Lagged atmospheric circulation response in the Black Sea region to Greenland Interstadial 10

Significance

Abrupt climate shifts in the Northern Hemisphere high latitudes during the last glacial propagate globally in a complex manner. Our understanding of this propagation is poor mainly due to cross-dating uncertainties between individual paleoclimate archives. We apply a record of the globally common ¹⁰Be production rate variations to synchronize the Black Sea sediment record to central Greenland ice cores and investigate the hemispheric propagation of Greenland Interstadial 10, with minimized uncertainties in the relative timing. Our results suggest a bipartite climate response in the Black Sea region to the onset of the Greenland Interstadial, characterized by an in-phase decrease in winter severity and lagged shift in atmospheric circulation, explained by a time-transgressive thermal adjustment of the regional ocean interior.

Abstract

Northern Hemispheric high-latitude climate variations during the last glacial are expected to propagate globally in a complex way. Investigating the evolution of these variations requires a precise synchronization of the considered environmental archives. Aligning the globally common production rate variations of the cosmogenic radionuclide ¹⁰Be in different archives provides a tool for such synchronizations. Here, we present a ¹⁰Be record at <40-y resolution along with subdecadal proxy records from one Black Sea sediment core around Greenland Interstadial 10 (GI-10) ∼41 ka BP and the Laschamp geomagnetic excursion. We synchronized our ¹⁰Be record to that from Greenland ice cores based on its globally common production rate variations. The synchronized environmental proxy records reveal a bipartite climate response in the Black Sea region at the onset of GI-10. First, in phase with Greenland warming, reduced sedimentary coastal ice rafted detritus contents indicate less severe winters. Second, and with a lag of 190 (± 44) y, an increase in the detrital K/Ti ratio and authigenic Ca precipitation point to enhanced regional precipitation and warmer lake surface temperatures. We explain the lagged climatic response by a shift in the dominant mode of atmospheric circulation, likely connected with a time-transgressive adjustment of the regional thermal ocean interior to interstadial conditions.

Northern Hemisphere climate during the last glacial cycle (Marine Isotope Stages 2 to 4) was primarily characterized by distinct switches between stadial and interstadial conditions during so-called Dansgaard−Oeschger or Greenland Interstadial (GI) events (1, 2). Abrupt interstadial warming was most pronounced in the high latitudes, occurred within decades or less, and lasted from a few hundred to some thousand years. Cold stadials were characterized by a more stable climate, but comparable ranges of durations (1, 2). Most-often proposed triggers of GIs are the glacial reinvigoration of the Atlantic Meridional Overturning Circulation due to fresh water perturbations and “spontaneous” reductions in sea ice cover in the Nordic Seas associated with thermohaline convective instability causing the release of accumulated oceanic heat to the atmosphere (2–5). Subsequently, these major shifts in high-latitude climate are expected to evolve globally in a complex manner, resulting in a range of region-specific expressions with possible lags in timing, and varying lengths, shapes, or amplitudes (2, 6). Proposed causal links for this behavior are interactions between fast atmospheric as well as slower cryospheric and oceanic processes (2, 6–8). However, the detailed progression of these abrupt climate shifts as well as their driving mechanisms are still poorly understood.

Main limitations for a better understanding of progressing GI climate variations are cross-dating uncertainties between individual environmental archives and the common practice of climate wiggle matching, impeding the detection of possible climate diachrony from the start (9–11). A way to circumvent these limitations is a climate-independent synchronization of the considered archives. In addition to tephra layers and atmospheric trace gases, cosmogenic radionuclides like ¹⁰Be can serve as climate-independent synchronization tools for natural archives (12, 13). The ¹⁰Be is produced in Earth’s upper atmosphere through cascades of nuclear reactions during collisions of galactic cosmic rays with N and O atoms (14). While the galactic cosmic ray flux outside the heliosphere is assumed to be stable over the last million years, the flux arriving at Earth is modulated through varying heliomagnetic and geomagnetic shielding (14, 15). Therefore, during periods of reduced (higher) solar activity and geomagnetic field strength, more (less) ¹⁰Be is produced. After its production, ¹⁰Be is globally mixed during its about 1-y stratospheric residence time and deposited on Earth’s surface through wet and dry deposition (16). Detecting and aligning the external ¹⁰Be production signal in environmental archives like sediments and ice allows us to synchronize these records via curve fitting methods (12, 13). Prerequisite is the assessment and removal of possible nonproduction variability introduced into ¹⁰Be records mainly by varying meteorological conditions during its about 1-mo tropospheric residence time and catchment processes (16–18).

Existing studies on the evolution of major Northern Hemispheric climate switches based on synchronized environmental records are noncomprehensive. A case study synchronizing two lake sediment records from central Norway and western Germany, using the Icelandic Vedde Ash isochrone (∼12.15 ka BP), suggests that a major shift toward windier conditions within the Younger Dryas cold period occurred locally abruptly, but time-transgressively with an offset of ∼120 y between both locations (6). Since the recorded proxy shifts are located close to the tephra layer in each archive, the chronological uncertainty connected with this result is likely low. Another study of central Greenland ice cores and tropical speleothems synchronized based on cosmogenic radionuclides suggests that climate shifts during the last glacial connected with GIs occur simultaneous, within errors of ± 189 y (19). Further studies applying synchronized paleoclimate records are required to understand the progression of high-latitude climate shifts throughout the Northern Hemisphere at down to multidecadal scales.

Synchronizing the Black Sea Sediment and Greenland Ice Core Records

Here, we provide constraints on the timing and succession of major climate changes in the Northern Hemisphere connected with GI-10 (∼41 ka BP) (Fig. 1). We measured a ¹⁰Be record at <40-y resolution along with environmental proxy records down to subdecadal step size from Black Sea sediment core M72/5-22-GC8 (SI Appendix, Supporting Information Text and Figs. S1 and S2). All time series were measured from the same sediment core to avoid chronological uncertainties introduced through core-to-core correlation of proxy records. To isolate the ¹⁰Be production rate and remove environmentally induced variability, we apply regression analysis between our ¹⁰Be record and the X-ray fluorescence (XRF) derived Ca, K, Ti, and coastal ice rafted detritus (IRDc) proxies representing major environmental shifts in the Black Sea region (Black Sea and large parts of its catchment) as well as ¹⁰Be/⁹Be normalization (13, 20) (SI Appendix, Fig. S3). The resulting estimates of the ¹⁰Be production rate (¹⁰Be_atmosphere) resemble each other and the original ¹⁰Be concentration record from Black Sea sediments within the studied time interval, indicating that environmental effects on ¹⁰Be deposition are minor (SI Appendix, Fig. S3). Further indication for the preservation of the globally common cosmogenic radionuclide production signal is the multicentennial double peak in ¹⁰Be, also present in the paleointensity record from Black Sea sediments around the Laschamp geomagnetic excursion (21–23) (SI Appendix, Fig. S3). However, since this structure is not that pronounced in the ¹⁰Be flux time series from Greenland ice cores, we cannot rule out remaining uncorrected environmental influences on ¹⁰Be deposition. Similarities between our ¹⁰Be_atmosphere records and ¹⁰Be fluxes from the central Greenland GRIP and GISP2 ice cores on multidecadal scales could possibly point to a solar modulation of ¹⁰Be production (Fig. 2). This result is supported by significant oscillations in the ¹⁰Be record around 90, 160, and 210 y, likely resembling known solar activity cycles (24, 25) (SI Appendix, Fig. S4). However, the low significance values for the 160- and 210-y oscillations point to a slight remaining nonproduction variability in the ¹⁰Be_atmosphere records (26). The preserved multidecadal cosmogenic radionuclide production rate variations suggest a minor influence of residence times in the water column and bioturbation on our ¹⁰Be record. Both effects are expected to smooth out these types of signals (27).

Fig. 1.

Proposed succession of climate events connected with the onset of GIs in the Black Sea region. (A) Initial situation during Greenland stadials. (B) Northward shift of the polar front following sea ice retreat causing reduced cold spells in the Black Sea region at the onset of GIs. (C) Lagged atmospheric circulation shift at the Black Sea associated with intensified cyclonic activity after interstadial adjustment of regional thermal ocean interior. The position of Black Sea sediment core M72/5-22GC8 is highlighted. Sea ice boundaries (46) and ice sheets extents (47) are indicated.

Fig. 2.

Synchronized ¹⁰Be records from Black Sea sediments and Greenland ice cores around GI-10 based on curve fitting. (A) Synchronizations of the individual environment-corrected ¹⁰Be_atmosphere and ¹⁰Be/⁹Be records from Black Sea sediments to the ¹⁰Be flux time series from the central Greenland GRIP and GISP2 ice cores on the GICC05 timescale (19) using the global optimal fit and the MATCH software (28). All records were normalized between 0 and 1. (B) Age difference between the Black Sea ¹⁰Be_atmosphere records and ice core ¹⁰Be flux. Maximum differences between the individual synchronizations are considered as synchronization error ranges. Colors in B correspond to those in A.

Our environment-corrected ¹⁰Be_atmosphere records from Black Sea sediments were synchronized to ¹⁰Be fluxes from central Greenland ice cores (19) using the global optimal fit and allowing for up to 25% chronological and sampling uncertainty (28) (Fig. 2 and SI Appendix, Supporting Information Text). To test the robustness of our results and provide error ranges, the ¹⁰Be_atmosphere records were individually synchronized to the ¹⁰Be flux time series from central Greenland ice cores (Fig. 2). Determined mean chronological offset between the Black Sea sediment and central Greenland ice core record on the Greenland Ice Core Chronology 2005 (GICC05) timescale around 41 ka BP is 155 y (−26/+11 y average uncertainty), whereby our sediment record leads the ice cores (Fig. 2). This is lower than the estimated 250-y younger age for the Greenland ice core GICC05 timescale around 41 ka BP compared to U/Th dated speleothems (19).

Delayed Climate Response in the Black Sea Region to GI-10

Corrected for the determined offset, the synchronized environmental records point to a bipartite climate response in the Black Sea region to GI-10. First, in phase with the abrupt temperature increase in Greenland, coastal sea ice decreases in the southern Black Sea during less severe winters, reflected by reduced sedimentary IRDc contents (29) (Fig. 3A). IRDc is composed of fine detrital grains incorporated in coastal sea ice formed along the southern Black Sea coast during severe winters with major polar cold spells (21). Subsequent ice drift and melting results in the offshore release of the material (21, 29).

Fig. 3.

Synchronized paleoclimate records from Black Sea sediment core M72/5-22GC8 and the Greenland NGRIP ice core (2) around GI-10. (A) Black Sea IRDc and NGRIP δ¹⁸O. (B) Black Sea K/Ti ratio (reversed axis) and NGRIP δ¹⁸O. (C) Black Sea Ca (reversed axis) and NGRIP δ¹⁸O. The horizontal spread of the gray Black Sea graphs depicts synchronization uncertainties.

Second, and with a lag of 190 (±44) y, abrupt increases in the sedimentary K/Ti ratio indicate enhanced regional precipitation causing higher riverine sediment supply into the basin (29) (Fig. 3B). K originates from weathering feldspars and mica and is transported into the Black Sea predominantly from the northern Danube, Dniester, and Dnieper Rivers (29). In addition, coinciding increases in Ca contents point to warmer temperatures (Fig. 3C). Enhanced authigenic carbonate precipitation in the Black Sea is controlled by CO₂ assimilation and pH increase through photosynthesizing phytoplankton in warmer surface waters (21, 30). A temperature-dependent carbonate precipitation during GI-9 to GI-11 is confirmed by in-phase changes between Ca and the TEX₈₆ paleothermometer on the multicentennial scale (31) (SI Appendix, Fig. S5). Coinciding changes in arboreal pollen contents in Black Sea sediments further support the interpretation of the Ca and K/Ti ratio proxies to reflect warmer and wetter climate conditions (30) (SI Appendix, Fig. S5). However, the direct application of the TEX₈₆ and arboreal pollen records for investigating leads and lags in the climate system connected with GIs is hampered due to their multicentennial resolution and the chronological uncertainties that would be introduced through matching data from multiple sediment cores.

To confirm the lag pattern seen in our Black Sea sediment proxies during GI-10, we measured the same set of IRDc, Ca, and K/Ti ratio data for the neighboring GI-9 and GI-11, also from sediment core M72/5-22-GC8 (Fig. 4). Supporting our results for GI-10, during both interstadials, the abrupt increases in the K/Ti ratio and Ca lag the decrease in IRDc (∼50 y for GI-9 and ∼200 y for GI-11) (Fig. 4). This systematic pattern provides confidence in the robustness of the detected succession of events. Furthermore, we rule out sampling artifacts as a reason for the proxy offsets, since all data were measured from the same Black Sea sediment core (SI Appendix, Fig. S2). A lagged in-lake response of the K/Ti ratio to one environmental forcing is also very unlikely, due to the external origin of the detrital material.

Fig. 4.

Proxy responses in Black Sea sediments and the Greenland NGRIP ice core (2) during GI-9 and GI-11. Shown are IRDc as well as K/Ti ratio and Ca from Black Sea sediment core M72/5-22GC8 in comparison with δ¹⁸O from the Greenland NGRIP ice core. The δ¹⁸O record from the Greenland NGRIP ice core was shifted to match the IRDc time series from Black Sea sediments. Comparable to GI-10, increases in the K/Ti ratio and Ca lag the decrease in IRDc in Black Sea sediments (light gray area), and changes in δ¹⁸O resemble those in IRDc (total gray area). Colors on Bottom correspond to those on Top.

Regional Oceanic Adjustment to Interstadial Conditions

Synchronizing our high-resolution sedimentological proxy records to the δ¹⁸O time series from the central Greenland NGRIP ice core on the GICC05 timescale allows us to put the succession of climate events in the continental Black Sea region connected with GI-10 in context with those from Greenland, with minimized relative chronological uncertainties (Fig. 3). The decrease in coastal sea ice on the Black Sea during less severe winters synchronous with Greenland warming at the onset of GI-10 points to an in-phase atmospheric transfer and deep inland propagation of the high-latitude temperature signal within years or decades (Fig. 3) (32). In-phase climate switches at both sites support the hypothesis of an abrupt sea ice reduction in the Nordic Seas at the onset of GIs (3, 5), causing both transfer of oceanic heat to the atmosphere and a northward displacement of the polar front, reducing the effects of polar cold spells in the Black Sea region (Fig. 1).

This initial reduction in winter severity is followed by an increase in regional precipitation and temperature 190 (±44) y later, reflected by higher K/Ti ratios and Ca contents in the investigated Black Sea sediments (Fig. 3), best explained by a switch in the dominant mode of regional atmospheric circulation (Fig. 1). Previous studies on Black Sea sediments indicate transitions from stadials to interstadials as accompanied by switches from more continental atmospheric circulation patterns, dominated by frequent polar cold spells, to more oceanic ones characterized by intensified traversing Mediterranean cyclones, enhanced precipitation, and warmer temperatures (Fig. 1) (29). A possible cause behind the now detected lag might be a time-transgressive thermal adjustment of the regional North Atlantic, as well as Mediterranean and Black Sea interior, to the interstadial annual mean warming of ∼3 °C (31).

A comparable lag pattern of, on average, 131 y was detected between δ¹⁸O and δ¹³C in the Sofular Cave speleothem from northern Anatolia, in close proximity to the Black Sea (33). This lag was interpreted to reflect a delayed vegetation response to an initial regional climate shift associated with the onset of GIs (33). However, chronological uncertainties inhibit the investigation of phase relationships between the Sofular speleothem and Black Sea sediment record.

Our results reveal the importance of integrating high-resolution, multiproxy, and synchronized paleoclimate records for providing novel insights into the complex and time-transgressive progression of major climate shifts from the high-latitude Northern Hemisphere, which are commonly assumed to be hemispherically synchronous. They reconcile previous studies suggesting either synchronous or asynchronous climate shifts, by revealing a bipartite climate response to GI-10 in the Black Sea region, characterized by a shift in atmospheric circulation ∼190 y after an initial decrease in winter severity. We propose a time-transgressive thermal adjustment of the regional ocean (and Black Sea lake) interior as trigger of the delayed atmospheric circulation change.

Methods

Sediment Core and Initial Chronology.

Sediment core M72/5-22-GC8 (42°23′N, 36°50′E) was retrieved from the Archangelsky Ridge in the southeastern Black Sea at a depth of 847 m below sea level using a gravity corer (SI Appendix, Fig. S1) (21). In this study, we focus on the sediment interval from 435- to 543-cm core depth covering GI-9 through GI-11. Initial chronological constraints for the SE Black Sea sediment stratigraphy were derived for the time interval 30 ka BP to 60 ka BP through tuning sedimentary IRDc and carbonate content records to δ¹⁸O from the Greenland NGRIP ice core on the GICC05 timescale, all reflecting GI-like climate variability (21) (SI Appendix, Fig. S6). Tuning results are supported by 16 bivalve accelerator mass spectrometry (AMS) ¹⁴C dates between 12.5 ka BP and 38.9 ka BP (21). The radiometrically determined ages of the Campanian Ignimbrite, 39.28 ± 0.11 ka BP (34) and 39.85 ± 0.14 ka BP (35), are in broad agreement with its appearance in the Black Sea stratigraphy of a neighboring sediment core at 39.40 ka BP (21). However, due to a hiatus between 20 ka BP and 39.4 ka BP, the Campanian Ignimbrite tephra is not present in the sediment core used here, M72/5-22-GC8 (21).

Be Extractions and Measurements.

For the construction of ¹⁰Be and ⁹Be records, a series of samples was collected from Black Sea sediment core M72/5-22-GC8. The central part of the Be time series around GI-10 (460- to 508-cm core depth) was continuously sampled in 1-cm steps resulting at a resolution of < 40 y. For the youngest (445- to 460-cm core depth) and oldest (508- to 540-cm core depth) parts, one 1-cm-thick sediment slice was extracted every 4 cm, resulting in a resolution of >100 y.

After the addition of 0.5 mg of ⁹Be carrier, Be was leached from 300 mg of freeze-dried and homogenized sediment aliquots with 8 M HCl at 60 °C for 16 h (36). The undissolved sediment fraction was subsequently removed from the solutions through filtering. Addition of NH₃ and H₂SO₄ triggered the precipitation of metal hydroxides and silicates, which were again separated from the solutions by filtering. Ethylenediaminetetraacetic acid treatment led to the separation of other metals (36). Then, the samples were passed through hydrogen-form ion exchange columns in which Be was retained. Be was removed from the columns through the addition of 4 M HCl, and Be(OH)₂ precipitated with NH₃ at pH 10. The precipitates were washed and dehydrated three times by centrifuging and oxidized to BeO at 600 °C in a muffle furnace (13). After mixing with Nb, the samples were pressed into copper targets for AMS measurements of BeO at the Ion Beam Physics Laboratory of the Eidgenössische Technische Hochschule Zurich. Final ¹⁰Be concentrations were calculated from measured ¹⁰Be/⁹Be ratios, normalized to the S2007N (¹⁰Be/⁹Be = 28.1 × 10⁻¹²) and S2010N (¹⁰Be/⁹Be = 3.3 × 10⁻¹²) reference standards (37).

The ⁹Be contents were determined from the same sediment samples as used for ¹⁰Be measurements by inductively coupled plasma mass spectrometry (iCAP Q, Thermo Fisher Scientific) using external calibration and Rh as internal standard to compensate matrix effects. Before the measurements, ⁹Be was leached from the material with 10 mL of 15 vol% HCl. Precision and accuracy of ⁹Be measurements based on acid digestions of the sediment reference material SGR-1b (US Geological Survey) were 3.6% and 5.0%, respectively. To investigate the performance of our extractions, we conducted a sensitivity test (SI Appendix, Fig. S7).

XRF Scanning and IRDc.

Geochemical XRF measurements at 500-µm resolution were conducted on Black Sea sediment core M72/5-22-GC8 using an ITRAX core scanner at 30 kV, 55 µA, and 5 s dwell time. The device was equipped with a Cr tube and an energy-dispersive silicon drift detector (38). To reduce matrix effects and allow a linear interpretation of the data in terms of changing sediment composition, measured element counts were log-ratio transformed (39). Effects of sediment microdisturbances were circumvented using the best preserved of six parallel line scans, distributed equally across the core diameter. IRDc contents were determined from sediment core M72/5-22-GC8 at 0.5-cm resolution by counting detrital grains >150 µm from a fixed volume of 4.5-cm³ wet-sieved sediments (21).

Estimating the Atmospheric ¹⁰Be Production Signal.

Two approaches were applied to isolate the ¹⁰Be production signal (¹⁰Be_atmosphere) and minimize nonproduction variability in our ¹⁰Be concentration (¹⁰Be_con) record from Black Sea sediments. First, four linear regression models were calculated between ¹⁰Be concentrations and the parallel Ti, Ca, K, and IRDc time series from sediment core M72/5-22-GC8, reflecting major environmental changes in the Black Sea region (13, 21, 29). Then, the common variability was subtracted from the original ¹⁰Be_con time series, to produce environment-corrected estimates of the ¹⁰Be production rate (¹⁰Be_atmosphere) (40).

Second, we applied ¹⁰Be/⁹Be normalization based on coupled Be measurements on the same sample material from Black Sea sediment core M72/5-22-GC8 (41, 42). In contrast to the atmospheric production of ¹⁰Be, the stable isotope ⁹Be originates mainly from weathering silicate rocks. However, once released into a waterbody, both isotopes are scavenged from the water column by settling particles with a similar efficiency (43). Therefore, the shared variance of sedimentary ¹⁰Be and ⁹Be concentrations is expected to reflect catchment effects, while the ¹⁰Be/⁹Be ratio is a proxy of the cosmogenic radionuclide production rate (41).

Timescale Synchronization.

For the synchronization to the ¹⁰Be flux time series from the central Greenland GRIP and GISP2 ice cores, we only use the part of the Black Sea ¹⁰Be_atmosphere and ¹⁰Be/⁹Be records at <40-y resolution, producing acceptable <100-y synchronization error ranges. Our ¹⁰Be_atmosphere estimates from Black Sea sediments were synchronized to the ¹⁰Be flux record from the central Greenland GRIP and GISP2 ice cores on the GICC05 timescale through automated wiggle matching using the MATCH software (28). MATCH calculates all possible alignments between two time series to find the optimal. Therefore, the time series are divided into small segments, and an alignment score is calculated for each possible matching. This score is defined as the square of the difference between the two signals (28). Evaluating a wide range of matches precludes the algorithm from getting trapped in a local solution. Geological realism is assured through penalty functions (28). In this study, we allow for up to 25% changes in relative sedimentation rate to account for chronological and sampling uncertainties. Synchronization error ranges are defined as the maximum differences between the five individual synchronizations.

Spectral Analysis.

Spectral analysis was applied to the average of the five ¹⁰Be_atmosphere realizations on the original Black Sea timescale using fast Fourier transform (44). An autoregressive model was fitted to the data, and 10,000 random versions of this model were used to estimate the spectral significance (45). Before the analysis, the ¹⁰Be_atmosphere time series was detrended and equidistantly resampled to 40-y resolution.

Data Availability Statement.

All study data are included in the article and SI Appendix.