Introduction

The last glacial period is characterized by centennial- to millennial-scale climate oscillations known as Dansgaard - Oeschger (D-O) cycles. At least 25 D-O cycles were identified in Greenland ice cores and characterized by warm Greenland interstadials (GI) and cold Greenland stadials (GS)1,2,3. Though the timing, duration, and spatial extent of D-O cycles are well-documented, uncertainties exist regarding their triggers and pacing. Proposed causes for D-O cycles include changes in the Atlantic Meridional Overturning Circulation (AMOC) and solar and volcanic forcing4,5,6,7,8,9. While D-O cycles are well documented for the last glacial period1,10,11,12,13,14,15, there are only very few terrestrial proxy records which provide unambiguous evidence for their existence during the preceding glacial periods12,13,14,16,17,18,19. This knowledge gap limits our ability to evaluate the influence of different glacial boundary conditions on D-O variability, which would contribute to a better understanding of their driving mechanisms. The penultimate glacial period corresponding to Marine Isotope Stage (MIS) 6 (191-130 kyrs BP) was most likely one of the strongest Quaternary glaciations, with much greater ice sheet extent throughout Eurasia20,21. Though MIS 6 is not directly covered by Greenland ice core records, a synthetic δ18O Greenland record suggests the occurrence of D-O like cycles during the penultimate glacial22. Further records are needed to better understand D-O cycles and to characterize their pattern, timing, and spatial extent. Such information can be obtained from speleothems, which are known to be an ideal archive for identifying D-O cycles15. However, speleothem-based climate reconstructions covering the penultimate and preceding glacial periods are still rare and most of them do not display a clear D-O pattern in their isotopic profiles11,23,24,25,26,27,28,29. At present, D-O cycles have been explicitly identified in tropical and subtropical speleothems from Sanbao Cave16 and Huagapo Cave17, respectively, but additional precisely dated and highly resolved paleoclimate records are urgently needed to verify the synthetic Greenland ice core record and to develop a more detailed stratigraphy for D-O cycles during MIS 6.

Here we present an extended Sofular record for MIS 2-4 and parts of MIS 6/7 to show D-O cycles in very close detail. Complementing previous studies30,31, we present so far unpublished δ13C isotope profiles to provide further evidence for a very close atmospheric coupling between the eastern Mediterranean and the North Atlantic during the penultimate glacial period. Because of the very strong atmospheric connection between the North Atlantic region and Northern Türkiye15,30,31, Sofular stalagmites are very well suited to investigate millennial-scale variability for glacial periods preceding MIS 2-4.

Results and discussion

Study site

Sofular Cave (41°25’N, 31°56’E) is located in the North Anatolian mountains in Türkiye, ~10 km inland from the southern Black Sea coast (Fig. 1). The cave is within the Lower Cretaceous limestones. Mean annual temperature and total precipitation average ~13.3 °C and ~1,200 mm yr−1, respectively. Although ~75% of total annual precipitation occurs between September and April, infiltration of rainwater occurs most likely throughout the year. Moisture originates predominantly from the Black Sea and, to a lesser extent, from the Marmara and Mediterranean Sea30,31. The soil above the cave is of heterogeneous thickness and covered by dense vegetation of deciduous trees (e.g., oaks, elms, and beeches), shrubs, and marginal grasslands.

Fig. 1: Map showing Eurasia including the location of Sofular Cave (red star).
figure 1

Black circles mark the location of Lake Van and Black Sea sediment cores GC25-1 and MSM33. The maximum extent of the Eurasian Ice Sheet during the Saalian (Marine Isotope Stage (MIS) 6; ice sheet boundaries redrawn after Svendsen et al.20) and Last Glacial Maximum (LGM)60 are shown by the blue and red lines, respectively. Arrows indicate potential meltwater pathways during MIS 6.

Full size image

Chronology

The chronology of the Sofular record is based on 150 230Th ages, all of which are in stratigraphic order. Uranium concentrations of ~0.5 ppm and low common thorium result in very precise 230Th ages for all stalagmites with typical age uncertainties of ~0.5%. The Sofular record covers the last 205,000 years before present (BP = 1950) almost continuously, except for two gaps between 24.1 - 21.6 kyrs BP and 160 - 134 kyrs BP (Figs. 2 and 3).

Fig. 2: Sofular stalagmite δ18O and δ13C records.
figure 2

a Sofular stalagmites So-1, So-2, So-4 and So-13 covering the Holocene and last glacial period (data for So-1 and So-2 taken from Fleitmann et al.30). b Sofular stalagmites So-4, So-6 and So-57 covering the penultimate glacial period (δ18O data for So-4 and So-6 taken from Badertscher et al.31). Color coded 230Th ages with 2σ-error bars are plotted below. Source data are provided in the Source Data file.

Full size image
Fig. 3: Comparison of climate variability during the last and penultimate glacial period.
figure 3

a Records showing Dansgaard-Oeschger (D-O) climate variability of the last glacial period. From top to bottom: δ18O of the Greenland ice core3, δ13C stacked record from Sofular Cave, and a sea surface temperature (SST) record from the Black Sea33. Blue bars mark Heinrich events. Numbers above the records denote D-O events. b Records showing D-O climate variability of the penultimate glacial period. From top to bottom: δ18O of the synthetic ice core22, δ18O of Sanbao Cave16, δ18O of Huagapo Cave17, δ13C stacked record from Sofular Cave, and a Black Sea SST record43. Numbers above the records denote D-O events. Source data are provided in the Source Data file.

Full size image

Sofular isotope profiles

The average temporal resolution of the entire Sofular record is around 40 years. Including previous studies30,31, 8540 stable oxygen (δ18O) and carbon (δ13C) isotope measurements were performed. Additional high-resolution data were obtained from stalagmites So-4, So-13 and So-57, and we now present the complete Sofular δ13C record. δ18O and δ13C values range from −7.6‰ to −17.8‰ (VPBD) and −2.7‰ to −10.6‰ (VPBD), respectively (Fig. 2). Oxygen and carbon isotope profiles of all stalagmites are very similar, with correlation coefficients (r) of stalagmites covering MIS 1-4 (So-1, So-2, So-4, So-13) typically ranging from ~0.4 to ~1.0 (determined using iscam32)(Supplementary Fig. 1a, b). For MIS 6 (So-4 and So-57) correlation coefficients range from ~0.7 to ~1.0 (Supplementary Fig. 1c).

For both MIS 2-4 and MIS 6, isotope profiles show distinct centennial- to millennial-scale variability and D-O like patterns, which are, even on multi-decadal timescales, very similar to those recorded in Greenland ice cores (Fig. 3). The strong resemblance between the Sofular and NGRIP isotope records demonstrates the very close and rapid atmospheric coupling between the North Atlantic and Black Sea region15,30,33.

D-O related variations of 0.5 to 1.5‰ in speleothem δ18O are most likely caused by multiple factors such as air temperature, variations in δ18O of the moisture source, seasonality, and amount of precipitation above Sofular Cave30,31. On orbital time-scales fluctuations in Sofular δ18O values track changes in the oxygen isotopic composition of Black Sea surface water through the so-called water vapor source effect30,31. δ18O values of around −8.5 ± 1‰ mark periods when the Black Sea was connected with the Mediterranean Sea when the sea level was higher than the Bosporus sill depth of ~35 meters below current sea level31. More negative δ18O values are therefore indicative of a Black Sea “Lake”34 without inflow of saline Mediterranean water. Pronounced negative excursions in δ18O of up to −17.8‰, however, indicate the inflow of isotopically depleted meltwater from the Caspian Sea and rivers entering the Black Sea”Lake”34 (Supplementary Fig. 2). Thus, the water vapor source effect in combination with variable mixing times of the Black Sea35 affected D-O related fluctuations in δ18O during MIS 2-4 and particularly during MIS 6. This effect is also apparent in Sofular stalagmites So-1 and So-2 δ18O profiles which show a suppressed Bølling-Allerød and Younger Dryas isotopic pattern31 (Fig. 2).

In contrast, D-O cycles are more clearly visible in the Sofular δ13C isotope profiles indicating a high sensitivity of δ13C to D-O climate variability. Millennial-scale δ13C fluctuations in Sofular speleothems are similar between the last and penultimate glacial ranging from −5‰ to −10‰. Though So-4 and So-57 δ13C profiles show an almost identical isotopic pattern, δ13C values of stalagmite So-57 are on average 1.3‰ more positive. This isotopic offset is most likely related to sample-specific effects caused by lower drip rates and stronger fractionation due to CO2 degassing. In order to reduce sample-specific effects, age uncertainties and to increase the signal-to-noise ratio, we developed a stacked δ13C record of MIS 1-4 and MIS 6 using the program called intra-site correlation age modeling (iscam)32. Iscam was designed to synchronize overlapping time series from the same site within their age uncertainties. D-O cycles in the Sofular δ13C stacked record are characterized by high amplitude shifts of up to 6‰ (Fig. 3a). We suggest that these shifts are primarily caused by temperature- and moisture-related changes in vegetation and soil microbial activity, with higher vegetation density and soil respiration rates associated with warmer and wetter GIs (Fig. 3a). Our assumption is based on the following lines of evidence. Firstly, a close coupling between climate and soil respiration rates is directly supported by paired δ13C and 14C measurements in stalagmite So-136. The paired measurements reveal a rapid response of soil respiration rates and microbial activity across the Bølling-Allerød and Younger Dryas in stalagmite So-1, with enhanced decomposition of soil organic matter (SOM) under warmer and wetter climatic conditions36.

Secondly, soil respiration rate is closely linked to temperature and soil moisture conditions on a range of timescales37,38, a relationship that is supported by modeling experiments39 and speleothem records from western Europe40,41,42. Thirdly, the Sofular δ13C stacked records show an inverse correlation (r2) of ~0.7 (MIS 2-4) and ~0.5 (MIS 6) with sea surface temperature (SST) records from the Black Sea (gravity core GC25-1 & MSM33)33,43 (Fig. 4, Supplementary Figs. 3 and 4). Such a close coupling between air temperature above Sofular Cave and Black Sea SSTs can be expected as the cave site is only ~10 km inland. Therefore, positive SST shifts of up to 4 °C associated with GIs33 (Fig. 3) would lead to a marked increase in temperature and precipitation above Sofular Cave, thereby promoting a change in the type and density of vegetation44 and an increase in soil respiration rates due to higher soil microbial activity and faster decomposition of SOM36. In contrast, cold-dry climatic conditions would affect vegetation density and type and reduce CO2 production due to lower microbial activity, and sparser vegetation is expected to result in lower soil pCO2, lower contribution of biogenic carbon, and, consequently, higher stalagmite δ13C values.

Fig. 4: Correlation between Sofular δ13C and sea surface temperature (SST) from the Black Sea.
figure 4

The scatterplots show Sofular δ13C values of Dansgaard-Oeschger (D-O) onsets, midpoints and peaks vs sea surface temperatures (SST) from the Black Sea33,43 during a the Marine Isotope Stage (MIS) 2-4 (n = 45 and p<<0.01) and b Marine Isotope Stage (MIS) 6 (n = 9 and p ~ 0.04). Shaded areas indicate the 95% confidence interval.

Full size image

The Sofular stack (Fig. 3a) exhibits the most positive δ13C values of around −3‰ at ~62 kyrs BP, concomitant with a major advance of the Eurasian Ice Sheet (EIS)45,46 and Heinrich event 6. In the Black Sea region, this interval was dominated by steppe pollen taxa44 and higher ice-rafted debris in Black Sea sediments47. Heinrich events H5 – H0 are also well expressed in the Sofular δ13C profile, and characterized by marked SST decreases of the Black Sea (Fig. 3a) and reduced precipitation in the Eastern Mediterranean33,47,48. Climate simulations suggest colder and drier conditions induced by a cyclonic atmospheric circulation anomaly modulating the eastward advection of cold air over Eurasia caused by a weakening of the AMOC during Heinrich events and GSs49,50,51,52.

D-O cycles during MIS 6

During MIS 6, a D-O like climate variability is apparent in the δ13C records of stalagmites So-4 and So-57. Similar to MIS 2-4, D-O cycles during MIS 6 are characterized by abrupt negative shifts of up to 4‰ in δ13C (Fig. 3b) in response to warmer and more humid climatic conditions and higher soil respiration rates, denser vegetation and higher proportions of C3 plant vegetation36. Phases of increasing temperature and higher effective moisture are broadly consistent with pollen evidence (dec. Quercus, Betula, Pinus) from Lake Van53 (Supplementary Fig. 2a).

D-O cycles in the Sofular stack mirror those of the synthetic Greenland ice core δ18O record22 (Fig. 3b), which was tuned to the absolutely-dated Sanbao Cave record11,54. Radiometrically dated records showing a clear D-O like pattern are currently very rare and their temporal resolution and/or chronological precision are not sufficient to identify D-O cycles26,27,55,56. Thus, records that can be used for a detailed comparison to Sofular are rather scarce. Suitable time series are speleothem δ18O records from Sanbao Cave16 in China and Huagapo Cave17 in Peru (Fig. 3b). Taking advantage of the precise and absolute chronologies of these stalagmite records, including our Sofular stack, we can gain further information on the pacing and timing of D-O like events during MIS 6. The comparison between the records show age offsets of several centuries for D-O 6.6, 6.8, 6.9 and 6.12 (Supplementary Fig. 5). Higher age offsets between the records of almost two millennia are observed for D-O 6.7, 6.10, 6.11, 6.13 and 6.14, likely caused by a combination of uncertainties related to the age models and uncertain assignment of D-O events.

For the time interval between 200 and 160 kyrs BP, the average pacing of D-O cycles is ~4.16 kyrs in the Sofular stack (Fig. 5b, Supplementary Fig. 6), very similar to the pacing of ~4.24 kyrs and ~4.03 kyrs in the Huagapo and Sanbao speleothem records, respectively. All speleothem-based estimates for the D-O pacing are in agreement with a pacing of ~4.00 kyrs in the synthetic Greenland ice core. In contrast, the average D-O pacing recorded by the Sofular stack during MIS 2-4 is ~2.07 kyrs (Fig. 5a, Supplementary Fig. 7), and therefore only half as long compared to MIS 6. The Speleothem Interstadial Onset Compilation data set (SIOC19) and Greenland ice cores show an almost identical D-O pacing of ~2.04 kyrs and ~2.09 kyrs respectively (Fig. 5a). Thus, the absolutely dated Sofular stack record presented here provides additional evidence for a longer pacing of D-O cycles during MIS 6. Since D-O events are closely linked to changes in the strength of AMOC, which in turn is dependent on the rate of North Atlantic Deep Water (NADW) formation and sea-ice extent52, the longer pacing of D-O cycles is most likely directly related to a different AMOC setting during MIS 6. Benthic foraminifera isotope data from the Portuguese margin suggest a shallower and weaker AMOC cell during MIS 6 than during MIS 313. This is in agreement with longer bipolar seesaw events57, greater sea ice cover, and lower amplitude SST variability in the South Atlantic58. Furthermore, iceberg discharge into the North Atlantic was muted during the first half of MIS 6 and ‘classic’ Heinrich events are missing13,57. Overall, there is strong evidence for a weaker AMOC during the penultimate glaciation. AMOC weakening leads to a considerable cooling in the North Atlantic realm and sea-ice advance over the Labrador and Nordic Seas, which in turn increases surface albedo and reduces heat loss from the ocean to the atmosphere52. As a result of these different mean climatic and glacial boundary conditions during MIS 6, thresholds for triggering D-O cycles were most likely higher and led to less frequent and longer pacing of D-O cycles. However, additional and more detailed reconstructions of AMOC intensity during MIS 6 are urgently needed to prove our hypothesis.

Fig. 5: Pacing of Dansgaard-Oeschger (D-O) cycles in Sofular and comparison of stalagmite and ice core chronologies.
figure 5

Comparison of D-O pacing between: a D-O cycles during the last glacial period, recorded in the Greenland ice core NGRIP (data from Rasmussen et al.3), the Sofular stacked record and a Speleothem Interstadial Onset Compilation (SIOC19) data set (data from Corrick et al.15) and b D-O cycles of the penultimate glacial period, recorded in the synthetic ice core record22, the Sofular stacked record, the Huagapo Cave speleothem record17 and the Sanbao Cave speleothem record16. Numbers within the legends refer to the average pacings of D-O events which are taken into account by black arrows.

Full size image

In conclusion, the stacked Sofular δ13C record shows D-O climate variability for the last and penultimate glacial in great detail and thereby provides additional evidence for D-O variability in the synthetic Greenland ice core record. The comparison of absolutely dated speleothem records covering MIS 6, including the Sofular stack, reveals an agreement in both the timing and pacing of D-O cycles. Thus, there is mounting evidence for a significantly longer pacing of D-O cycles during MIS 6 compared to MIS 2-4. This could be related to a generally weaker and shallower AMOC and reduced northward heat transport and significant cooling of the North Atlantic realm, thereby inducing a lower pacing of D-O events during MIS 6.

Methods

Stalagmite samples

A total of six stalagmites from Sofular Cave (So-1, So-2, So-4, So-6, So-13, and So-57) with heights ranging between 0.8 and 1.7 meters were analyzed. So-1, So-2, and parts of So-4 and So-6 have been investigated in previous studies30,31, while this study provides complementary δ13C profiles on So-4 and So-6, and previously unpublished data of stalagmites So-13 and So-57.

230Th dating and age model development

U-series dating (230Th) was performed on a multi-collector inductively coupled plasma mass spectrometer (MC-ICP-MS, Thermo-Finnigan-Neptune) at the Department of Geology and Geophysics, University of Minnesota and Institute of Global Environmental Change, Xi’an Jiaotong University (Supplementary Dataset 1). Further 230Th dating on stalagmites So-1 and So-2 was done on a Nu Instruments® MC-ICP-MS at the Geological Institute, University of Bern (Supplementary Dataset 2). Detailed information on analytical procedures is provided in Supplementary Texts 1 and 2 accompanying this article. Age models of all stalagmites were constructed using the StalAge algorithm59 and the Sofular stacked record was developed using the iscam algorithm32.

Stable isotope analysis

Stable isotope analyzes were performed on a Finnigan Delta V and Delta V Plus IRMS equipped with an automated carbonate preparation system (Gas Bench-II) at the Institute of Geological Sciences, University of Bern, and Department of Environmental Sciences, University of Basel. The precision of δ13C and δ18O measurements is 0.06% and 0.07% (1σ-error) respectively.