among suitable habitats 31 . Clearly the MD sites with their very high nitrate concentrations would not be good refugia candidates, even though they were ice-free throughout the Pleistocene.
So the question becomes, where were high altitude refugia that also had suitable habitats? It is now very clear that abiotic spatial gradients that cut across landscape units (e.g. streams to soils) and elevational gradients, rather than latitudinal gradients, greatly impact the distribution of soil invertebrates and other organisms in Antarctic soils 8,9,13,29,32,33 . These gradients in abiotic properties can exist over very short distances 13 . This local patchiness may be related to the slow lowering of the ice sheet via sublimation and/or differential snow accumulation and consequent melting through solar radiation absorption by the surrounding low albedo soil. Scarrow et al. 13 point out that this small-scale variation in the abiotic properties of these high elevation soils cannot be effectively mapped. Thus the overlap of ice-free refugia and suitable habitat may have only existed on such scales that make it extremely difficult to map and hence identify. If this is indeed the case, the exercise of predicting the past locations of refugia will require describing both past soil conditions and glacier/ice elevation on a much more detailed scale.

Conclusions
Soil samples analyzed from the Beardmore Glacier area of the central Transantarctic Mountains (TAM) of Antarctica have very different geochemical characteristics. These differences are driven by the present or recent past availability of liquid water that both leaches soluble salts and acts as a chemical weathering agent of the soils. The soils with the highest concentrations of salt do not provide suitable habitat for metazoan life even though they were undoubtedly ice-free during the Pleistocene glaciations. These data and previous soil biogeochemical/ ecological research at lower latitudes in Victoria Land suggest that gradients in abiotic properties exist on many scales in Antarctic soils. Some of these variations, including those most closely associated with habitat suitability, occur at very small spatial scales, complicating the interpretation of past biogeochemical patterns and the identification of LGM refugia. Aliquots of dried soil were leached with deionized water (DI) using a 1:5 soil:water ratio 13 . The leachates were filtered with a 0.4 μ m polycarbonate membrane filter and analyzed for major cations and anions (Na + , K + , Ca 2+ , Mg 2+ , Cl − , SO 4 2− , NO 3 − ) using ion chromatography. Phosphorous, as phosphate, was analyzed using a Skalar Nutrient Analyzer. The ion chromatographic data had a precision of ≤ 5%. Bicarbonate was determined by subtracting the sum of Cl − , NO 3 − and SO 4 2− equivalents from the sum of Na + , K + , Ca 2+ , and Mg 2+ in equivalents 34 . Another subset of bulk soil was analyzed for major element oxides, including P 2 O 5, using X-ray fluorescence according to the lithium borate dissolution technique 35 . USGS standard reference materials were measured similarly and were within 12% of the reported values. Total carbon, organic carbon and total nitrogen were measured on bulk samples using a Leco C/N analyzer. The percentage of sand and silt grain size of subsets of soils was also determined using standard sieving techniques. All of these analyses were done at The Ohio State University. Several samples were analyzed using an FEI Quanta FEG field emission SEM. Samples were affixed to carbon tape on a stub and coated with Au-Pd before analysis. Finally, another subset of samples was leached with water and the nitrate in the leachate was analyzed for δ 15 N, δ 18 O, and δ 17 O at Purdue University using the techniques of Michalski et al. 36 .