and temperature estimates (Schwarcz et al., 1976; Fleitmann et al., 2003; van Breukelen et al., 2008; Kluge et al., 2008; Affolter et al., 2015; Labuhn et al., 2015). The analysis of the tem- perature dependent preference for heavy nuclides to bond to each other, rather than to a lighter isotope is referred to as clumped isotope geochemistry (Ghosh et al., 2006; Eiler, 2007) is a promising technique, despite some complications of applying the system to speleothems (Affek et al., 2008; Daëron et al., 2011; Wainer et al., 2011; Kluge and Affek, 2012; Kluge et al., 2013).
8.1.4 Conclusions
In summary, the use of speleothem δ18O values in terms of paleo climate reconstruction is unchallenged. The fundamental problem of interpreting speleothem δ18O values is the discrimination of temperature and precipitation effects. In most cases, the analysis of δ18O and δ13Calone is not enough to derive satisfying results in terms of climate reconstruction.
In particular in tropical regions which are dominated by the amount effect, the oxygen iso- topic signature is regarded a reliable proxy of past hydrological variability, when taking into account second order effects, that may change the interpretation of δ18O
csignificantly, e.g. on glacial-interglacial timescales. However, this also presumes the knowledge of a quantitive estimate past temperature changes.
In the mid-latitudes, a deep understanding of the climatological framework is indispens- able, but still the controls of speleothem δ18Ovalues may remain obscure, because a variety of effects may compete each other. Independent information from ’paleo thermometers’ or hydrologically forced proxies constitute valuable opportunities to disentangle the climate information of the stable isotopic composition of the speleothem.
8.2 Magnesium and Strontium - tracers of hydrological
processes?
Magnesium and strontium are amongst the most commonly used elemental proxies in speleo- them research. Despite the existence of a number of processes influencing the incorporation of Mg and Sr (section 2.4), most published records are interpreted to be linked with rainfall variability when Mg/Ca and Sr/Ca covary (Hellstrom and McCulloch, 2000; McMillan et al., 2005; Cruz et al., 2007). However, two general mechanisms - limestone diagenesis and water- rock interactions - might be responsible for co-fractionating Sr and Mg in the dripwaters. In this thesis, the model of Sinclair et al. (2012) was used to make inferences about the processes governing apparent co-variations in Mg/Ca and Sr/Ca.
8.2.1 Results of the case studies
In Chapter 5, Mg/Ca and Sr/Ca ratios were found to be generally positively correlated in speleothem PR-LA-1, however, little variability is observed throughout the record. Monitoring of dripwaters in Cueva Larga did not provide evidence for the existence of strong prior calcite precipitation (PCP) (Vieten et al., in press).
Applying the model of Sinclair et al. (2012) to the elemental ratios of speleothem calcite confirms, that the influence of calcite-water interactions such as PCP or ICD was comparably small in the main part of the record. Temperature as well as growth rate variability were also found to be potential drivers of second order-effects in Mg/Ca and Sr/Ca, respectively, as well as the influence of sea spray or detrital contamination could not be excluded. The only interval showing a notable signature of prior calcite precipitation (PCP) and/or (ICD) and being thus mainly controlled by calcite-water interactions is the last section between 15 and 16.5 ka indicating drier conditions during Heinrich stadial 1. For the earlier part of the record, additional evidence from other trace elements, such as P/Ca, or stable isotope δ18Oand δ13C values, was necessary to reconstruct climate conditions at the surface above the cave.
A similar picture is observed in stalagmite CM (Chapter 6). With 82 - 7 ka, the record spans a longer time frame than PR-LA-1 (Chapter 5). Mg/Ca and Sr/Ca are generally positively correlated, however, the slopes of the ln(Sr/Ca) vs ln(Mg/Ca) plots are systematically lower than expected for PCP/ICD. This is interpreted to be caused by an additional Mg source along the pathway of the dripwater through the host rock, e.g. high Mg-calcite or dolomite (section 6.4.1). However, the partly very variable slopes of the logarithmic plots could also be caused by calcite recrystallization, which is in general a highly non-linear process (Sinclair et al., 2012). Despite this is still a process determined by the amount of water percolating through the host rock, it is therefore not possible to make inferences from the absolute variations to the actual amount of rainfall above the cave.
Moreover, in CM, the incorporation of Sr and Mg in speleothem calcite may also be ob- scured by temperature and growth rate effects, variable soil leaching or sea spray (section 6.4.1). However, the generally positive correlation of Mg/Ca with δ18Oand δ13Cvalues sug- gests that high Mg/Ca values in CM most likely reflect a higher degree of calcite-water inter- actions through PCP, ICD and/or calcite recrystallization and thus generally drier conditions. In stalagmite C09-2 from southwestern Romania (Chapter 7), a different pattern in Mg/Ca and Sr/Ca systematics is noted. Both elements are correlated in the dripwater, but anti- correlated within speleothem calcite (section 7.4.1). The anti-correlation is based on the de- coupling of Mg and Sr signals due to seasonal growth rate effects. Thus, the model of Sinclair et al. (2012) is not applicable to the Sr and Mg values in the stalagmite. However, in the drip- water, the slopes of the logarithmic plots between both elements follows the theoretically derived values by Sinclair et al. (2012), indicating that PCP and/or ICD exerts a first order control prior to calcite precipitation on the stalagmite. This signal is subsequently modu- lated by the growth-rate dependency of the incorporation of Sr.
8.3 P - a proxy for vegetation and soil dynamics? The control on Mg/Ca as tracer for rainfall variability is further constrained by a strong correlation of annual mean Mg/Ca in the speleothem with winter precipitation in southwest- ern Romania through the 20th century. This correlation has proven to be robust against age uncertainties and autocorrelation effects and could even be used to deduce a quantitative reconstruction of precipitation estimates for the last ∼4000 years.
Conclusions
In summary, Mg/Ca and Sr/Ca have shown to be mainly controlled by hydrological processes in all records. For Cloşani, a direct link of Mg/Ca in stalagmite C09-2 with rainfall was found. The interpretation of the Romanian record clearly benefits from the direct comparison with instrumental data, which is not possible for the other studies. Here, the case is not that clear and the hydrological control on the Mg/Ca ratio remains ambiguous. Reconstruction of past hydrological conditions thus has to rely on the combination of several proxies such as δ18O values (section 8.1) to derive concrete conclusions.
However, care must be taken when interpreting trace element signals from samples which exhibit a variable calcite precipitation rate. If growth rate varies over more than an order of magnitude, it may significantly affect the Sr/Ca ratio of speleothems, which was shown to apply for all samples in this thesis to a greater or lesser extent. Also temperature variability may exert a non-negligible second-order effect on Mg/Ca. The same may apply at some loca- tions to detrital contamination, the effect of soil processes and sea spray as well as limestone diagenesis. These aspects have to be properly assessed and evaluated and, if quantitative es- timation of the processes is possible, compared to the present variability in the trace element signal.
The model of Sinclair et al. (2012) provides a helpful tool to reach a better understand- ing of the processes which may be responsible for the observed trace element variability in speleothem records. In remote cave locations, it is often not possible to establish a cave mon- itoring program, or the results may not be transferable to the analyzed sample. It clearly shows, that a correlation of Mg/Ca with Sr/Ca may not be necessarily linked with hydrolog- ical processes, and if so, they may not underly a linear relationship.