First insights into the age of the giant ice deposits in the Eisriesenwelt cave (Austria)

Frozen water is the most widespread type of ice present in ice caves and forms ice stalagmites and stalactites as well as floor ice, which is often several meters thick. Organic macroremains are commonly rare in this type of cave ice, which makes it difficult to establish a chronology and severely limits the use of such ice deposits as paleoenvironmental archives. Here, the chronology of such ice deposits in the inner part of the glaciated Eisriesenwelt, one of the world’s largest ice caves located in the European Alps of Austria, is determined by a combination of radiocarbon and 230Th dating of cryogenic calcite. The data suggest that this cave ice has formed over the last three millennia, with a marked increase in the average accumulation rate during the thirteenth century, coinciding with the onset of the Little Ice Age in the Alps. Data from a second site closer to the entrance suggests that large parts of this tourist cave were likely ice-free during the Medieval Warm Period and that a substantial part of the ice is probably a relic of the Little Ice Age. The current warming has already penetrated deeper into the cave than during the Medieval Warm Period, although air exchange during the warm season is restricted by a door at the cave entrance.


Study site
Eisriesenwelt is an extensive labyrinthic system of almost 40 km of passages in the western part of the Tennengebirge, a karst massif in the Northern Calcareous Alps of Austria (47° 30′ 40′′ N, 13° 11′ 35′′ E, Fig. 1).The main (lower) entrance opens in a cliff at an elevation of 1641 m a.s.l.; a much smaller, second (lower) entrance is located at 1838 m a.s.l.The upper entrance (possible multiple entrances) are likely located in dolines on the plateau of the Tennengebirge some 500 m above the main galleries and are not fully explored yet.As a result of this cave geometry, cold winter air enters through the lower entrances, and ascends through the cave's passageways, eventually exiting on the plateau (chimney effect).This influx of winter air leads to a strong cooling of the first 550 m behind the main lower entrance and the preservation of congelation ice, covering an area of approximately 11,000 m 2 .In late spring, the atmospheric dynamics in the cave are reversed, as the cave air is cooler and denser compared to the warmer outside air, resulting in a strong, cold cave wind emanating from the lower entrances.
The main ice-forming season is spring and early summer, when the snow melts above the cave and this cold water seeps into the strongly undercooled frontal part of the cave.In contrast, ice sublimation typically occurs in winter by the fairly dry and cold air entering the cave.Minor ablation also occurs at the end of summer, particularly close to the inner border of the ice-bearing part.A large hall known as Eispalast (Fig. 1) is the demarcation point for year-round ice presence in the cave and is characterized by minimal seasonal temperature fluctuations when compared to the entirety of the ice-bearing portion of the cave 28 .Microclimatic measurements performed at the Eispalast during the period from December 2007 to November 2008 revealed that the mean cave air temperature in summer was approximately 0.0 °C, while the mean winter temperature registered at − 1.1 °C29 .Deeper into the cave, only localized seasonal ice formations are observed, particularly during anomalously cold winters.
Eisriesenwelt offers spectacular exposures of the ice stratigraphy revealing well-bedded ice with varying amounts of air inclusions, but without organic macroremains.Hence, the age of this ice is largely unknown.The first attempt to date this ice was made half a century ago by studying pollen.Four ice samples from Mörkgletscher and Großer Eiswall (Fig. 1) yielded pollen spectra that included pollen of cultivated plants and were interpreted as probably formed during the Little Ice Age 30 .May et al. 20 examined a core drilled into the floor ice of Eispalast.An estimate based on radiocarbon dating of particulate organic matter suggests a basal ice age of around 5000 yr BP, albeit with a large uncertainty.
Exposures at several ice cliffs in Eisriesenwelt show the presence of fine CCC, most prominently at Mörkgletscher (Fig. 2).This 8 m high cliff (which extends higher up on the back side and comprises about 11.7 m of total ice exposure) shows several white to light brown layers in the lower part consisting of fine CCC, while the ice above is conspicuously cleaner and lacks CCC layers.These "dust layers" can be traced up to several meters along the ice cliff and pinch out laterally.In some of these layers angular blocks of limestone are intercalated.
Based on an earlier study 31 , we systematically sampled the different CCC-bearing layers in this exposure, ranging in thickness from about 1 to several mm.In addition, we sampled thin CCC-lenses in the ice on the top of this cliff (on the back side) and extended the 11.7 m-long profile downward by coring the ice floor in front of the cliff (total almost 14 m) Finally, we also analyzed CCC extracted from ice drilled in Eispalast and from an exposure in Eistunnel (Fig. 1).

CCC mineralogy and petrography
All samples consist of calcite.This also applies to two samples transported in cold condition to the laboratory and examined by XRD (see Methods).No evidence of precursor phases such as ikaite was found.
Under the microscope the samples are composed of > 99% euhedral calcite crystals and aggregates thereof.Granulometric measurements show median grain sizes between 44 and 199 µm (Table 1).The crystals show a homogeneous whitish color in reflected light.Impurities, i.e., detrital grains derived from the host limestone, are rare and easy to recognize based on their color (commonly light red to yellow) and anhedral shape.
Optical microscopy and SEM observations revealed a range of particle shapes.The most abundant type are crystal aggregates showing a radial arrangement of intergrown crystals (Fig. 3).These aggregates are often asymmetrical in structure and show a conspicuously flat upper side and a lower side covered with crystal tips.The upper side sometimes shows holes with a diameter of 5-15 µm.The second most abundant type are skeletal crystals or skeletal crystal aggregates.These show a very high microporosity.There are all transitions from crystal aggregates to skeletal crystals.Finally, there are (hemi)spherulitic aggregates showing a concentric shell structure.These particles often have a flat upper surface and are partly porous.

CCC stable isotope composition
Fine CCC from Mörkgletscher are characterized by highly positive δ 13 C values and moderately negative δ 18 O values (Fig. 4, Table 1).δ 13 C values reach as high as + 13.5‰.Aliquots measured of each CCC sample commonly show a tight clustering with only a few tenths of permil variability in both C and O isotopes; a few samples yielded a slightly larger variability of up to about 1.5‰ in δ 13 C and up to about 0.5‰ in δ 18 O.The C and O isotope data are linearly correlated (r 2 = 0.52) with a positive slope of 0.83 (Fig. 4).These isotopic values are distinct from both calcite derived from weathering of the host rock limestone and from coarse grained CCC which are present in     www.nature.com/scientificreports/interior (and currently ice-free) parts of this cave and which document episodes of much more extensive cave glaciations prior to the Holocene (ref. 32and unpublished data by the authors).

Radiocarbon results
15 CCC samples from Mörkgletscher were analyzed for radiocarbon and the resulting conventional ages range from 921 to 3517 BP (Table 1).There is no relationship between 14 C ages and δ 13 C values (r 2 = 0.0002) suggesting that there is no significant detrital contamination by radiocarbon-free host rock particles characterized by much lower δ 13 C values (Fig. 4).Replicates of samples from three CCC layers yielded conventional ages that agree within 4.4, 7.0 and 9.0% and two samples taken laterally within one CCC layer deviate by 12% (Table 1).Plotted against depth, the conventional ages decrease from the bottom of the ice section up, with the uppermost samples being approximately 2000-2500 14 C years younger than the deepest ones (Fig. 5).DIC of a sample (ERW-B) taken from the water coming from the Alphorn shaft yielded 95.2 ± 0.5% pmC, which indicates a low dead carbon effect (4.8%) and a small reservoir age (395 ± 42 14 C yr). CCC formed experimentally in small pools in Eispalast 33 yielded conventional radiocarbon ages of 356 ± 20 BP (sample L1-1, using water from the Alphorn shaft) and 511 ± 19 BP (samples L2-2, using water from the Elefant site-Fig.1, Table 1).Table 2. 230 Th ages of CCC samples from Mörkgletscher.U decay constants: λ 238 = 1.55125×10 -10 (Jaffey et al. 55 ) and λ 234 = 2.82206×10 -6 ref. 54 .Th decay constant: λ 230 = 9.1705×10 -6 ref. 54 .*δ 234 U = ([ 234 U/ 238 U] activity -1) x1000.**δ 234 U initial was calculated based on 230 Th age (T), i.e., δ 234 U initial = δ 234 U measured x e 234 ×T .Corrected 230 Th ages assume the initial 230 Th/ 232 Th atomic ratio of 4.4 ± 2.2 ×10 -6 .Those are the values for a material at secular equilibrium, with the bulk earth 232 Th/ 238 U value of 3.8.The errors are arbitrarily assumed to be 50%.***BP stands for "Before Present" where the "Present" is defined as the year 1950 A.D. Three CCC samples were analyzed for 230 Th.Sample M-G1 from near the top of the ice section yielded an age of 953 ± 303 yr BP.Sample M-A20 from the lower part of the cliff and sample M-E154-156 taken from a drill core at the foot of the cliff yielded ages of 2262 ± 4346 and 3737 ± 1954 yr BP, respectively, which are in stratigraphic order but very imprecise due to the large Th correction (Table 2).

Origin of the CCC layers and the ice cliff
Several lines of observations show that the discontinuous "dust layers" at Mörkgletscher are deposits of fine CCC with only tiny traces of other, i.e., detrital (limestone) particles.These observations include the euhedral shape and the stable isotopic composition of these carbonate crystals and crystal aggregates.The stable isotopic composition is not only distinct from that of the local limestone, but also consistent with fine CCC extracted from an ice core drilled in the Eispalast 20 and fine CCC found on top of modern ice surfaces in this cave (unpublished data by the authors).The C isotope values overlap with fine CCC from ice caves in Romania 34 and the Canadian Arctic 35 , but the O isotope values are shifted towards higher (Romania) and lower values (Canada), reflecting the overall δ 18 O values of meteoric precipitation (higher and lower in Romania and Canada, respectively).These data suggest fairly rapid crystallization from a thin water film on ice surfaces.The abundance of skeletal crystals and spheroidal forms is consistent with this mode of formation.
A special feature of the CCC layers at Mörkgletscher is their varying thickness and their lateral discontinuity.Both indicate that the thicker layers in particular were not deposited in situ, but represent redeposition of CCC.Our own observations made in winter in the ice-bearing part of Eisriesenwelt showed that fine CCC are being released from ice surfaces due to sublimation, forming tails of white to light brown powders.Similar observations were reported from ice caves in the Canadian Arctic where sublimation released CCC disseminated in ice, forming CCC deposits up to 5 cm thick 24 .These fine powders can be transported by the cave wind, which in Eisriesenwelt can reach more than 10 m/s on cold days in areas of narrow cross sections.It is interesting to note that cavers who explored and surveyed the cave about a century ago reported "flour-like dust often in fingerthick layers" at a narrow squeeze known as Sturmsee 36 (this narrow passage had to be overcome by crawling and was later artificially widened; Fig. 1).These deposits were likely re-worked fine CCC deposited in the lee of this formerly strongly ventilated squeeze.A similar setting likely gave rise to the CCC layers at Mörkgletscher, located in the lee of another squeeze called Eistor (Fig. 1).Back in 1913, the first explorers had to crawl through a narrow gap between the ice and the rock ceiling to enter the hall called Mörkdom (Fig. 1) which hosts Mörkgletscher 37 .It is thus likely that these fine CCC had formed slightly further west (i.e., closer to the cave entrance) and became released from the ice by sublimation as a result of cold winter air being drawn into the cave through the main entrance.This hypothesis could also explain the fact that the CCC layers pinch out laterally towards the east, i.e., leeward from the former Eistor squeeze.Angular blocks of limestone embedded in some of the thickest CCC layers (but absent in clean ice above) represent breakdown from the ceiling, likely as a result of strong cooling and frost shattering.The thicker CCC layers therefore likely represent microhiatuses reflecting anomalously cold winter conditions giving rise to strong sublimation and short-distance eolian transportation of fine CCC and redeposition behind narrow passages followed by burial by younger layers of ice.There is no evidence, however, that the thicker CCC layers mark major (long lasting) episodes of ice retreat.The presence of the CCC layers marking microhiatuses therefore indicates that the mean rate of ice accumulation in the lower part of the profile was likely lower compared to the middle and upper part, where CCC layers are rare, thinner and can only be traced laterally for less than a few meters.
The ice cliff of Mörkgletscher did not exist in 1913, when the ice body extended all the way to the opposite cave wall.By about 1925, a 2 m-wide gap had opened due to ice sublimation 37 , which since then has widened to almost 8 m, exposing the internal structure of this floor ice and its CCC layers.The average rate of ice wall retreat due to sublimation during the cold season (plus a smaller contribution of summer melting) was about 6 cm per year.

Radiocarbon and 230 Th constraints on the age of the CCC layers
Although the (thicker) CCC layers in the lower part of the ice section likely mark microhiatuses, they are nevertheless the key to obtaining quantitative information on the age of the surrounding ice.The purity of the CCC samples in terms of detrital contamination and the replicated samples are strong arguments that the overall decrease in conventional radiocarbon ages with increasing distance from the base of the section reflects an age gradient of approximately 2400 14 C years, using the youngest radiocarbon age of each layer (Fig. 5).To constrain the timing of the start and end of ice accumulation, the radiocarbon dates must be corrected for the effect of dead carbon, followed by calibration to calendar years.
The dead carbon effect was assessed using radiocarbon analyses of modern CCC and DIC and compared to less precise constraints provided by 230 Th dating.The radiocarbon dates of two modern CCC samples formed in the nearby Eispalast (356 ± 20 BP and 511 ± 19 BP) differ somewhat, reflecting the different water sources used for the two freezing experiments 33 .Such small radiocarbon variabilities of DIC are not uncommon in caves 38,39 .The value of DIC of one of the waters (395 ± 42 BP) is consistent with the CCC formed from it (356 ± 20 BP; the second water was not analyzed).We used the mean of all three measurements and its standard error (421 ± 17 BP) to correct the radiocarbon dates of CCC samples from Mörkgletscher.This approach assumes that the dead carbon effect has remained constant during the past few millennia.This appears justified because major changes in the composition and bioproductivity of the thin and patchy soil present above this part of the cave and of water-rock interactions in the vadose zone above the cave are unlikely in this time frame.This also includes land use changes which can be ruled out for the steep and rocky terrain above the cave.The resulting dead carbon-corrected dates were then converted to calendar ages using IntCal20 40 (Fig. 6).The final ages are plotted on Fig. 6  www.nature.com/scientificreports/this is the following: Several processes may result in calibrated radiocarbon ages being variably older than the "true" age.These include (a) contamination by detrital carbonate, (b) a higher dead carbon fraction, and/or (c) reworking of CCC derived from older ice strata.On the other hand, processes that would result in anomalously young CCC ages such as contamination by younger CCC or organic matter can be ruled out.The proposed linear age model is constrained by the youngest ages of each CCC layer following the rationale outlined above.A more accurate and precise depth-age model is currently not possible given the nature of the radiocarbon dates of such CCC samples (which represent maximum age estimates) and the limited number of CCC layers.The validity of our proposed age model, however, is confirmed in the upper part by a 230 Th age whose 2σ uncertainty almost overlaps with the age model (Fig. 6).The 230 Th age obtained from the deepest CCC sample is imprecise but still provides an independent age control, and it is consistent with the age model even within its 1σ uncertainty range (Fig. 6).The third 230 Th age is unfortunately too imprecise to be used.The proposed depth-age relationship assumes that no melting had occurred at the base of this ice body.Basal melting would remove older ice and hence truncate the stratigraphic record.This process has been documented in some ice caves (e.g. 9,41), but several lines of observations suggest that it does not play a significant role in Eisriesenwelt.(i) In all places in the cave where the contact between the basal ice and the rock or gravel beneath is well exposed, there is no macroscopic evidence of melting (based on observations during all seasons, but primarily during summer and autumn).If cave ice reaches the melting point it changes in crystal fabric resulting in a honeycomb-shaped structure (known as Wabeneis 42 ).Such features were locally observed at the surface of ice stalagmites in late summer but were never observed in the basal ice.(ii) Ice temperature profiles obtained in Eispalast recorded negative temperatures year-round in the deepest ice layer at 7.3 m depth close to the base of the ice (unpublished data by the authors).Also, an earlier study of ice temperatures at Eispalast based on temperature profiles found no evidence of basal melting 29 .(iii) There is no evidence of measurable ice movement in this cave (due to basal sliding), even in parts where the ice forms steep slopes.This is based on decadal-long observations by the show cave personnel (pers.comm. A. Rettenbacher).If there had been ice movement, fortifications for the tourist trails drilled vertically into the ice slopes would have been deformed and/or tilted, which was not observed.This argues for the ice being permanently frozen to the rock or gravel substrate.In some exposures ice strata are inclined and change their dip laterally, mimicking folding.This structure, however, is not related to ice folding, but reflects the growth structure of large ice stalagmites and their variable internal stratification when exposed along an ice tunnel or cliff.In summary, there is no evidence of basal melting in Eisriesenwelt.During colder climate periods such as the Little Ice Age basal melting was thus even more unlikely.The proposed age model suggests that ice accumulation at Mörkgletscher commenced about 3.2 cal kyr BP and terminated at about 0.2 cal kyr BP.The age model suggests a mean accumulation rate of 0.25 cm per year during the first two millennia until about the thirteenth century AD, followed by a mean rate of about 0.79 cm per year in the subsequent six centuries.This overall increase in growth rate is consistent with the scarcity of CCC layers (and the lack of thick CCC layers) in the middle and upper part of the section.In contrast, the abundance of CCC layers in the lower part suggests repeated episodes of intensive ice sublimation (possibly also minor ablation), hence overall lower accumulation rates, but still an overall positive ice mass balance.The onset of ice accumulation at Mörkgletscher around 3.2 kyr BP may have been related to an episode of glacier advance in the Alps, known as the Göschenen 1 advance and more recently called Iron Age Advance Period 1, starting around 3 kyr BP 43 .The change towards a higher mean accumulation rate during approximately the thirteenth century AD occurred at the onset of the Little Ice Age, which started in the Alps around 1260 AD and ended at about 1860 AD 44 .In addition, the end of the ice accumulation phase-based on extrapolation (Fig. 6)-falls at the end of the Little Ice Age.In the 1920s, the ice had started to retreat from the southern cave wall and already exposed part of the Mörkgletscher ice cliff, i.e. the mass balance in this inner part of the show cave had become negative by then.Since then, the ice cliff has retreated by about 6 cm per year.
We also tested the CCC-dating approach in two other parts of Eisriesenwelt and briefly report here preliminary results.First, we applied this approach to Eispalast (Fig. 1), where an ice core drilling had encountered 7.3 m of ice 20 .We re-drilled this ice body and sampled a CCC layer at 3.2 m depth (sample F320, 3955 ± 23 BP).Using the same correction for the dead carbon effect as applied above, the calibrated age suggests that this ice is not more than 3.9-3.7 cal kyr BP old (2σ range, Table 1).Whether the basal ice is about 5 millennia old as suggested by imprecise radiocarbon data of particulate organic matter 20 remains an open question.
Secondly, we sampled fine CCC layers in a place called Eistunnel, which is a tunnel that was artificially cut through the ice in the middle part of the show cave in 1963 (Fig. 1; pers.comm. A. Rettenbacher) and which has since widened substantially by sublimation.Multiple layers of fine CCC are present in this ice whose gravel base is well exposed.A sample from a CCC layer about 45 cm above the ice base yielded a radiocarbon age of 1046 ± 20 BP (sample ETB).When corrected for the dead carbon effect the same way as the Mörkgletscher samples, the calibrated age is 653-553 cal yr BP (2σ range, Table 1), which suggests that this ice was likely formed between about the fourteenth and fifteenth century, assuming that no ice was lost due to basal melting.As detailed above for the Mörkgletscher site, basal melting does not play a significant role in this cave today.Given the more proximal location of Eistunnel with respect to the lower cave entrance, basal melting is even more unlikely there compared to Mörkgletscher.In essence, this CCC age falling into the Little Ice Age implies that the central (and likely also the frontal) part of the show cave may have been largely ice-free during medieval times.The observation that older ice is preserved in the inner part (Mörkgletscher, Eispalast) is in contrast to what would be expected from the point of maximum winter cooling (more pronounced in the near-entrance part).Part of this discrepancy may be explained by (i) more pronounced ice sublimation in the near-entrance part due to a greater temperature difference, (ii) the presence of large depressions in the inner part allowing thick and less vulnerable floor ice deposits to form compared to the more proximal parts that are characterized by steep slopes, and (iii) local ingress of water.Today, large ice stalagmites form in the cooler parts of the cave close to the entrance due to the drip water, while water entering the inner parts of the ice cave mostly forms sheet-like deposits (unless artificially diverted).
The fact that the ice at Mörkgletscher has been retreating for about a century suggests that the current warming has already penetrated deeper into the cave than during the Medieval Warm Period.In this context, it is important to mention that the cave's microclimate has been artificially modified for about a century by closing the lower entrance with a door during the warm season (to prevent the cold cave air from escaping).Old surveys indeed shows more ice in Eisriesenwelt during the 1950s and 1960s compared to 1920 32 .Without this door, the summer cave temperatures would be higher, the near-entrance part of the show cave would most likely be ice-free (only seasonal ice) and the ice deposits in the inner ice parts would probably become increasingly vulnerable.

Perspectives
Radiometric dating of congelation ice remains a challenging task, which ultimately limits attempts to unravel the full potential of this poorly understood paleoenvironmental archive.
Given the wide-spread lack of organic matter in this type of cave ice (and the challenge to constrain soil and karst reservoir effects), the use of fine CCC as a chronometer seems an attractive way forward.In terms of radiocarbon, two aspects are important.First, radiocarbon dates of CCC are maximum age estimates and the accuracy of the final age hinges on the magnitude of the local dead carbon effect.Secondly, fine CCC disseminated in ice and thin CCC layers likely formed together with the surrounding ice.In contrast, thicker CCC layers represent microhiatuses and require sublimation of pre-existing CCC-bearing ice followed by short-distance transportation by cave wind to accumulate in such a concentrated manner.The issue of the dead carbon effect could be tackled by a systematic DIC study of modern cave water, but we are unaware of such a kind of study in ice caves.Depending on the size of the cave, the thickness of the rock overburden, the types of water pathways, and the structure of the soil, it is likely that such a study would yield a range of values for the dead carbon effect, probably also on a seasonal scale.Applying such a modern average correction factor to millennia-old ice requires independent age control.This is why 230 Th data of fine CCC would be highly valuable.Given their very large surface area, the delicate CCC crystals and crystal aggregates are inherently less pure with respect to detritally bound initial 230 Th than regular speleothems such as stalagmites.Attempts to purify fine CCC samples by repeated washing in an ultrasonic bath and hand-picking of contaminant particles under a binocular were only partially successful as shown by the still very low 230 Th/ 232 Th ratios (Table 2) resulting in large corrections.

Figure 1 .
Figure 1.Simplified plan view of the frontal part of Eisriesenwelt showing the ice distribution, the study site Mörkgletscher, as well as other sites in this cave mentioned in the text.Map was produced using Adobe Illustrator (v27.2, www.adobe.com).

Figure 2 .
Figure 2. Exposure of stratified congelation ice at Mörkgletscher.CCC-bearing "dust" layers in the lower third of the cliff stand out because of their light brown color.Sampled CCC layers are labelled, whereby the lowermost sample (M-E154-E156) was taken from a core drilled vertically into the ice floor.The three uppermost samples (M-G1 to M-G3) were obtained from laterally discontinuous CCC layers on the opposite side of the ice cliff and stratigraphically above the highest ice shown on this picture.Photo courtesy Robbie Shone.

Figure 3 .
Figure 3. Microscopic appearance of fine CCC from Mörkgletscher.(a) and (b) are optical images of aggregates of euhedral calcite crystals showing a variety of particle shapes.(c) and (d) are SEM images of rafts and spheroids, respectively, both characterized by a conspicuously flat side where these particles touched the ice during growth.

Figure 4 .
Figure 4. Carbon and oxygen isotopic composition of fine CCC from Mörkgletscher (n = 66) compared to Dachstein limestone (n = 38), the host rock of the cave (host rock data31 ).Also shown is the stable isotopic composition of coarse grained CCC from inner and currently ice-free parts of Eisriesenwelt (unpublished data by the authors).

Figure 5 .
Figure 5. Conventional radiocarbon ages of fine CCC from the Mörkgletscher ice cliff plotted against ice thickness.The analytical uncertainty of the measurements is well within the size of the symbols.Black arrows indicate CCC layers where duplicate samples were analyzed.The green arrow marks two samples taken laterally within a CCC layer.

Figure 6 .
Figure 6.Radiocarbon ages of fine CCC samples from the Mörkgletscher ice cliff corrected for a constant reservoir age of 421 ± 17 14 C years and calibrated, showing median ages and associated 2σ uncertainties.The two yellow-green data points are 230 Th ages with their 2σ uncertainties.The thick dashed red line is the proposed depth-age model based on the youngest ages of each sampled CCC layer.