The magnesium isotope record of cave carbonate archives

Introduction Conclusions References

record changes in outside air temperature as driving factor rather than rainfall amount.The alpine Pleistocene flowstones from Austria (SPA 52: −3.00 ± 0.73 ‰; SPA 59: −3.70 ± 0.43 ‰) are affected by glacial versus interglacial climate change with outside air temperature affecting soil zone activity and weathering balance.Several data points in the Austrian and two data points in the German speleothems are shifted to higher values due to sampling in detrital layers (Mg-bearing clay minerals) of the speleothems.The data and their interpretation shown here highlight the potential but also the limitations of the magnesium isotope proxy applied in continental climate research.An
More recently, it has been proposed that multi-proxy studies including novel approaches such as fluid inclusions, noble gases, clumped isotopes and non-traditional isotope systems (Kluge et al., 2008;Scheidegger et al., 2008;Immenhauser et al., Introduction Conclusions References Tables Figures

Back Close
Full 2010; Da ëron et al., 2011;Reynard et al., 2011) may provide valuable information and complement more traditional proxy data.Amongst these more recently established proxies are Mg isotope (δ 26 Mg) data.Previous work dealing with Mg isotope fractionation in Earth surface environments includes investigations from carbonate archives and field and laboratory precipitation experiments.Data available in the literature include δ 26 Mg ratios from rain, snow, riverine, soil and drip water as well as hostrock and cave carbonates and soil and cave silicate data (Galy et al., 2002;Young and Galy, 2004;de Villiers et al., 2005;Tipper et al., 2006bTipper et al., , 2008Tipper et al., , 2010;;Buhl et al., 2007;Brenot et al., 2008;Immenhauser et al., 2010;Pokrovsky et al., 2011;Riechelmann et al., 2012b).Nevertheless, with the exception of the preliminary work shown in Buhl et al. (2007), a critical application of the above-cited work to speleothem time series data is lacking.
Here we report on the potential and limitations of Mg isotope time series proxy data interpretation from eight speleothems collected in six caves on three different continents.The goals of this study are to (1) review parameters known to affect δ 26 Mg fractionation between land surface and cave carbonate deposition, to (2) apply these findings to all available δ 26 Mg time-series data sets from Pleistocene/Holocene speleothems and flowstones.

Case settings
This study makes use of δ 26 Mg isotope data from six stalagmites and two flowstones.
These include two stalagmites from Germany (Atta Cave and Bunker Cave), two stalagmites from Morocco (Grotte d'Aoufous and Grotte Prison de Chien), two stalagmites from Peru (Cueva del Tigre Perdido) and two flowstones from Austria (Spannagel Cave).The reader is referred to Fig. 1 and Table 1 for detailed information regarding cave parameters and speleothems investigated.

Magnesium isotope analysis
Powder subsamples from longitudinally cut stalagmites and flowstones were dissolved in 6 M HCl.The solution was dried and re-dissolved with 250 µl of a 1 : 1 mixture of HNO 3 : H 2 O 2 (65 % : 31 %).Subsequently, the solution was evaporated and again redissolved in 1.25 M HCl.The Mg-bearing fraction was extracted by using ion-exchange columns (BioRad ion exchange resin AG50 W-X12, 200 to 400 mesh), evaporated to dryness and a 500 ppb Mg-solution (in 3.5 % HNO 3 ) was prepared.The samples were analyzed with a Thermo Fisher Scientific Neptune MC-ICP-MS using DSM3 standard solution.The external precision was determined by measuring the mono-elemental solution Cambridge 1 repeatedly (n = 89, δ 25 Mg: −1.34 ± 0.01 ‰ and δ 26 Mg: −2.58 ± 0.03 ‰).For details of the analytical procedure refer to Immenhauser et al. (2010).
The stalagmite AH-1 (Atta Cave) provided material that was tested for the impact of fluid inclusions in the calcite fabric on the bulk δ 26 Mg.The sample was taken because it contains numerous fluid inclusions.Calcite samples from this stalagmite were extracted and half of each sample grounded in acetone to remove the fluid inclusions and subsequently analysed for their magnesium isotope signature.Aliquots of non-treated samples were analysed too and results compared.2  and 3).Stalagmite AH-1 yielded overall more depleted values between 9.0 and 6.0 ka BP and less depleted values between 6.0 and 0.7 ka BP.Between 6.0 and 4.2 ka BP, a trend to lower δ 26 Mg values is observed and between 9.0 to 6.0 ka BP and 4.2 to 0.7 ka BP, a weak trend to 26 Mg-enriched values and a higher overall variability is found.The lowest δ 26 Mg value is found at 8.7 ka BP.An enriched δ 26 Mg value at 7.9 ka BP coincides with a detritus layer.Between 4.2 to 3.5 ka BP, a trend to higher values is found.At 3.2 ka BP, the δ 26 Mg value becomes slightly depleted whilst subsequently the value is again more enriched at 2.5 ka BP.Between 2.5 and 1.6 ka BP δ 26 Mg values decrease again and then increase between 1.6 and 1.5 ka BP.At 1.5 ka BP the least   2 and     3).Between 8.  2 and 3).Magnesium isotope values in speleothem NC-B (13.1 to 3.6 ka) are invariant with fluctuations mostly within the analytical error.Speleothem NC-A δ 26 Mg value is slightly enriched at 2.8 ka BP and at 1.9 ka BP the value is 26 Mg-depleted.The last point at 0.06 ka BP is again 26 Mg-enriched.These shifts, however, barely exceed the error bars of NC-B.

Conclusions References
Tables Figures

Back Close
Full the University of Berne, Switzerland), therefore error bars are large.From the base of the stalagmite GDA δ 26 Mg values show more depleted values in the lower third of the stalagmite; i.e. between 13.6 and 9.7 cm distance from the top (dft) of the stalagmite.At 8.9 cm dft, i.e. the middle part, a more 26 Mg-enriched value is recorded reaching plateau values between 8.9 and 7.7 cm dft.The magnesium isotope value is again more enriched at 6.3 dft and a second plateau phase is reached between 6.3 and 3.1 cm dft, whilst δ 26 Mg is again depleted at 2.3 cm dft.

Stalagmite HK3, Grotte Prison de Chien, Morocco
The Pleistocene to Holocene stalagmite HK3 consists of several primary aragonitic and calcitic layers reflecting alternatingly more arid (aragonite) and less arid (calcite) climate conditions.Data from samples across two transitions from aragonite to calcite and from calcite to aragonite are used and were described in detail in Wassenburg et al. (2012).The transitions are located between two age points -27.48 and 23.53 ka BP (Wassenburg et al., 2012).Samples from the low-Mg aragonite layers were not analysed as the Mg content of aragonite is too low.The calcite layer yields δ 26 Mgvalues between −4.29 ± 0.02 ‰ and −3.89 ± 0.06 ‰ (mean: −4.17 ± 0.05 ‰; Figs. 2 and 8a, Tables 2 and 3).Overall, δ 26 Mg calcite increase towards the overlying aragonite intervals, whilst isotope ratios remain invariant in the central portions of the calcite interval (Fig. 8a).  2 and 3), respectively.The highest δ 26 Mg value for SPA 52 is located at 0.7 cm distance from base (dfb) in a detrital layer followed by a depleted δ 26 Mg value at 7.0 cm dfb (Fig. 9a).Between 7.0 and 9.1 cm dfb the values slightly increase again.At 10.3 cm dfb the δ Introduction

Conclusions References
Tables Figures

Back Close
Full is slightly depleted.The sample at 13.2 cm dfb is higher in its δ 26 Mg and is located in a calcite layer containing slight amounts of detritus.After a hiatus of ca.53 ka the δ 26 Mg values are again higher (at 14.4 cm dfb) and are again lower at 16.7 cm dfb (Fig. 9a).These two sample points coincide with detrital layers.Between 25.4 to 33.6 mm distance from base (dfb) the δ 26 Mg values of SPA 59 are low before the δ 26 Mg is higher at 41.1 mm dfb (Fig. 10a).This sample was taken at a hiatus with detrital material at the top.At 56.4 mm dfb the value is 26 Mg-depleted before the highest value is reached at 60.3 mm dfb (δ 26 Mg: −2.47 ‰; Fig. 10a).Samples taken at 47.5 and 60.3 mm dfb contain detrital material.Afterwards, the δ 26 Mg values are again depleted 65.7 mm dfb and show a decreasing trend to 80.5 mm dfb.A slight increase occurs between 80.5 and 94.2 mm dfb followed by a decrease to 110.9 mm dfb (Fig. 10a).

Interpretation and discussion
The study sites are grouped into the following present-day climate regions (i) warm temperate and semi-arid (dry and hot summers) to arid (desert and hot arid) climate in Morocco; (ii) warm temperate (warm summers) and humid climate in Western Germany; (iii) equatorial-humid climate in Peru and (iv) alpine climate (humid, snow-rich winters and cold summers) in Austria (Kottek et al., 2006).Given that most of the following factors directly relate to climate, the discussion of Mg-isotope data is organized according to climate zones.compared to carbonate weathering, which is always dominant but particularly so under more humid and cooler conditions (Buhl et al., 2007;Immenhauser et al., 2010)  their high Mg-content and higher δ 26 Mg values (Riechelmann et al., 2012b).Where dolostones or dolomite-rich limestones form the cave hostrock, the impact of silicatederived 26 Mg on water isotope ratios will be subdued (Galy et al., 2002) due to the Mg-rich hostrock mineralogy.The Moroccan stalagmite HK3, characterized by alternating aragonitic and calcitic intervals, documents the above relationships (Fig. 8).Due to the low abundance of Mg in aragonite (Okrusch and Matthes, 2005), sampling was restricted to the calcite intervals.Based on high-resolution carbon and oxygen isotope data as well as elemental abundances, Wassenburg et al. (2012) documented that during more arid periods, aragonite precipitation was favoured whilst calcite layers typify less arid conditions.During more arid phases, two main factors trigger aragonite precipitation.These include (i) the increased residence time of water in the soil as well as in the karst system, where 26 Mg enriched clays are leached, and (ii) prior calcite precipitation, which takes place in the karst aquifer zone (i.e. the precipitation of calcite prior to reaching the stalagmite surface), leading to a higher Mg-content of the fluid.The fact that the hostrock contains dolomite (Table 1) results in an enriched Mg-content of the drip water, too.In addition, prior calcite precipitation also increases the pH of the fluid and lowers the fluid saturation state.These factors give rise to aragonite precipitation (Frisia and Borsato, 2010;Wassenburg et al., 2012).We used the stalagmite HK3 Mg-isotope data to test the above-described impact of increasing aridity and soil water residence time on the Mg-isotope signature.In essence, rainfall amount predictably decreases towards the calcite-to-aragonite transition (27.48 to 23.53 ka BP), whilst soil water residence time and silicate weathering rates in the aquifer should increase.Under increased input of isotopically more positive, clay-derived Mg a shift towards higher values near the calcite-to-aragonite transition is expected.Indeed, this shift in isotope ratios in the calcite layer near the transition to aragonite (∆ 26 Mg=0.36 ‰) is observed and documented in Fig. 8a.Carbon, oxygen and magnesium isotopes show a highly similar trend in the calcite layer (Fig. 8).However, it should be noticed that near the calcite-to-aragonite transition the possibility of Introduction

Conclusions References
Tables Figures

Back Close
Full a mixing of predominantly calcite with small amounts of aragonite is found.Since the fractionation of carbon and oxygen isotopes is different for aragonite and calcite (Romanek et al., 1992;Kim and O'Neil, 1997), it may be possible that different fractionation factors also apply for Mg.Therefore, changes in the Mg-isotope composition near the calcite-to-aragonite transition may also be induced by a change of the fractionation factor.
The Mg-isotope record of the calcitic stalagmite GDA from Grotte d'Aoufous is equally well understood in the context of aridity changes.The Mg-isotope data (Fig. 7a) were thought to reflect changes between more or less arid conditions, which were in accordance with δ 18 O, δ 13 C, Sr and Mg elemental data and Sr isotopes (Buhl et al., 2007).Similar to the Moroccan stalagmite HK3, the δ 26 Mg values of the GDA show a similar pattern as the δ 13 C and δ 18 O values (Fig. 7).Problems arise because stalagmite GDA is considerably older (U-Pb age of 2.134 ± 0.115 Ma) and speleothem growth in the Grotte d'Aoufous ceased several times in the Pleistocene (Buhl et al., 2007).Climate reconstructions from Larrasoa ña et al. ( 2003 , 2008).This overall pattern is documented in the δ 18 O calcite data that decrease during the Holocene due to a gradual increase of rainfall amount (van Breukelen et al., 2008).It seems likely, that water residence times were short due to the high and constantly high rainfall amount at this site.Therefore, no changes in the Mg-isotope values should be expected.
In contrast to the Moroccan case examples, the Peruvian stalagmites NC-A and NC-B lack a correlation between δ 26 Mg and δ 13 C and δ 18 O ratios, respectively (Fig. 6).
Applying the concepts laid out above, the limited changes in the (short) soil water Introduction

Conclusions References
Tables Figures

Back Close
Full residence and a rather constant carbonate versus silicate weathering ratio throughout the Late Pleistocene to Late Holocene -the soil is mainly composed of kaolinite and Mg-bearing montmorillonite (Sanchez and Buol, 1974) -resulted in invariant δ 26 Mg ratios in stalagmite NC-B (Fig. 6a).In contrast, the Late Holocene to recent NC-A speleothem recorded subtle changes in δ 26 Mg calcite barely exceeding the analytical error that are not reflected in the C or O isotope ratios.Considering the limitations of the low temporal resolution of δ 26 Mg sampling points and the invariant Mg-isotope signatures, it is here proposed that the Peruvian stalagmites represent an end-member case setting that is less than suitable for Mg isotope proxy data.

Warm-humid climate: speleothem time series δ 26 Mg data from Germany
Figure 4 documents all δ 26 Mg values of stalagmite AH-1 from samples that were crushed prior to analysis and that are isotopically depleted relative to aliquots of the same sample that have not been treated in this manner.The trend in Mg-isotopes with time is the same for the crushed samples and the bulk samples (Fig. 4).From this it is concluded, that the δ 26 Mg of the fluid contained in these inclusions shifts the bulk calcite Mg-isotope composition by 0.01 to 0.08 ‰ towards higher values.This is because the fluid representing the drip water chemistry -enriched in δ 26 Mg relative to the calcite (Galy et al., 2002;Young and Galy, 2004;Immenhauser et al., 2010;Riechelmann et al., 2012) -is influenced by the same factors as the calcite that has precipitated from this fluid (Riechelmann et al., 2012b).drip rate of the AH-1 drip site is ten times faster than that of the BU 4 drip site probably suggests that the water feeding the AH-1 drip site (Atta Cave) has a shorter water residence time compared to that of the BU 4 drip site (Bunker Cave).This assumption is supported by field evidence given that speleothem BU 4 is characterized by a lower mean δ 26 Mg value (−4.20 ± 0.10 ‰; Fig. 2 and Table 3) compared to speleothem AH-1 (−4.01 ± 0.07 ‰, Fig. 2 and Table 3).This because the aquifer water in Bunker Cave has a longer reaction time with the isotopically depleted hostrock in the karst zone.Nevertheless, differences in drip water chemistry might also be the reason for the differences in mean Mg-isotope composition of 0.19 ‰ between these two caves.Drip water samples taken above AH-1 is considerably enriched in Mg (Mg/Ca ratio: 0.67; Niggemann, 2000) relative to drip water samples taken above BU 4 (Mg/Ca ratio: 0.06).Higher Mg concentrations in the aquifer may be due to dolomitisation of the limestone along fissures in the hostrock (Krebs, 1978).Water feeding the drip site of AH-1 is possibly percolating along these dolomitized fissures.This seems not to be the case for the drip water of BU 4. Given that dolomite is generally elevated in δ 26 Mg relative to limestone (Li et al., 2012), higher mean δ 26 Mg value of AH-1 drip water are expected and found.
The portion of stalagmite AH-1 representing the time interval between 9.0 to 6.0 ka BP is characterized by a dark and compact crystal fabric.Calcite precipitated between 6 ka BP and Present is typified by a white and porous crystal fabric.Lower δ 26 Mg values correspond with the dark and compact crystal fabric and 26 Mg-enriched values correspond with the white and porous crystal fabric.
In AH-1, carbon and oxygen isotope values (Fig. 3b) are enriched in the dark, dense fabric (9.0 to 6.0 ka BP), a fact that was interpreted as evidence for overall drier climate conditions (Niggemann et al., 2003).As argued in Riechelmann et al. (2012b), drier Introduction

Conclusions References
Tables Figures

Back Close
Full to 26 Mg-depleted values as expected for dominant carbonate weathering under wetter conditions.Based on the δ 18 O record of stalagmite AH-1, Mangini et al. (2007) proposed an overall climatic warming for the last 6 ka BP.
The following lines of evidence (Riechelmann et al., 2012b) suggest that under an overall humid climate and increasing mean air temperature microbial soil-zone activity is favoured (Harper et al., 2005).Related to this, the weathering of Mg-bearing clay minerals such as montmorillonite, chlorite and illite in the soil zone is increased.The general shift in δ 26 Mg AH-1 towards higher values may bear witness of an overall increased influx of clay-derived 26 Mg.If these considerations hold true, the general trend in AH-1 δ 26 Mg is mainly outside air temperature controlled.Between 9 to 6 ka BP carbonate weathering increased relative to clay mineral weathering in the soil zone due to colder temperatures.Between 6 ka BP to Present, a renewed increase in outside air temperature favoured silicate weathering in the soil zone (Figs. 3a and 11).
Using oxygen isotope data of AH-1 (Niggemann, 2000;Niggemann et al., 2003) as a tentative benchmark against which the smaller variations in δ 26 Mg are calibrated, the following pattern emerges: Shifts towards higher δ 26 Mg values took place when mean rainfall amounts decreased whilst shifts towards lower δ 26 Mg values coincided with overall more humid climate (Figs. 3 and 11).
The stalagmite BU 4 consists of a columnar fabric with the exception of a thin interval dated as 7.6 ka BP that is characterized by dendritic crystals (Riechelmann, 2010).It is of interest to note that δ 26 Mg ratios shift to higher values at this level (Fig. 5a) but the relation between crystallography and Mg isotope ratios is still underexplored.Columnar crystal fabrics are formed under stable precipitation conditions, while dendritic fabric forms preferentially under conditions far from equilibrium (Frisia et al., 2000;Frisia and Borsato, 2010).
Following Fohlmeister et al. (2012), δ 18 O data of BU 4 recorded changes in both temperature and rainfall amount -at higher values the climate was cooler and drier and vice versa (Fig. 5b).The first order signal of δ 13 C is influenced by vegetation cover above the cave while the second order signal is influenced by drip rate (Fohlmeister Introduction

Conclusions References
Tables Figures

Back Close
Full  , 2012).Between 9 to 7 ka BP soil respiration was lower leading to higher δ 13 C values.After 7 ka BP the δ 13 C signal decreases, which is due to a denser vegetation cover and thus higher soil respiration (Fohlmeister et al., 2012).Periods of lower drip rate lead to higher δ 13 C values and vice versa (Fohlmeister et al., 2012).
Between 8.2 and 6.0 ka BP, the δ 13 C and δ 18 O data of BU 4 indicate a trend to drier and cooler climate with less soil respiration (Fig. 5b), whilst the Mg-isotope values decrease (Fig. 5a).Following our previous set of arguments, this points to a shift towards carbonate weathering with temperature overruling rainfall amount as the driving factor (Fig. 11).From 5.5 to 1.8 ka BP, δ 13 C and δ 26 Mg values both show a decreasing trend (Fig. 5).Soil respiration expectedly increased during this time due to the development of a denser vegetation cover (Fohlmeister et al., 2012).Hence, the production of CO 2 is increased and leads to higher dissolution rates of the low-Mg calcite hostrock due to the increased amount of carbonic acid in the karst system.Also the decalcification of the loess (at present a loess loam) may have lead to decreasing δ 26 Mg values.
Therefore, decreasing δ 26 Mg values may be due to enhanced dissolution of carbonate between 5.5 to 1.8 ka BP.The processes might also apply between 5.7 and 4.2 ka BP for speleothem AH-1.Here, a decreasing trend in the δ 26 Mg is noted (Fig. 3a), whilst δ 13 C values decrease but to a lesser degree (Fig. 3b).The direct comparison of the Bunker Cave stalagmites δ 18 O record with that of the Atta Cave stalagmite AH-1 reveals similar patterns (Fohlmeister et al., 2012).This observation is considered evidence that stalagmites in both caves recorded essentially the same environmental signals.
BU 4 magnesium-isotope ratios between 1.6 and 0.7 ka BP anti-correlate with BU 4 δ 13 C and δ 18 O data (Fig. 5).As a consequence, it seems that temperature is again the factor dominating soil processes and linked to this, weathering ratios (Fig. 11).Between 1.1 to 0.7 ka, the Medieval Warm Period (Fohlmeister et al., 2012) is characterized by lower δ 13 C and δ 18 O and higher δ 26 Mg values (Fig. 5), perhaps explained by an overall Introduction

Conclusions References
Tables Figures

Back Close
Full wetter and warmer climate (Esper et al., 2002;Soon et al., 2003;Negendank, 2004;Mann et al., 2009).The Little Ice Age, dated between 0.7 to 0.2 ka, is characterized by a shift to higher values in δ 13 C and δ 18 O and lower δ 26 Mg ratios (Fig. 5; Fohlmeister et al., 2012) indicating overall drier and cooler climate (Esper et al., 2002;Soon et al., 2003;Negendank, 2004;Mann et al., 2009).Published mean air temperature differences from Western Germany between Little Ice Age and Medieval Warm Period are on the order of 2 to 2.5 • C (Negendank, 2004).Given the low dependency of Mg isotope fractionation on temperature during calcite precipitation (0.02 ‰/AMU/ • C; Galy et al., 2002), the observed isotope shifts would require cave temperature changes in excess of 10 • C.
Evidence for a temperature shift in this order of magnitude is lacking and temperature is thus ruled out as a main factor driving Mg isotope fractionation during calcite precipitation.
High δ 26 Mg values in BU 4 coincide with a hiatus at 5.5 ka BP containing detrital material (Fig. 5a) as does one sample taken from stalagmite AH-1 at 7.9 ka BP (Fig. 3a).
Besides quartz, feldspar and mica, detrital layers contain high amounts of clay minerals that are in part dissolved in 6M HCl especially during heating (Madejov á et al., 1998;Komadel and Madejov á, 2006).It seems thus likely that clay mineral-derived, isotopically heavy magnesium is analysed in parallel with the calcite-derived magnesium (Figs.3a and 5a).In essence, the two cave records from Germany exemplify the full complexity of speleothem δ 26 Mg records in a setting characterized by comparably subtle climate changes over time.Based on the arguments brought forward here, it is suggested that weathering patterns in the soil zone above the caves as well as in the karst zone, driven by mean outside air temperature and rainfall amount, are the main driving factors in the δ 26 Mg records obtained (Fig. 11).In principle, all of these factors are present at all times but temperature seems to overrule rainfall amount in general.Introduction

Conclusions References
Tables Figures

Back Close
Full Speleothem deposition in the high-elevation Spannagel Cave is biased towards warm interglacial conditions.Cold-climate precipitation, however, has also been observed and attributed to the presence of sulphides in the hostrock (providing acidity for karstification) and a warm-based ice cover (Sp ötl et al., 2007;Sp ötl and Mangini, 2007) During interglacials the availability of water (rain, snow and ice melt) is higher than during glacial periods (Sp ötl et al., 2007).Hence, changes in rainfall amount that in turn affect soil activity can be excluded as main factor controlling Mg-isotope fractionation, because speleothems can not form without water availability.Having ruled out water availability as a controlling factor, temperature remains the main parameter controlling subtle changes in silicate versus carbonate weathering ratios (Fig. 11).This probably applies for the Holocene with warm summer months and cold winter months (7 to 8 months of snow cover per year; Sp ötl et al., 2002).
Flowstones yield numerous detrital layers, which were sampled (red triangles in Figs. 9 and 10).As found in the case of BU 4 and AH-1, Mg-bearing clay minerals Introduction

Conclusions References
Tables Figures

Back Close
Full were also partly dissolved during sample preparation and shift the Mg-isotope ratios towards higher values.Sample SPA 52 f (δ 26 Mg: −1.42 ± 0.04 ‰;  2) and therefore closer to the mean isotopic composition of SPA 52 relative to SPA 59, a feature that is in good agreement with a short hostrock water residence time.
In conclusion, the Spannagel flowstones represent a cold, alpine endmember setting characterized by mainly temperature-driven weathering changes in the thin soil cover -where present -above the cave (Fig. 11).Warmer periods favour more soil-zone activity and increased silicate weathering leading to enriched δ 26 Mg.

Potential, problems and future work
Data shown here suggest that a limited, albeit complicated bundle of parameters controls the Mg isotope composition of speleothems.Here, we show data from a series In the cold, alpine end member, outside air temperature seems the dominant driver of carbonate-silicate weathering patterns driving the δ 26 Mg variations.
One of the obvious strengths of the Mg isotope proxy is its sensitivity for subtle processes in the soil zone (weathering ratios; Buhl et al., 2007) and differences in soil and reservoir water residence time (Riechelmann et al., 2012b).Combined with elemental evidence and other isotope systems such as Sr, C, and O, an improved understanding of soil zone processes can be achieved.Nevertheless, despite significant advances in the application of Mg-isotopes as a continental climate proxy, unsolved issues remain.Due to the low Mg content of many speleothems, the sample size required for δ 26 Mg analysis is large and -combined with the time-consuming and expensive analytical work -inhibit high-resolution speleothem Mg isotope records similar to C or O isotopes.Specifically, aragonite speleothems are low in Mg so that an interpretation of their δ 26 Mg records seems not feasible at present.Therefore, comparisons between high-resolution C, O and elemental data records with δ 26 Mg data in speleothems are difficult at present.The sampling of speleothem calcite containing detrital material should be avoided due to contamination problems.The dependency of the δ 26 Mg on the calcite precipitation rate is known from laboratory precipitation experiments (Immenhauser et al., 2010), but awaits further work using water compositions and precipitation rates that are more similar to those found in nature, since there was no influence on δ 26 Mg in natural samples observed.
Finally, an improved understanding of the factors that control Mg leaching from soil zone clay minerals and clay contained in hostrock carbonates is required.Carefully established leaching experiments under controlled laboratory conditions might represent the way forward.

Conclusions References
Tables Figures

Back Close
Full

Conclusions
The analysis of Mg as well as C and O isotope ratios from eight speleothems representing five climate zones leads to the following conclusions: 1. Germany are more difficult to interpret.This is because factors interfere in a more complex manner.These factors include mean air temperature outside the cave and rainfall amount.These factors drive soil-zone activity and affect the subtle balance between silicate and carbonate weathering rates.Here, multi-proxy C and O data must be used in combination with δ 26 Mg ratios in order to obtain a coherent interpretation and to separate the relative effects of mean air temperature and rainfall amount on soil zone activity.
3. Layers of detrital material in speleothems must be avoided due to contamination problems.Particularly, Mg-bearing clay minerals shift bulk values to higher δ 26 Mg ratios.
4. In high-latitude and alpine climate settings exposed to pronounced changes in glacial and interglacial air temperature, the main driving factor of soil-zone activity and weathering ratios outside of the cave is temperature controlling weathering ratios and hence speleothem δ 26 Mg values.

CPD Introduction Conclusions References
Tables Figures

Back Close
Full  Full  Full Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | and oxygen isotope time series data as well as U-Th age data shown in Figs. 3 and 5-10 are published in Niggemann et al. (2003), Holzk ämper et al. (2005), Buhl et al. (2007), Sp ötl et al. (2007), van Breukelen et al. (2008), van Breukelen (2009), Wassenburg et al. (2012) and Fohlmeister et al. (2012).Time series stalagmite and flowstone Mg isotope data are shown in Tables 2 and 3 and Fig. 2. Below, the main geochemical characteristics are summarized for each speleothem analyzed.Moroccan stalagmites GDA and HK3 and stalagmite BU 4 from Germany yielded the most depleted mean values, whilst flowstones SPA 52 and SPA 59 from Austria Discussion Paper | Discussion Paper | Discussion Paper | are least depleted in terms of their mean δ 26 Mg signatures.Stalagmite AH-1 from Germany and the Peruvian stalagmites NC-A and NC-B yielded intermediate mean δ 26 Mg values (Fig. 2).The flowstones SPA 52 and SPA 59 show the highest variations in their Mg-isotope composition.In contrast, stalagmite δ 26 Mg is less variable.Stalagmite HK3 (Morocco) and stalagmite BU 4 (Germany) show the highest amplitudes in δ 26 Mg, whilst stalagmites NC-A and NC-B (Peru) reveal the lowest degree of variability and therefore the smallest ∆ 26 Mg.Stalagmites AH-1 (Germany) and GDA (Morocco) show intermediate amplitudes (Fig. 2).4.1 Stalagmite AH-1, Atta Cave, Germany Holocene to recent (1997 AD) stalagmite AH-1 δ 26 Mg values range from −4.18 ± 0.05 ‰ to −3.86 ± 0.03 ‰ with a mean of −4.01 ± 0.07 ‰ (Figs. 2 and 3a, Tables depleted δ 26 Mg ratio is present.At 1.2 ka BP, δ 26 Mg show again a more depleted value whilst the most recent data point, dated at 0.7 ka BP is less depleted.The results from the fluid inclusion test runs are presented in Table 4 and Fig. 4. The ∆ 26 Mg between bulk (with intact fluid inclusions) and grounded samples (without fluid inclusions) is not exceeding the second decimal.Nevertheless, grounded samples are systematically lower by 0.01 to 0.08 ‰ in their Mg-isotope composition relative to their aliquots containing fluid inclusions (Fig. 4).Error bars of all these samples, Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | SPA 52 and SPA 59, Spannagel Cave, Austria The Pleistocene flowstones SPA 52 and SPA 59 δ 26 Mg values show ranges from −3.63 ± 0.05 ‰ to −1.42 ± 0.04 ‰ (mean −3.00 ± 0.73 ‰) and from −4.06 ± 0.04 ‰ to −2.47 ± 0.05 ‰ (mean: −3.70 ± 0.43 ‰; Figs. 2, 9a and 10a, Tables Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

5. 1
Environmental, equilibrium and non-equilibrium factors affecting the δ 26 Mg-isotope fractionation in continental karst systemsEnvironmental, equilibrium and non-equilibrium factors affecting Mg isotope fractionation in Earth surface systems have been described inGaly et al. (2002),Young and Galy (2004), de Villiers et al. (2005), Tipper et al. (2006b, 2008, 2010), Discussion Paper | Discussion Paper | Discussion Paper | et al. (2007),Brenot et al. (2008), Immenhauser et al. (2010), Pokrovsky et al. (2011)   andRiechelmann et al. (2012b).Considering the published work, the following parameters are of significance for the interpretation of speleothem time-series δ 26 Mg isotope data: (i) δ 26 Mg signature of all involved Mg sources -rain water, soil cover (silicates) and hostrock (limestone or dolostone); (ii) the biogenic activity of the soil zone controlled mainly by water availability and air temperature; (iii) the balance between the silicate (Mg-bearing clay minerals) and carbonate weathering ratio in the soil zone, with clays being 26 Mg-enriched and diagenetically stabilized low Mg-calcite limestones being 26 Mg-depleted; (iv) the residence time of soil and karst water in the aquifer above the cave and (v) carbonate precipitation rates in the cave environment.In contrast, even when considering the maximum range of Pleistocene to recent air temperature change in caves in Germany (ca. 9 to 10 • C; Richter et al., 2011), temperature-related δ 26 Mg fractionation during calcite precipitation is a minor factor with experimental data suggesting a δ 26 Mg dependency of 0.011 ‰ per • C from 4 to 45 • C (Li et al., 2012) and 0.02 ‰/AMU/ • C between 4 and 18 • C (Galy et al., 2002).Rain water is 26 Mg-enriched with increasing distance to the marine aerosol source and may reveal temperature-related fractionation patterns, i.e. snow is26 Mg depleted compared to rain(Tipper et al., 2010;Riechelmann et al., 2012b) but the existing data set is perhaps too limited to draw solid conclusions.The δ 26 Mg ratio of dolostone (−2.2 to −1.1 ‰; Li et al., 2012) is in general higher than that of limestone (−5.3 to −1 ‰; Li et al., 2012), whilst silicate rocks and clay minerals have the most 26 Mg-enriched composition (−1.1 to +0.1 ‰; Immenhauser et al., 2010).Biological Mg fractionation by plants and soil microorganisms affect the Mg-isotope composition of the soil water in variable degrees depending on the amount of Mg involved and plant and microbial metabolisms (Tipper et al., 2010; Riechelmann et al., 2012b).Clearly, under warmer and wetter conditions soil activity is increased (Harper et al., 2005) and enhanced plant growth induces a stronger biological fractionation.Grass, for example, is 26 Mg-enriched in comparison with its rain water source (Tipper et al., 2010).Silicate (Mg-bearing clay minerals) weathering is relatively increasing during warmer and drier climate conditions Introduction Discussion Paper | Discussion Paper | Discussion Paper | well as biogenic soil activity are considered significant(Tipper et al., 2006a(Tipper et al.,  , 2010;;Li et al., 2010).Related to these factors, changes in the soil clay mineral to host limestone weathering balance affect the soil and aquifer water δ 26 Mg value.Following the above discussion, a relative increase in the weathering of soil clay minerals induces higher δ 26 Mg values.Under most hydrogeochemical conditions, the limestone and dolostone solubility in soil water and aquifer waters exceeds that of clay minerals (Appelo and Postma, 2005).Nevertheless, even a slightly increased solubility of Mg-bearing clay minerals will be recorded in soil-zone and karst aquifer waters due to Introduction Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | ) andFeddi et al. (2011), however, support the interpretation of stalagmite GDA fromBuhl et al. (2007).For the Gelasian (1.806-2.588Ma), i.e. the time interval that coincides with the U-Pb age of stalagmiteGDA, Larrasoa ña et al. (2003)  found changes in the dust amount, interpreted as changes in the amount of rainfall that coincides with changes of the 400-kyr component of the eccentricity.According to these authors, changes between more arid and less arid conditions occur during the Gelasian.Similarly,Feddi et al. (2011) documented evidence for temperature changes in the Gelasian from pollen analyses.These general patterns are arguably also recorded in stalagmite GDA, which is therefore showing changes between more arid and warm and less arid and warm conditions in the Mg-isotope composition.The hostrock δ 26 Mg of Grotte d'Aoufous (−3.63 ± 0.08 ‰; Fig.2;Buhl et al., 2007), a Turonian marine limestone, is slightly higher than that of the Devonian hostrock (−3.72±0.07‰; Fig.2; Immenhauser et al., 2010) values of the caves in Germany discussed below.Aquifer water residence time under the arid conditions that characterize Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | The interpretation of speleothem δ 26 Mg time series data from Atta and Bunker Caves is challenging.Given the very comparable climate setting, hostrock composition (δ 26 Mg: −3.72 ± 0.07 ‰; Fig. 2; Immenhauser et al., 2010) and soil properties, we suggest that the same set of factors drive Mg fractionation in these two caves.Differences between the two sites include drip characteristics as well as drip water chemistry.Both drip sites show only slight differences in their drip rate (AH-1: 0.019 ± 0.001 ml min −1 ; Niggemann, 2000; BU 4: 0.002±0.001mlmin −1 ) but the same drip characteristic (seepage flow) using the terminology of Baker et al. (1997).The fact, that the Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | et al.
Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | 5.5 Cold-humid climate: speleothem time series δ 26 Mg data from Austria Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | of climatic zones.Obviously, caves situated in climate domains characterized by pronounced changes in mean air temperature or rainwater availability are expectedly characterized by more pronounced shifts (Fig. 11).A prominent example is found in the data set from Morocco.Conversely, speleothem records characterized by rather constant climate parameters such as the Peruvian data, show no variability in δ 26 Mg.The interpretation of the δ 26 Mg records from caves in Germany are complex.This is because several of the factors, namely temperature and rainfall amount, control speleothem δ 26 Mg in a manner that differs from the Moroccan stalagmites δ 26 Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | 5. Comparing hostrock and mean δ 26 Mg stalagmite signatures, semi-quantitative estimates of aquifer water residence time can be obtained.Depleted mean stalagmite δ 26 Mg values are found where hostrock values are low and long residence time resulted in equilibration of δ 26 Mg fluid with δ 26 Mg hostrock .6.The results shown here summarize the present level of knowledge in speleothem Mg isotope proxy research.An improved understanding of the complex processes involved requires well calibrated experimental work including leaching and precipitation experimentsDiscussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

Fig. 1 .
Fig. 1.Map with the locations of studied caves.Atta and Bunker Cave (Germany) are represented by a single green dot.

arid climate: speleothem time series δ 26 Mg data from Morocco Factors
. Longer water residence times in the karst aquifer lead to lower δ 26 Mg values in the drip waters, because the mixing time between younger, newly infiltrating 26 Mg-enriched water and older 26 Mg-depleted aquifer water is longer and hence leading to a bigger portion of the 26 Mg-depleted water.For shorter water residence times it is the opposite (Riechelmann et al., 2012b).Based on laboratory precipitation experiments (Immenhauser et al., 2010), lower carbonate precipitation rates lead to decreasing δ 26 Mg values and vice versa.This, however, refers to experimental fluids that are considerably more saturated than those commonly found in cave environments and to precipitation rates (50 to 630 µmol h −1 ) that exceed those of most natural speleothems (range of 0.0008 to 9.94 µmol h −1 /0.1 to 1140 µm yr −1 for AH-1, BU 4, NC-A, NC-B, SPA 52 and SPA 59).A statistically significant impact of experimental precipitation rate on calcite δ 26 Mg is found at rates exceeding 200 µmol h Immenhauser et al. (20Li et al., 2010;y of the precipitation experiment data shown inImmenhauser et al. (2010)to natural systems is at present unclear.5.2 Warm-(semi-)affecting weathering rates in carbonate-silicate settings are complex, spatially diverse and debated in the literature(Tipper et al., 2006a;Li et al., 2010; Pokrovsky  et al., 2011).Nevertheless, rain water abundance and residence time in the soil as the northern rim of the Sahara were expectedly longer than those in the warm-humid climate in Germany.Hence, lower δ 26 Mg values for the Moroccan stalagmite GDA could be expected.In essence, the concepts as explained above provide a comparably robust and internally consistent interpretation for the δ 26 Mg record of the two Moroccan speleothems shown here.Combined with evidence from Sr and Mg abundances and calcite 5.3 Equatorial-humid climate: speleothem time series δ 26 Mg data from PeruThe Holocene climate in Peru differs significantly from that of the Moroccan case examples.The present-day rainfall amount in the vicinity of the Cueva del Tigre Perdido is around 1344 mm yr −1 (Table1; Str äßer, 1999); and was probably in the order of 940-1140 mm yr −1 at the beginning of the Holocene (15-30 % increase of rainfall amount during the Holocene;van Breukelen et al., 2008), a value that indicates an overall very humid climate.Average annual temperatures seem to have remained rather constant through the Holocene (van Breukelen et al. . In the present-day setting, sparse alpine meadows cover a soil above the cave and soil thickness does not exceed 20 cm.Portions of the hostrock above the cave are bare, meaning not covered by soil or vegetation.Carbon isotope ratios of both studied flowstones (Figs.9b and 10b) are influenced by the composition of the marine marble hostrock (δ 13 C of +2 to +3 ‰; Holzk ämper et al., 2005) and the input of soil-derived organic C in the drip water.Oxygen isotope values are high during warm and low during cold climate conditions.Low δ 13 C values imply more biological activity in the soil zone and vice versa (Sp ötl et al., 2007; Holzk ämper et al., 2005).Decreasing δ 13 C values thus reflect a genuine vegetation signal during periods, when carbon is anti-correlated with oxygen (Sp ötl et al., 2007; Holzk ämper et al., 2005).Both flowstones show similar patterns in δ 26 Mg and δ 18 O (Figs. 9 and 10).Higher values during warm conditions may thus imply more silicate weathering and vice versa.
Table 2) and sample SPA 59 b (δ 26 Mg: −2.47 ± 0.05 ‰; Table 2) are taken from thick detrital layers with a high clay mineral content and are therefore highly 26 Mg-enriched.Given that these data points are not representing the calcite δ 26 Mg ratios, mean values should be cal- The sensitivity of speleothem δ 26 Mg values to environmental and kinetic factors strongly depends on the overall climate setting.Pronounced variations in mean air temperature and rainfall amount with time, such as the case in the Pleistocene of Morocco, result in changes in the silicate versus carbonate weathering ratio that in turn affects drip water δ 26 Mg in a predictable manner.Caves situated in invariant climate domains, such as tropical Peru during the Pleistocene to Holocene, are characterized by invariant δ 26 Mg records.2. Stalagmite δ 26 Mg values from warm-temperate climate zones such as Western

Table 4 .
Magnesium isotope composition of several samples of the stalagmite AH-1 measured with and without fluid inclusions and the calculated ∆ 26 Mg between these samples.