Introduction

To understand hemisphere-wide climate variability during glacials, profound knowledge of former oceanic and atmospheric coupling in a zonal and meridional perspective is essential requiring well-preserved palaeo-archives that are widely distributed over the oceans and continents. While especially records along the North Atlantic recording last glacial orbital- and millennial-scale climate variability are available, high-resolution continental records from the last glacial are still sparse. However, just the deeper continental palaeoclimate records may allow retracing high to mid-latitude teleconnection patterns of the atmospheric circulation system tightly coupled to the size of the Eurasian Ice Sheet during glacials. The former glacial isolated Black Sea, bridging the North Atlantic-Eurasian corridor, is one of such inland candidates. It could help to extend our view of high- to midlatitude atmospheric coupling during glacial climate variability.

During the relatively mild period within the last glacial between 57 and 29 ka BP, the Marine Isotope Stage 3 (MIS 31), temperatures fluctuated on a millennial time scale during the so-called Dansgaard–Oeschger cycles (DO cycles) with abrupt warming into interstadials and gradual cooling into stadials across the North Atlantic–Mediterranean corridor2,3,4,5,6,7. During MIS 3, the largely enclosed, lacustrine Black Sea experienced pronounced DO climate and environmental variability as well, which were tightly linked to changes in the North Atlantic Meridional Overturning Circulation (AMOC) and temperature variations in Greenland7,8,9. Temperature amplitudes between stadials and interstadials in the Black Sea were as high as 4 °C7. During the warmer interstadials, the amount of rainfall and riverine freshwater supply into the basin increased8,9. The increased freshwater contribution resulted in increased lake level and lower salinity, whereas higher nutrient supply and warmer conditions favoured primary productivity. In Northern Anatolia, temperate forest expanded and wind strength was reduced8,9. During the colder stadials, the hydroclimate gradually changed from humid to arid conditions, which resulted in increased evaporation, a lowered lake level, and higher salinity. Along with reduced riverine supply of nutrients, primary productivity decreased. During stadial winters, temperatures were much lower than during interstadial winters resulting in ice formation at the southern coast of the Black Sea ‘Lake’7,8,9,10. On land, the vegetation in Northern Anatolia became dominant in xerophytic steppe8. In a long-term perspective, similar changes in temperature and especially in rainfall amounts on an orbital timescale are superimposed to the DO-cycles with much more humid conditions during the first half of MIS 38,9. During this time window it is supposed that the enhanced supply of freshwater by rainfall and rivers caused a rising lake level of the Black Sea that probably resulted into an overflow into the Marmara Sea until ca. 42 ka BP9,11,12,13.

Previous studies have contributed to the knowledge about the changes in temperature and rainfall amount and the resulting environmental implications for the Black Sea ‘Lake’ and Northern Anatolia during MIS 37,8,9,10. In addition to the Black Sea sediment record, the δ18O records of speleothems from Northern Anatolia document these climate cycles exceptionally well and are suggested to reflect changes in the δ18O composition of the Black Sea surface water14. However, there is no clear explanation for the underlying processes and mechanisms of these apparent changes in the δ18O composition of the Black Sea surface waters during the intermediate climate state of Marine Isotope Stage 3 (57–29 ka BP). Therefore, it has to be unravelled how these changes relate to the variability of temperature, the amount and sources of rainfall and riverine supply as well as to the hydrological activity in the Black Sea itself. With a deeper knowledge of the coupling between hydroclimate variability and hydrological activity, it may help understanding continental high- to mid-latitude teleconnection patterns of the atmospheric circulation during glacials. Here, we present a new continuous high-resolution δ18O record obtained from benthic ostracod carbonate shells from a well-dated SE Black Sea sediment core in order to investigate the hydrological dynamics in the former Black Sea upper water column and their relation to migrating air mass trajectories subjected to hydroclimate reorganisations. In combination with a well-dated δ18O record from speleothems of the Northern Anatolian Sofular Cave that has been suggested to mainly reflect the δ18O composition of the former Black Sea surface water during the period of interest14,15, we will propose potential atmospheric-hydroclimatic processes explaining the similar and diverging trends in both records. We further will elaborate on why the δ18O record has a significant orbital timescale component and how this is related to the Eurasian Ice Sheet (EIS) dynamics and changes in atmospheric circulation patterns over Eurasia. A simple hydrologic-isotope-balance model will show that the δ18Oostracods pattern reflects an integrated central European precipitation signal with its imprint strongly affected by hydrological activity and mixing times of the large water volume of the Black Sea. A new 87Sr/86Srostracods record obtained from the same sediment core, will serve for estimating changes in the dominant source of riverine supply that might be partially associated with meltwater release from the EIS in turn potentially affecting the δ18O composition of the Black Sea.

Today, the Black Sea is a semi-enclosed, brackish basin and represents the largest anoxic water body on Earth16,17. It receives freshwater by rivers and rainfall and saltwater by the inflow of water from the Mediterranean and Marmara Seas through the Dardanelles and Bosporus straits17. Due to the higher input of freshwater by precipitation (P) and riverine runoff (R) compared to the loss of water by evaporation (E), the modern Black Sea has a positive water balance (P + R > E)17.

Controlled by the northern hemispheric atmospheric circulation over Europe, the climate in the Black Sea region is mainly controlled by the Polar Front Jet, the mid- to high latitude westerlies, the Subtropical Jet, and the Siberian and the mid-latitude subtropical high pressure systems18,19 (Fig. 1). The dynamics and intensity of Mediterranean cyclones arriving in the Black Sea are affected by the East African and Indian monsoons in combination with the Siberian high-pressure system and relate to the positions of the Polar Front and Subtropical Jet18,19. With dry summers and wet winters, the Black Sea climate is generally under Mediterranean influence and its interannual variability is strongly linked to the North Atlantic Oscillation20,21,22,23,24,25. Mediterranean atmospheric moisture mainly originates from the North Atlantic, but evaporation of Mediterranean waters also strongly controls the isotopic composition of rainfall in the Black Sea drainage basin26.

Fig. 1: Map showing Eurasia.
figure 1

Location of the coring site of M72/5 25GC-1 in the Black Sea (yellow star), and the maximum extent of the Eurasian Ice Sheet during the Weichselian (blue shading; MIS 2 and 3; modified from Larsen et al.73). Idealised scheme of the modern atmospheric circulation patterns with broad position of the polar front during January and the Subtropical Jet during July influencing regional climate in the Black Sea region (modified after Wigley and Farmer18; Akçar and Schlüchter19). Yellow numbers show values for the δ18O composition of modern rainfall53. AeS = Aegean Sea, BS = Black Sea, CS = Caspian Sea, MD = Manych Depression, MS = Marmara Sea, SC = Sofular Cave (Map modified after Wegwerth et al.39).

Full size image

Northeastern Anatolia and the eastern Black Sea are the most summer rain-laden Black Sea regions due to orographic precipitation in the Pontic and Caucasian Mountains27,28,29, whereas the Black Sea itself forms an additional moisture source for rainfall14,15,24.

On an over-regional scale, the δ18O composition of rainfall across Eurasia decreases progressively from the mid-latitudes to northern latitudes and generally eastwards to the continental interior (Fig.1) due to the Rayleigh condensation30,31,32. The δ18O composition of rainfall in the Black Sea region ranges today from −4‰ to −8‰33, whereas other studies reported even values of around −9‰34. Additionally to the rainout effect with depletion towards northeast and the continent, the altitude of the NE Anatolian Mountains also affects the rainfall isotopic composition leading to values of about −10‰26. Today, in the Black Sea the δ18O composition of the surface water ranges between −2.44‰ and −2.84‰15,33,35,36, whereas the deeper water is slightly heavier and reveals a composition between −1.65‰ and −1.8‰33,35,36. However, it has to be noted that during strong precipitation periods, the oxygen isotopic composition of the surface water can decrease to −7.1‰33, demonstrating the pronounced sensitivity of the Black Sea to variations in freshwater input. With 50–70% of the total riverine input, the Danube river is the main freshwater source from all rivers draining into the Black Sea today17,37. Near the river mouth, the Danube reveals a δ18O composition of −10.5‰36. While the freshwater contributed by rivers and rainfall is isotopically highly depleted, the high-salinity Mediterranean water entering the Black Sea has a δ18O composition of about +1.8‰33.

Results

The presented data derive from the well-dated10 gravity core M72/5 25GC-1 recovered at the Archangelsky Ridge in the SE Black Sea (42°06.20′N, 36°37.40′E) in a water depth of 418 m (Figs. 1 and 2).

Fig. 2: Core chronology.
figure 2

Age-depth model of sediment core M72/5 25GC-1 between 20 and 65 ka BP, error bars define age uncertainties (slightly modified from Nowaczyk et al. and Wegwerth et al.7,9,10).

Full size image

δ18Oostracods

The Black Sea δ18Oostracods record covering the last glacial from MIS 4 to MIS 2 (64–20 ka BP; Fig. 3b) shows values ranging between −2.7‰ and −6.1‰ (Supplementary Data 1). During the first half of MIS 3, the values of δ18Oostracods are less depleted (or show maximum values) reaching a level similar to the last two glacial–interglacial transitions38,39. Else than several other parameters already presented from the same sediment core for that period7,8,9,10, the δ18Oostracods apparently do not reveal only the typical millennial-scale Dansgaard–Oeschger pattern. Instead, the δ18Oostracods signature clearly shows four cycles with sharply increasing (lasting 1–3 kyrs) and gradually decreasing (lasting 3–8 kyrs) values with minima occurring close to the onset of North Atlantic Heinrich stadials. From MIS 4 towards MIS 2, δ18Oostracods values are getting lighter during each cycle finally culminating in minimum values around the last glacial maximum (LGM) at 23 ka BP.

Fig. 3: Climate and environmental conditions in the Black Sea during the last glacial.
figure 3

a δ18ONGRIP62,63,64,65 (Greenland) b δ18Oostracods (original and modelled (Supplementary Information), this study); c TEX86-temperature corrected δ18Oostracods (this study); d TEX86-based lake surface temperature7 (LST); e δ18OSofular Cave (stalagmites So-1, So-2, and So-414,15); f relative sea level (light blue: Arz et al.66; dark blue: Grant et al.67); g Sr/Caostracods indicating changes in salinity9 (raw data and simple 21-point moving average); h 87Sr/86Srostracods indicating changes in river sources (this study), error bars define standard errors, estimation of relative changes in the contributions between Danube (0.7089)50 and Dnieper (0.7084)50; i Eurasian ice volume45 (m sea level equivalent). The red numbers and grey bars denote the Greenland Interstadials (GI) during Dansgaard-Oeschger cycles and the blue numbers denote Heinrich event (HE) equivalent periods.

Full size image

87Sr/86Srostracods

The Sr-isotope record of the ostracods shows a relatively low variability ranging between 0.708546 and 0.708789 (average 0.708670; n = 35) with minimum values occurring during mid-MIS 3 and maximum values during MIS 4 and MIS 2 (Fig. 3h; Supplementary Data 1). Apart from some outliers during the middle part of MIS 3 between 40 and 45 ka BP, the 87Sr/86Srostracods pattern bears resemblance to the long-term pattern of the Sr/Caostracods record, with its minimum level interpreted to reflect a freshening phase during MIS 39 (Fig. 3g).

Discussion

Potential factors changing the Black Sea δ18Oostracods composition during the last glacial

The high-resolution δ18Oostracods record presented here on a centennial timescale reveals at first glance a subdued nature of DO variability suggesting that it neither reflects simply temperature nor rainfall amount since both abruptly changed during last glacial stadial-interstadial cycles in the Black Sea region7,8,9. The variations in the δ18Oostracods record represent a combined response to changes in the δ18O composition of the water and in the ambient temperature during the growth of the ostracod shells. The δ18O composition of the glacial lacustrine Black Sea water column, in turn, was affected by changes in a) evaporation of the lake surface and b) the amount, sources, and δ18O composition of freshwater input from direct precipitation at the lake surface and indirectly via rivers arriving from the Black Sea drainage basins. The latter isotopic signature comes from a mixture of precipitation and potential meltwater from adjacent glaciers and ice sheets feeding the rivers38,40,41. In case of a connection to the Black Sea via the Manych depression, Caspian water could have additionally contributed to an altered (lighter) δ18O composition because rainfall as well as river Volga (δ18O: −12.5‰) contributing by 80% of total riverine discharge into the Caspian Sea are isotopically very light15,42. Otherwise, a connection to the Mediterranean Sea would have led to increased δ18O values38.

In addition to the above mentioned potential external factors, internal features such as long residence times of the water could modulate the δ18O composition of the ostracods. Thus, longer mixing times of the water column associated with reduced freshwater input may cause an overprint of the pattern of the δ18Oostracods resulting in an orbital-scale pattern by the loss of an original millennial-scale climate component, especially for mixing times that are on the scale of individual DO-cycles.

When affected only by temperature, the range of about 3.4‰ in the δ18Oostracods would suggest temperature amplitudes of about 13.6 °C during the investigated period43,44. However, temperature changes between stadials and interstadials reached only up to 4 °C during the period 64–20 ka BP as revealed by reconstructions of TEX86-based annual lake surface temperatures7 (Fig. 3d). Similarly, estimations of temperature changes in the intermediate water depth of ~400 m by Mg/Caostracods imply temperature amplitudes of 2–4 °C during the DO-cycles7. This circumstance and because the record a) lacks clear millennial-scale DO variability and b) shows generally higher values during warmer periods although the opposite pattern (negative correlation) would be expected, the temperature impact on the δ18Oostracods was most likely very low43,44.

By considering a relation of 0.25‰ decrease in δ18O per rise of 1 °C in temperature43,44, we corrected the δ18Oostracods values by removing this temperature effect using the TEX86-based lake surface temperatures from the same sediment core7 to unfold a potential DO variability not coupled to temperature. Although this lake surface temperature correction relative to the conditions at the ostracod’s habitat depth at 418 m represents an overcorrection, the temperature-corrected δ18Oostracods values are very much the same as the non-corrected values (Fig. 3c) confirming another superordinate environmental factor for the δ18Oostracods composition of the Black Sea ‘Lake’.

Similar to an enhanced temperature, increased input of freshwater by rivers including meltwater discharge would result in lowered δ18O values. Previous studies have demonstrated a pronounced variability in the amount of riverine input on orbital- and millennial timescales from different proxies from the same sediment core presented in this study8,9. Since riverine discharge includes both direct rainfall and meltwater, we try to unravel the impact of these different freshwater sources on the δ18Oostracods record in this chapter. Irrespective of the freshwater source, an increase must have lowered the δ18O composition of the water in the Black Sea basin. On a millennial timescale, rainfall was much higher during the warmer interstadials as indicated by an increased arboreal pollen abundance reflecting the expansion of moisture-dependent temperate forest vegetation8. The related riverine discharge likely lowered the surface water salinity as mirrored by decreasing Sr/Caostracods and increased freshwater dinoflagellate assemblages during interstadials8,9 (Fig. 3g). Furthermore, the amount of leaf wax derived odd-numbered long-chain n-alkanes along with the sedimentary ln(K/Ti) ratio, both reflecting relative changes between riverine and aeolian input, support the observation of an amplified riverine discharge during interstadials9.

In addition to the DO-patterns, the long-term patterns of the aforementioned proxies suggest a pronounced orbital-scale component with increased freshwater supply during the first and warmer half of MIS 3 strongly resembling the pattern of the coevally reduced Eurasian Ice Sheet volume45 (Fig. 3i). Therefore, in addition to rainfall, increased meltwater discharge originating from the Eurasian Ice Sheet could have affected the hydrology and salinity of the former Black Sea ‘Lake‘. Similar as during other meltwater pulses during the last two glacial terminations and the deeper penultimate glacial, the Dnieper could have played an increasing role in freshwater discharge relative to the Danube39,46,47. A reliable tool for estimating changes in river sources for the glacial Black Sea is the Sr-isotopic composition of ostracods39,47,48, because it depends on the catchment geology in the respective drainage regions48,49. Our new 87Sr/86Srostracods record (Fig. 3h) shows generally less radiogenic values when compared to the last two glacial-interglacial terminations (T I48: 0.709102; T II47: 0.709452) and the penultimate glacial (MIS 639; max. 0.709405) and lowest values are reached during MIS 3 (0.7085460) with a slightly elevated glacial background during the colder periods (MIS 4 and 2: 0.708789). These values indicate an increased contribution of Dnieper and Don relative to Danube50 including meltwater from the partly disintegrating Eurasian Ice during the earlier part of MIS 345. When disregarding runoff from the remaining smaller rivers, our estimations imply that the contribution by Dnieper relative to Danube could have increased up to 71% during MIS 3, whereas during MIS 4 and 2 the relative contribution by Danube would have amounted up to 78% (Fig. 3h). A certain contribution from Caspian waters (87Sr/86Sr: 0.70821–0.7085) via spill out along the Manych depression (Fig. 1) is possible as well50,51,52. Although not explicitly suggested from the Sofular Cave speleothem δ18O record, the dinoflagellate assemblages from the Black Sea sediment core would be in line with a potential Caspian connection8,15.

Despite the clear evidence for early MIS 3 meltwater contribution, the amount released by the medium sized MIS 3 ice sheet45 was obviously insufficient for decreasing the δ18O composition of the Black Sea water column as it was the case for the meltwater pulses of the last two glacial terminations and the deeper penultimate glacial38,39,47. Comparable trends in the δ18Oostracods, 87Sr/86Srostracods, and EIS volume records (Fig. 3b, h, i), however, suggest changes in the atmospheric circulation over the North Atlantic and Eurasia as a potential trigger of the observed patterns. Alike the Black Sea ostracod record, the δ18O values from the speleothems of Sofular Cave in north-western Anatolia are also heavier during the more humid interstadials than during the stadial periods14,15 (Fig. 3e). While the speleothem data are suggested to be closely linked to the surface water conditions of the former Black Sea ‘Lake’14, it is conceivable that the δ18OSofular Cave was affected not only by the Black Sea as moisture source14, but incorporated a variable portion of heavier atmospheric moisture signal originating in the (eastern) Mediterranean Sea and the North Atlantic.

Since a retreated Eurasian Ice Sheet was most likely associated with a northward migration of the atmospheric polar front, precipitation patterns and moisture sources may have changed as well. The northward migration of the frontal systems over Eurasia possibly caused a progressively stronger influence of the North Atlantic/Mediterranean moisture sources to the precipitation in the Black Sea region and could explain a shift towards a heavier δ18O signature. Arppe and Karhu53 pointed out that present precipitation patterns over Europe with a westerly moisture source and a progressive isotopic depletion in the heavier isotopes of air masses towards northeast existed also during MIS 3. Based on δ18O analyses of mammoth tooth enamel and palaeogroundwaters across Europe, the authors reconstructed the oxygen isotopic composition of rainfall for the period 52–24 ka BP, showing ca. 2‰ lighter values relative to present53. Although there are no reconstructions of δ18O-precipitation patterns over Eurasia available for more discrete intervals or even on a millennial timescale, the general reconstruction for the MIS 3 implies lighter δ18O values in SE Europe when the Eurasian Ice Sheet was present (MIS 3) than it was absent (today)53. This supports the idea of the MIS 3 Eurasian Ice Sheet dynamics as an important forcing factor in modulating the Eurasian atmospheric circulation pattern and moisture sources of the Black Sea region.

According to reconstructions by Bintanja and van de Wal45, the Eurasian Ice Sheet began to expand again since ca. 40 ka BP (Fig. 3i). Accordingly, a repeated southward migration of the atmospheric polar front is indicated from 40 ka BP by increasing wind intensity and drier conditions in SE Europe as reconstructed from grain-size distribution of loess-palaeosol sequences in the Lower Danube Basin54. The authors argue for an increasing role of the Siberian/Eurasian high-pressure system in modulating the continentality towards the peak glacial phase (Fig. 1). Such a gradual change likely reduced the North Atlantic/Mediterranean sourced precipitation in the Black Sea region causing a strong negative trend in the δ18Oostracods record due to increased contribution of rainfall sourced in the Eurasian continental interior. On the millennial scale, the stronger imprint of Heinrich stadials compared to the regular stadials in the δ18Oostracods record would then similarly point to a major reorganisation of the atmospheric circulation with a stronger continentality and much lighter δ18Oostracods. This goes in hand with the expected direct effect from the condensation temperature of moisture sources above the cave and more generally the Black Sea drainage basin with lighter (heavier) δ18O at low (high) stadial (interstadial) temperatures.

Changes in the Black Sea ‘Lake’ hydrology during the last glacial

On the millennial scale the δ18Oostracods record is not following simply the DO pattern, but appears to be subdued and slightly delayed regarding the peak interstadials. This may be caused by variable mixing times of the relatively large Black Sea ‘Lake’ water body in response to the hydrological balance of the basin. Similar to our MIS 3 record, δ18Oostracods records from the western Black Sea of the last glacial maximum and Termination I show rather smoothly increasing values and lack a clear shift during the Bölling-Alleröd warming and the cold Younger Dryas stadial38,55,56.

The present Black Sea water residence time is in an order of 1000 years57 but it might have varied significantly during the last glacial period. It can be assumed that the mixing time in a basin of that size and geometry could have lasted as long as or even longer than an average DO cycle resulting in a significant modification of an original freshwater input signal. The mixing time during the late MIS 3/MIS 2 increased potentially to several millennia in response to the generally aridification trend towards the LGM with a negative hydrological balance in the basin. Higher evaporation and a decreasing Black Sea ‘Lake’ level is indicated e.g., by increased salinities as reconstructed by Sr/Caostracods9, sequence-stratigraphic evidence of an LGM lake level low stand in the Danube fan area58, and significantly higher reservoir ages of up to 1450 years towards the LGM10,59,60.

Following the approach by Bahr et al.38 performing a simplified hydrologic-isotope-balance model (HIBAL61) for the Black Sea since the LGM, we simulated the δ18O composition of the Black Sea for the period 64–20 ka BP by considering potential changes in the amount of precipitation, runoff, and evaporation, which consequently involves changes in the mixing time of the Black Sea water. As a high-latitude rainfall source we used the δ18ONGRIP62,63,64,65 record (Supplementary Methods, Supplementary Table S1, Supplementary Fig. S1). Supplementary Table S1 shows under which specific but simplified boundary conditions during four time slices our simulated δ18O record best mimic the measured δ18Oostracods record (Fig. 3b, c; Supplementary Fig. S1).

Considering a reservoir age of 1450 years for the LGM around 23 ka BP10 lying in our fourth time slice 20–28 ka BP of our HIBAL estimations (Supplementary Table S1), mixing times for the other three times slices can be roughly estimated from the hydrologic input volumes and are ~1500 years for 65–60 ka BP, ~700 years for 60–45 ka BP, and ~1000 years for 45–28 ka BP. These simulations support not only a high-latitude rainfall source, but also long and varying mixing times of the Black Sea water explaining the subdued nature of DO-variability in the intermediate water depth of the glacial, lacustrine Black Sea. For instance, during the earlier part of MIS 3 predominantly more humid conditions and a certain contribution by meltwater8,9 suggest a generally more enhanced hydrological activity. These conditions caused a rising lake level that probably culminated into an overflow into the Marmara Sea9,11,12,13 associated with reduced mixing times and lower reservoir ages10,60 being half as high as during the ending MIS 4 and towards the LGM.

Interestingly, the Bond-cycle like pattern in the δ18Oostracods with a subdued DO-variability is remarkable similar to the pattern of global sea-level changes66,67 (Fig. 3 b, c, f). It was proposed that gradual sea-level rises were related to interstadial warming and resulted mainly from melting of northern hemisphere ice sheets66. Obviously, the oxygen isotopic composition in the isolated Black Sea ‘Lake’ was not directly affected by such changes, but centennial scale changes in the dynamics of the Eurasian Ice Sheet and modified atmospheric teleconnections between the North Atlantic and Eurasia7 might have had an additional imprint on the δ18Oostracods.

The new high-resolution δ18Oostracods record covering the last glacial from 64 to 20 ka BP along with the hydrologic-isotope-balance estimations demonstrates that the hydroclimate regime in the Black Sea region was strongly coupled to the northern hemisphere climate variations. During periods of a retreated Eurasian Ice Sheet, the δ18Oostracods generally increase, which is likely related to the northward migrations of atmospheric frontal systems and a reduced continentality with an increasing influence from North Atlantic/Mediterranean moisture sources for the Black Sea region. Although the decreased 87Sr/86Srostracods values during the milder MIS 3 point to an increased meltwater contribution from the Dnieper, Don, and possibly Volga rivers, the effect is not evident in the δ18Oostracods record. The subdued nature of the Dansgaard–Oeschger pattern recorded in the δ18Oostracods we attribute to the long mixing times of the former Black Sea ‘Lake’ reacting sensitive to hydrological changes in the basin. We suggest generally shorter mixing times in an order of ~700 years in an overflowing and fresher Black Sea in the first warmer and more humid part of MIS 3 and increasing mixing times up to ~1500 years during the later part of MIS 3 and during MIS 2 and late MIS 4 with generally more arid conditions and a lowered lake level.

Methods

Sediment core and chronology

This study bases on the gravity core 25GC-1 recovered during RV Meteor cruise M72/5 in the SE Black Sea in 2007 (Fig. 1; Archangelsky Ridge, 42°06.20′N, 36°37.40′E, 418 m water depth). The sediment core consists mainly of detrital mud intercalated by carbonate-rich intervals. The core chronology10 (Fig. 2) bases on AMS radiocarbon dating, the Campanian ignimbrite tephra (Y5; 39.28 ± 0.11 ka68), the Laschamp geomagnetic excursion (40.70 ± 0.95 ka BP69), and proxy-tuning to the δ18O record from the North Greenland Ice core Project on the GICC05 timescale62,63,64,65 (NGRIP). A study on cryptotephra horizons from the same core largely confirmed the NGRIP-tuned stratigraphy70. The sequence of interest in the present study covering the last glacial from the late MIS 4 to the MIS 2 (64–20 ka BP) derives from core depth 307–952 cm.

Ostracod isotope geochemistry

For the analyses of the δ18O and 87Sr/86Sr isotope signatures of the ostracods, well-preserved valves of adult specimens of Candona spp. were hand-picked from the wet-sieved and dried sediment fraction >150 µm.

For the oxygen isotopic composition of the ostracods, the carbonate valves from 447 samples in total were cleaned with a thin brush and deionised water under a binocular microscope. Then, the pre-cleaned ostracods (20–80 µg) were measured with a Finnigan MAT 253 isotope ratio mass spectrometer coupled with an automated KIEL IV carbonate preparation device (103% H3PO4 at 70 °C) at the GFZ Potsdam, Germany. Oxygen isotope values are given in delta notation relative to Vienna Peedee Belemnite. Repeated measurements of the international reference material NBS 19 and a laboratory internal standard (C1) assured an analytical precision better than ±0.07‰ for δ18Oostracods.

For the analyses of the strontium isotopic composition, 35 samples evenly distributed along the core with each 10–15 ostracod valves were cleaned with a few drops of H2O2 (5%), methanol, and deionised water48,71,72. After dissolution of the valves in 300 µL 2 vol% HNO3, the samples were dried on a hot plate at 80 °C. Strontium was separated from the sample matrix by using the HNO3–H2O technique in microcolumns filled with Eichrom© Sr-spec resin. Isotope analyses were performed with samples re-dissolved in 0.3 N HNO3 and all isotopic beams measured simultaneously in static mode on a MC-ICP-MS (Neptune Plus, Thermo) at the University of Tübingen, Germany. Instrumental mass bias was corrected using exponential law and an 88Sr/86Sr ratio of 8.375209. External reproducibility for NBS SRM 987 was 0.710253 ± 0.000015 (2 SD, n = 16) for the 87Sr/86Sr ratio and total procedural blank was <95.8 pg Sr.