Western Mediterranean Climate Response to Dansgaard/Oeschger Events: New Insights From Speleothem Records

The climate of the western Mediterranean was characterized by a strong precipitation gradient during the Holocene driven by atmospheric circulation patterns. The scarcity of terrestrial paleoclimate archives has precluded exploring this hydroclimate pattern during Marine Isotope Stages 5 to 3. Here we present stable carbon and oxygen isotope records from three flowstones from southeast Iberia, which show that Dansgaard/Oeschger events were associated with more humid conditions. This is in agreement with other records from the Iberian Peninsula, the Mediterranean, and western Europe, which all responded in a similar way to millennial‐scale climate variability in Greenland. This general increase in precipitation during Dansgaard/Oeschger events cannot be explained by any present‐day or Holocene winter atmospheric circulation pattern. Instead, we suggest that changes in sea surface temperature played a dominant role in determining precipitation amounts in the western Mediterranean.


Introduction
During the last glacial period, global climate underwent a series of rapid changes superimposed on a longterm cooling trend (Deplazes et al., 2013;Martrat et al., 2007;North Greenland Ice Core Project members, 2004;Wang et al., 2008). In particular, Greenland ice core records show large, rapid changes in δ 18 O values interpreted as changes in temperature (Johnsen et al., 2001;Wolff et al., 2010). This climate variability on millennial timescales with warm Dansgaard/Oeschger (D/O) events-also referred to as Greenland Interstadials (Dansgaard et al., 1993)-and Greenland stadials (Hemming, 2004) is reflected in marine sediment cores from the North Atlantic (Heinrich, 1988;Hemming, 2004;Martrat et al., 2007) as well as the Mediterranean Sea (Cacho et al., 1999;Frigola et al., 2012;Incarbona et al., 2013;Martrat et al., 2004). These millennial-scale fluctuations are assumed to have been linked to the Atlantic Meridional Overturning Circulation (AMOC), which strongly affected climate variability in the Northern Hemisphere during Marine Isotope Stage (MIS) 5b to 3 (96-29 ka; Henry et al., 2016;Li & Born, 2019). The strength of the AMOC controls oceanic heat transport to high northern latitudes (Böhm et al., 2015;Menviel et al., 2014). Concomitantly, the AMOC has a large impact on sea surface temperatures (SST) in the North Atlantic (Pailler & Bard, 2002) as well as the western Mediterranean Sea (Bagniewski et al., 2017;Martrat et al., 2004). Changes in SST and the position and topography of the ice sheets, in turn, have an impact on the atmospheric circulation (Cacho et al., 2000;Merz et al., 2015;Moreno et al., 2005;Naughton et al., 2009) by influencing the pathways of North Atlantic storm tracks and the position of the Intertropical Convergence Zone (ITCZ; Naughton et al., 2009;Stríkis et al., 2018) and, as a consequence, on effective precipitation in the Mediterranean (Hodge, Richards, Smart, Andreo, et al., 2008;Hodge, Richards, Smart, Ginés, & Mattey, 2008). A decrease or even shutdown of the AMOC is coupled to lower SSTs in the North Atlantic and a southward shifted oceanic thermal front, which results in a more southerly route of the Atlantic jet stream and its associated westerlies (Naughton et al., 2009).
The present-day climate of the Iberian Peninsula is dominated by several regional atmospheric circulation patterns. For instance, the Western Mediterranean Oscillation (WeMO) and the North Atlantic Oscillation (NAO) lead to strong spatial differences in precipitation (Comas-Bru & McDermott, 2014;Hurrell & Loon, 1997;Martin-Vide & Lopez-Bustins, 2006). Pollen records from offshore sediments provide a wealth of information on last interglacial to glacial vegetation and climate changes, but these data reflect a large region encompassing subregions dominated by different atmospheric circulation patterns. Thus, in order to understand the importance of regional atmospheric circulation patterns, precisely dated terrestrial climate records are needed. However, so far, only very few terrestrial records from the Iberian Peninsula covering MIS 5b to 3 are available. Here we present carbon and oxygen isotope records from three precisely dated flowstones from Cueva Victoria (CV, SE Spain), covering the period from 96 to 45 ka (MIS 5b-3, including D/O events 22 to 12). These records allow to examine how climate variability in the western Mediterranean was related to abrupt climate change in the North Atlantic during the last glacial.

Sample Site and Methods
CV (37.63°N, 0.82°W, 40 m above sea level; Figure 1) is located in SE Spain and developed in Triassic limestones and dolostones of the Inner Betic Cordillera (Manteca Martínez & Pina, 2015). The cave system consists of more than 3 km of galleries that were artificially widened by mining activities during the early twentieth century. The climate of the region shows a strong seasonality with warm and dry summers and precipitation maxima in spring and autumn (up to 300 mm/year, mean annual temperature ≈ 17°C; Figure 1c). This seasonality is also reflected by the monthly δ 18 O values of local precipitation, which show an inverse correlation with rainfall and temperature (Araguas-Araguas & Diaz Teijeiro, 2005). On the interannual timescale, the δ 18 O values of precipitation do not show a significant correlation to temperature or precipitation (Budsky et al., 2019). Climate is classified as a cold semi-arid climate (BSk) according to the Köppen-Geiger classification (Kottek et al., 2006). The vegetation period lasts from spring to summer and highly depends on rainfall during these seasons (Camarero et al., 2015;Pasho et al., 2011). Precipitation is prevalent during periods characterized by a negative WeMO (Cortesi et al., 2014;Martin-Vide & Lopez-Bustins, 2006;Moreno et al., 2014) index ( Figure 1a). The main moisture sources of precipitation are the surrounding Mediterranean and Alboran Sea as well as the North Atlantic (Budsky et al., 2019). There is no direct influence of the NAO (Figure 1b; Hurrell & Loon, 1997) or the ITCZ (Broccoli et al., 2006). Similarly, the main modern European winter circulation patterns, such as the Eastern Atlantic or the Scandinavian pattern (Barnston & Livezey, 1987), have almost no impact on precipitation in SE Spain (Comas-Bru & McDermott, 2014).
Flowstone SR01t (6 cm thick) was sampled from the center of "Sala de las Reuniones", while cores Vic-III-1 and Vic-III-3 were drilled in thick (>50 cm) flowstones in room "Victoria 3" (Ros & Llamusí, 2015). For Vic-III-1 (42 cm) and Vic-III-3 (40.5 cm), we focus on the upper 23 and 8 cm here (supporting information Figure S1), which correspond to the last 96 ka (MIS 5b/c transition). Cave monitoring at CV is not possible 10.1029/2019GL084009 Geophysical Research Letters due to the lack of active drip sites and the artificial widening of the cave, which strongly altered the natural cave system. For 230 Th/U dating, small pieces (0.05-0.3 g) were cut from the flowstone, prepared by column chemistry (Gibert et al., 2016;Yang et al., 2015), and analyzed using multicollector inductively coupled plasma mass spectrometry at the Max Planck Institute for Chemistry in Mainz (Obert et al., 2016). Samples for stable isotope analysis were milled at an equidistant spacing of 500 μm (Vic-III-1, SR01t) and 250 μm (Vic-III-3), respectively. The obtained powders were analyzed at the University of Innsbruck with a Delta plus XL isotope ratio mass spectrometer linked to a Gasbench II (Spötl, 2011;Spötl & Vennemann, 2003).
for Vic-III-1 and 0.37 ± 0.19 for Vic-III-3, respectively, which are in agreement within uncertainty. For sample SR01t, the correction is negligible due to its low content of detrital material (( 230 Th/ 232 Th) ≫ 200; Richards & Dorale, 2003). An exception is subsample SR01t-11 (( 230 Th/ 232 Th) = 49.01). Therefore, we used the mean ( 232 Th/ 238 U) activity ratio of samples Vic-III-1 and Vic-III-3 for detrital correction (( 232 Th/ 238 U) = 0.31 ± 0.16) for SR01t. The corrected ages range from 95.7 ± 4.7 to 46.2 ± 0.6 ka (Vic-III-1), excluding the uppermost age, which corresponds to the Holocene (Budsky et al., 2019), 92.8 ± 1.8 to 49.9 ± 0.4 ka (Vic-III-3) and 85.4 ± 1.2 to 49.5 ± 1.3 ka (SR01t). The final age models for all flowstones were constructed using the corrected ages and the StalAge algorithm (Scholz & Hoffmann, 2011; Figure S2). Visible hiatuses were included manually into the StalAge age model by fitting the corresponding flowstone sections separately (see supporting information). For very short growth phases consisting of only one 230 Th/U-age, we used a mean growth rate of the corresponding longer growth intervals to establish an age-depth model. These short growth intervals were stratigraphically identified by dark layers in thin sections. There is no evidence of dissolution or diagenesis at these growth stops. Thin sections show a pristine elongated/open columnar fabric (cf. Frisia, 2015).

Interpretation of the CV Speleothem Record
The three flowstone records from CV cover a long period between the last interglacial and the Last Glacial Maximum, which is only sparsely covered by other paleoclimate archives from SE Spain. The typical D/O pattern as recorded by the North Greenland Ice Core Project ice core (North Greenland Ice Core Project members, 2004; Obrochta et al., 2014;Svensson et al., 2013) is reflected in both the carbon and oxygen isotope records of the CV flowstones with lower values occurring during D/O events and vice versa. This indicates a strong link between climate in the North Atlantic region and SE Spain on millennial timescales. In the overlapping sections, all three flowstones show consistent δ 13 C and δ 18 O values (Figures 2 and 3). This replication of the proxy signals confirms that the observed variability is related to climate change above the cave rather than processes within the cave or the karst aquifer. Temporal discrepancies are likely largely due to uncertainties in the chronology of our flowstones, in particular for a few short growth periods that cannot be constrained by more than one 230 Th/U-age, and probably to a smaller extent due to the uncertainties of the chronology of North Greenland Ice Core Project (up to 1.5 ka; Svensson et al., 2008; Figure 2). Budsky et al. (2019) demonstrated a strong influence of effective precipitation on vegetation density and microbiological activity in the soil above CV during the Holocene. Higher δ 13 C values were interpreted as decreased precipitation during the season of vegetation growth (spring to summer). This interpretation is in agreement with other studies using δ 13 C values as a proxy for vegetation density (Cerling et al., 1993;Fohlmeister et al., 2011) and soil microbiological activity (Genty et al., 2003;Meyer et al., 2014), which in turn are related to effective precipitation during the growing season (Denniston et al., 2018;Hodge, Richards, Smart, Andreo, et al., 2008;Hodge, Richards, Smart, Ginés, & Mattey, 2008). In addition to these processes occurring in the soil zone, several processes within the aquifer and the cave can result in carbon isotope fractionation, such as prior calcite precipitation, cave ventilation, and the distance of flow on flowstones (Hansen et al., 2017;Johnson et al., 2006;Mühlinghaus et al., 2009;Spötl et al., 2005). Stronger cave ventilation and increased distances of flow lead to enhanced degassing of CO 2 from the solution and precipitation of calcite prior to the sampling site. This may be particularly relevant for sample SR01t, which is associated with the longest distance of flow due to its position in the middle of the cave chamber and may explain the elevated δ 13 C values compared to the other flowstones. In general, despite their complexity, all processes result in higher δ 13 C values during drier conditions above the cave. Thus, we interpret higher δ 13 C values in the CV flowstones as reflecting periods of reduced precipitation/infiltration.
Lower δ 13 C values during D/O events thus suggest increased precipitation in the western Mediterranean during these warm events in the North Atlantic region (Budsky et al., 2019;Genty et al., 2003). This is in agreement with other climate archives from the Mediterranean, such as pollen records from the western Mediterranean (Burjachs et al., 2012;Camuera et al., 2019;Combourieu Nebout et al., 2002;Sánchez Goñi et al., 2008) and Italy (Allen et al., 1999).  (Moreno et al., 2014) preventing the possibility of disentangling precipitation originating in the Atlantic from that originating from the Mediterranean region. This also implies that the transport distance of the water vapor and potential rain-out effects (McDermott, 2004;Mook, 2001) are not dominant because the Mediterranean is the more local moisture source compared to the more distant Atlantic. In addition, temporal changes in the δ 18 O value of surface ocean water have to be taken into account, but unfortunately, there is no seawater δ 18 O reconstruction available from the region. δ 18 O values of planktonic foraminifera in sediment cores from the Iberian margin and the Alboran Sea (Vautravers & Shackleton, 2006), a proxy for both SST and the δ 18 O value of seawater, reflect all D/O events (Figure 3b). However, considering the temperature dependence of the δ 18 O values of planktonic foraminifera of~−0.21‰/°C (Bemis et al., 1998) suggests only minor changes in the δ 18 O value of seawater during the D/O events (Figures 3b and 3c). Therefore, we interpret our flowstone δ 18 O record as reflecting a combination of the amount of winter precipitation and cave air temperature, with more negative flowstone δ 18 O values corresponding to warmer and more humid conditions. This relationship is potentially weakened by the positive relationship between surface air temperature and rainfall δ 18 O values.

Climate Variability on Orbital Timescales
On orbital timescales, the flowstone δ 18 O record follows 65°N July insolation, whereby high insolation is associated with low δ 18 O and δ 13 C values (warm and humid) and vice versa (Figures 3d and 3e; Berger, 1978). Only few terrestrial climate archives in southern Europe and the western Mediterranean cover MIS 5 to 3. For MIS 5c-a, coastal sediments suggest more humid conditions in SE Spain (Mauz et al., 2012). Enhanced precipitation during interglacials is also corroborated by stalagmite growth in northern Iberia (Muñoz-García et al., 2007;Stoll et al., 2013). Located close to CV, the low-resolution Gitana Cave record (Figure 1) is in good agreement with our records on orbital timescales, with higher effective precipitation during interglacials and a cessation of speleothem growth during Heinrich stadial 5 (≈ 46 ka; Hodge, Richards, Smart, Andreo, et al., 2008). This coincides with a prominent sea level drop at circa 45 ka (Siddall et al., 2008) and the termination of calcite deposition in CV. Speleothem δ 13 C values from Portugal also suggest that high 65°N summer insolation is associated with higher precipitation (Denniston et al., 2018). In summary, on orbital timescales, precipitation on both the western and the eastern Iberian Peninsula responds to 65°N July insolation.
The 65°N July insolation depends on the interplay between obliquity and precession (Davis & Brewer, 2009). Both lead to a varying latitudinal insolation gradient, which in turn drives the latitudinal temperature gradient and thus climate in higher and lower latitudes by a latitudinal displacement and varying intensity of the midlatitude storm tracks and the tropical Hadley Cell/ITCZ (Schneider et al., 2014;Stríkis et al., 2018). In particular, precession minima are associated with stronger latitudinal shifts of the ITCZ and the midlatitude storm tracks on seasonal timescales and are thus associated with higher seasonality and enhanced autumn/winter precipitation due to higher storm activity in the Mediterranean (Bosmans et al., 2015;Kutzbach et al., 2014;Toucanne et al., 2015). This phenomenon may explain the wetter conditions observed around 80-85 ka in our record, but not those at 50-60 ka (Figure 3). This suggests that the combined signal of 65°N July insolation is more important than precession alone. Higher 65°N July insolation during interglacials is associated with a weaker latitudinal temperature gradient. A higher temperature gradient during glacial periods was associated with a weakened AMOC (Böhm et al., 2015) and leads to stronger and southward shifted westerlies (Merz et al., 2015). Consequently, this should lead to more precipitation on the Iberian Peninsula during glacial periods (Hofer, Raible, Merz, et al., 2012). Interestingly, this is not observed in the western Iberian Peninsula, where glacial periods were characterized by drier conditions, whereas interglacials were relatively wet (Denniston et al., 2018). This apparent controversy may be explained by the fact that the glacials (interglacials) were associated with a reduced (stronger) AMOC and lower (higher) 65°N July insolation, which lead to lower (higher) SSTs. On orbital timescales, SSTs at the Iberian margin are correlated with precipitation at both the western (Denniston et al., 2018) and the eastern Iberian Peninsula (CV; this study). This strongly suggests that SST controlled precipitation on the Iberian Peninsula on orbital timescales, although it remains difficult to assess whether this is related to an increase in winter or summer precipitation or both (Kutzbach et al., 2014).
Warming of the North Atlantic during D/O events is associated with an enhanced AMOC, which results in a decreased temperature gradient. In general, a weakened AMOC during stadials reduces the heat transport to the North concomitant with reduced SSTs (Bagniewski et al., 2017). In combination with the presence of the Laurentide ice sheet, this induces a southward shift of the Hadley Cell, associated with stronger and southward shifted westerlies (Menviel et al., 2014). Stronger and southward shifted westerlies during stadials lead to decreased precipitation over the Iberian Peninsula and vice versa (Bagniewski et al., 2017;Menviel et al., 2014). Nevertheless, from a sediment core off the northwest Iberian coast, a more complex response of precipitation to Heinrich events 4, 2, and 1 was observed (Naughton et al., 2009). Naughton et al. (2009) suggested that the first phase of the Heinrich events was associated with relatively wet and cold conditions, whereas the second phase was characterized by dry and cold conditions following the displacement of the ocean polar front. However, since the CV record does not cover Heinrich events 4, 2, and 1, this cannot be verified for the eastern Iberian Peninsula. Moreover, the CV record shows drier conditions during stadials and wetter conditions during D/O events.

Precipitation Patterns: Present-Day Versus Last Glacial Period
The present-day precipitation distribution on the Iberian Peninsula is strongly influenced by several atmospheric circulation patterns including the WeMO and the NAO (  Figures 1a and 1d). This bipolar precipitation pattern has also been discussed in detail for the Holocene (Budsky et al., 2019). In contrast, the NAO particularly affects precipitation in the regions of the Iberian Peninsula that are not affected by the WeMO (Figure 1b).
During the last glacial D/O events, it is well documented that many regions in addition to southeastern Spain, such as western Europe (Genty et al., 2003;Wainer et al., 2009) and the eastern Mediterranean (Grant et al., 2012;Ünal-İmer et al., 2015), also experienced increased precipitation. This is in agreement with higher tree pollen percentages in Greece (Tenaghi Philippon, Tzedakis et al., 2003) and Italy (Lake Monticchio, Allen et al., 1999) -Bustins, 2006). Instead, we suggest that changes in North Atlantic and Mediterranean SSTs controlled the water vapor content of the atmosphere and regulated changes in precipitation. We emphasize that this general increase in precipitation neither excludes changes in atmospheric circulation during D/O events nor must have been restricted to a specific season.
Such a strong link between SST and precipitation has also been suggested for the last glacial (Denniston et al., 2018;Hodge, Richards, Smart, Ginés, & Mattey, 2008), primarily for the Mediterranean due to instabilities in winter associated with high SSTs during D/O events (Bosmans et al., 2015). In particular, it is well known that the water vapor content of the air over the North Atlantic increased during warmer periods, and these warm and moist air masses were then transported to the western Mediterranean causing an increase in precipitation (Bosmans et al., 2015;Kutzbach & Liu, 1997;Trenberth et al., 1998). During glacials, cool SSTs in the North Atlantic decreased the energy budget over the ocean and the moisture uptake in winter. This resulted in drier conditions in the western Mediterranean (Daniau et al., 2007;Dumitru et al., 2018;Hodge, Richards, Smart, Ginés, & Mattey, 2008;Moreno et al., 2005).

Conclusions
Three overlapping flowstone δ 13 C and δ 18 O records from CV demonstrate that precipitation in SE Spain between MIS 5b and MIS 3 (96-45 ka) was related to North Atlantic climate variability. Warm D/O events were associated with higher precipitation and an expansion of vegetation, even during short D/O events, such as D/O 15 and 18. Cold stadials were associated with lower precipitation and reduced vegetation cover. Warm and humid conditions during D/O events are also recorded by pollen and were associated with an expansion of forests in the Mediterranean region.
Climate of the Iberian Peninsula during the Holocene and the present day shows strong regional differences due to different controlling factors, such as the NAO and the WeMO. However, vast regions in the Mediterranean and western Europe show coherently more humid conditions during D/O events and drier conditions during Greenland stadials. We conclude that this coherent large-scale climate response cannot be explained by present-day winter atmospheric circulation patterns alone. Instead, the SST of the North Atlantic and the Mediterranean Sea played a key role in determining the water vapor content of the atmosphere that controlled precipitation in the western Mediterranean and western Europe.