Simulated European stalagmite record and its relation to a quasi-decadal climate mode

Introduction Conclusions References


Introduction
Speleothems are a valuable archive of past climate variability since they allow precise dating (Richards and Dorale, 2003;Fairchild et al., 2006;Scholz and Hoffmann, 2008) and provide high-resolution climate proxy data. The most commonly used climate proxies in speleothems are stable carbon and oxygen isotope signals (δ 13 C and δ 18 O) (McDermott, 2004;Lachniet, 2009) as well as various trace elements such as 20 magnesium or strontium (Fairchild and Treble, 2009). Their potential for paleoclimate research is related to the question whether they reflect local climate conditions above the cave or large-scale climate variability modes. Such modes show coherent spatial structures and were identified both in the tropical Pacific (Philander, 1990) and the North Atlantic (Deser and Blackmon, 1993). Part of the problem of understand-

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | patterns (e.g. Rimbu et al., 2001;Felis et al., 2004;Lohmann et al., 2004;Langebroek et al., 2011). A climate-speleothem proxy relationship is postulated through a correspondence between a speleothem δ 18 O record and large-scale surface temperature and/or rainfall amount (e.g. Fairchild and Treble, 2009;Drysdale et al., 2009). Baker et al. (2011) analyze the climate-proxy relationship for a high-resolution speleothem δ 18 O 5 and found little correspondence to instrumental data, although a clear relationship between local rainfall δ 18 O and atmospheric circulation is observed.
Here, we follow their idea and trace the speleothem δ 18 O which stems from the composition of infiltrated water in a simulated cave. We analyze the climate variability pattern related to variations in a simulated cave system in Central Europe, which is under the influence of maritime climate. Climate over the North Atlantic sector varies on quasi-decadal to multi-decadal timescales (Deser and Blackmon, 1993;Hurrell, 1995;Sutton and Allen, 1997). In this pattern, the atmospheric circulation, similar to the North Atlantic Oscillation (NAO) (Walker, 1924), generates a tripole pattern in sea surface temperature (SST) anomalies (Bjerknes, 1964;Deser and Blackmon, 1993;Kushnir, 15 1994). Modeling studies with atmospheric general circulation models (AGCMs) of different complexity forced by global SST variability over the last century show that the atmospheric circulation over the North Atlantic is predictable if global SST variability can be predicted (Rodwell et al., 1999;Latif et al., 2000;Robertson et al., 2000;Sutton and Hodson, 2003;Grosfeld et al., 2007). For instance, North Atlantic SST can be 20 used as a predictor for the NAO pattern. However, it remains poorly understood how changing climatic boundary conditions affect the strength and dynamics of these natural oscillations in the North Atlantic realm on long time scales. Such information can be inferred from the past using climate proxy data.
Here, we elaborate the large-scale relation of the δ 18 O signal recorded in a sim-25 ulated stalagmite, for the location of Bunker Cave (51 • N, 7 • E). The cave is located in the Rhenish Slate Mountains in the western part of Germany (Riechelmann et al., 2011;Fohlmeister et al., 2012). Our model approach is based on an AGCM including water stable isotopes (Werner and Heimann, 2002) as well as a proxy model for the 3515 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | general processes influencing the δ 18 O signal of cave drip water and speleothem calcite (Wackerbarth et al., 2010). This model was developed in order to better understand the influence of climate change on the δ 18 O values of speleothem calcite. We examine the related large-scale variability on interannual to multi-decadal timescales in the North Atlantic realm. In addition, our approach helps to study the relationship between 5 climate change and the recorded speleothem proxy signals.

Atmospheric model
The applied model was the Hamburg AGCM ECHAM4 (Roeckner et al., 1996) with both water stable isotopes H 18 2 O and HDO explicitly cycled through the water cycle of 10 the model (Werner and Heimann, 2002). The simulation was performed in T30 resolution (3.75 by 3.75 spatial grid; 19 vertical levels). Observed monthly values of the global sea ice and sea surface temperature data set (GISST2.2) of the United Kingdom Meteorological Office were prescribed for the period 1908-1994 (Rayner et al., 1996). Atmospheric concentrations of greenhouse gases (CO 2 , CH 4 , N 2 O) but no additional 15 aerosol forcing were also prescribed according to the observations. With the same SST data and the same model, ensemble integration with three members were performed in order to study interannual to multi-decadal variability (Arpe et al., 2000;Grosfeld et al., 2007). Furthermore, the water isotope module has been applied for recent and past interglacial variability (Werner and Heimann, 2002;Herold and Lohmann, 2009; Cruz the biosphere, pedosphere and the karst system, the signal is influenced by various processes, such as evapotranspiration, mixing of water of different age and host rock dissolution, which in turn depend on different parameters such as temperature, the properties of the soil and karst layer, the pCO 2 of soil air and the type and seasonal state of vegetation (Dreybrodt and Scholz, 2011). Isotope fractionation between the δ 18 O value of the drip water and the precipitated speleothem calcite also depends on several conditions inside the cave, such as cave temperature, drip interval and supersaturation of the drip water with respect to calcite (e.g. Dreybrodt, 2008;Mühlinghaus et al., 2009;O'Neil et al., 1969;Scholz et al., 2009 (Riechelmann et al., 2011). The mixing of water parcels in the soil and karst matrix is assumed to be 48 months (Wackerbarth, 2012), the mean value of the infiltration related drip interval is 3600s, and the mixing parameter is set to 1 (the latter two parameters are needed for calculating kinetic isotope fractionation according to the Mühlinghaus et al. (2009) (Riechelmann, 2010;Fohlmeister et al, 2012).

Results
We start with local temperature and the hydrological cycle in the AGCM. Figure 1 shows the temporal evolution of the simulated surface temperature (panel a), local precipitation minus evaporation (panel b), and the δ 18 O value of precipitation at the cave site. 15 The panels indicate pronounced seasonal, interannual, and decadal climate variability. Figure 2 shows the relation between the simulated monthly δ 18 O values and surface temperature, indicating a positive correlation between the δ 18 O value of local precipitation and temperature. The climate information is used as an input for the ODSM stalagmite proxy model. Figure 3 (Fig. 3). Figure 4 displays the corresponding spectra of the cave temperature and speleothem calcite δ 18 O values in Bunker cave. Interestingly, the interannual variability 25 in calcite δ 18 O is suppressed (Fig. 4b), and the power spectrum shows a significant peak at about 14 yr (p = 0.002). In order to relate the recorded δ 18 O signal to the large-scale temperature, the correlation of simulated speleothem calcite δ 18 O in the Bunker cave and SST is evaluated (Fig. 5). Areas showing a significant correlation (95 % confidence level, t-test) are colored. The correlation map with SST is characterized by zonal bands of SST stacked in the meridional direction.

5
Due to the delay between the infiltration of a water parcel and its inflow into the cave (i.e. a lag of 2 yr for Bunker Cave, Kluge et al., 2010) and the propagation of the climate pattern, we also calculate the lag-correlation and lag-composite maps. To extract the patterns that coincide with a maximum index, we applied a Composite Map Analysis (von Storch and Zwiers, 2003) between the δ 18 O calcite (Fig. 3b) and differ-10 ent horizontal quantities (SSTs and precipitation). For the calculation we use all time slices that are above −5.2 ‰. We apply composite analysis for different lags prior to the maxima in δ 18 O calcite (Fig. 6). The results are not sensitive to the exact choice of the threshold (not shown). The SST anomalies develop around the Gulf Stream area south of Newfoundland (Fig. 6a) and propagate 1 yr (Fig. 6b) and 2 yr (Fig. 6c) further 15 downstream. Of interest is also the hydrological budget and its spatial extension. Figure 7 displays the composite map of the δ 18 O value of precipitation (‰) with respect to speleothem calcite δ 18 O (Fig. 3b), indicating a regional coherence in Central Europe. Because the surface signal is transported within ∼ 2 yr, we again calculate a lagged composite map.

Discussion
Instrumental surface temperature data over the last century depict strong variability at interannual to multidecadal time scales. There is evidence that the global climate system contains modes of climatic variability operating on decadal to multidecadal time scales involving temperature and atmospheric circulation (e.g. Mann et al., 1995; Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | speleothem climate archive for a specific cave site in western Germany where largescale patterns may play a role. We combine two models, one AGCM and a speleothem proxy model for Bunker cave. We build pseudo proxy data by calculating the local speleothem calcite δ 18 O values in Bunker Cave (Fig. 3b). This allows attributing dominant signals of variability in observed and proxy data of the North Atlantic realm to 5 changes in SST forcing.
The decadal SST-signature is characterized by a remote North Atlantic tripolar pattern (Figs. 5 and 6) and has a quasi-decadal time scale (Fig. 4). Due to the lagged response, this also projects to a European regional pattern. It has been proposed that this mode with a time period of 12-14 yr results from ocean-atmosphere and tropics-10 midlatitudes interactions in the North Atlantic basin (Deser and Blackmon, 1993;Dima et al., 2001;Dima and Lohmann, 2004). The clear distinction of peaks in the climate spectra suggests that other than random processes may be responsible for the Atlantic quasi-decadal mode (Deser and Blackmon, 1993).
An important question is the mechanism of the filtering through the speleothem cli-15 mate archive. Due to mixing processes in the soil and karst matrix, the δ 18 O prec signal is smoothed to an infiltration weighted mean δ 18 O value. The extent of mixing determines the variance of the simulated δ 18 O drip . We find that the spectrum of the calcite δ 18 O value in Bunker Cave is dampened for interannual time scales providing for a natural filter. We emphasize that this parameter is location-dependent and can be 20 calibrated to agree with the observed natural variance in the particular cave system (Wackerbarth et al., 2010). However, for the interpretation of stalagmite proxy records, the smoothing of the signal may be an advantage. Werner and Heimann (2002) found that simulated δ 18 O records at ice core sites indicate year-to-year variations masked by internal atmospheric variability. Indeed, one can interpret ice core proxy data in terms 25 of the frequency of weather patterns (Rimbu and Lohmann, 2010). If a proxy filters out the inherent noise of the climate system, it may be easier to detect a deterministic response to a large-scale SST pattern (Rodwell et al., 1999;Grosfeld et al., 2007;Latif et al., 2000;Sutton and Hodson, 2003;Robertson et al., 2000). The smoothing, Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | of course, makes it more difficult to detect the corresponding mechanism, especially when dealing with relatively short periods as done in Baker et al. (2011). We admit that the actual mixing process in the real caves are more complex than in our ODSM model (Wackerbarth et al., 2010) and might be climate dependent. The mode indicates the propagation of SST anomalies from the Gulf Stream region 5 along the gyre circulation (Dima and Lohmann, 2004, cf. Fig. 5). Evidence for the Gulf Stream SST anomalies to be transferred from mid-latitudes into the tropics through surface advection is further supported by a lag composite analysis between speleothem calcite δ 18 O and the SST field (Fig. 6). Local surface temperature is largely determined by the surrounding SST. The moderate correlation (0.4) with temperature indicates that 10 other processes than SST affect the calcite δ 18 O. Besides the effect through the modulation of δ 18 O via temperature, the hydrological cycle shows a spatially coherent pattern (Fig. 7). Thus, we would expect a similar temporal behaviour in the areas showing a positive correlation in Fig. 7.

15
Several attempts to reconstruct reliable climate information from stalagmites over the last few centuries have been made to reconstruct large-scale climate patterns for the last millennium (see, for instance, the review papers by McDermott (2004) and Lachniet (2009)). At present, the modes of climate variability and their modulation through longer-term background climate, and how this has varied in the past, is not well known. 20 Accordingly, climate models used to assess potential changes of these climate modes in the future are only poorly constrained. On the other hand, cave monitoring programs are not in a stage that they could cover decades to study the environmental processes relevant for climate reconstructions. Using a pseudo-proxy approach extracted from AGCM simulations and a proxy module, we analyze the (modelled) reconstructions in the light of variability modes. We find that the regional response in speleothem calcite δ 18 O is sensitive to environmental