Weakened AMOC related to cooling and atmospheric circulation shifts in the last interglacial Eastern Mediterranean

Abstract

There is limited understanding of temperature and atmospheric circulation changes that accompany an Atlantic Meridional Overturning Circulation (AMOC) slowdown beyond the North Atlantic realm. A Peqi'in Cave (Israel) speleothem dated to the last interglacial period (LIG), 129-116 thousand years ago (ka), together with a large modern rainfall monitoring dataset, serve as the base for investigating past AMOC slowdown effects on the Eastern Mediterranean. Here, we reconstruct LIG temperatures and rainfall source using organic proxies (TEX₈₆) and fluid inclusion water d-excess. The TEX₈₆ data show a stepwise cooling from 19.8 ± 0.2° (ca. 128-126 ka) to 16.5 ± 0.6 °C (ca. 124-123 ka), while d-excess values decrease abruptly (ca. 126 ka). The d-excess shift suggests that rainfall was derived from more zonal Mediterranean air flow during the weakened AMOC interval. Decreasing rainfall d-excess trends over the last 25 years raise the question whether similar atmospheric circulation changes are also occurring today.

Fig. 1. Study region.

A Locations discussed in the study (Black, red, white and blue points). SCH= Schrattenkarst, COR= Corchia, TR= Trentino, BAR= Baradla and GIB= Gibraltar. B Median sea surface temperatures (SST; °C) calculated for October to April from 2000 to 2022 (blue to red gradient). For comparison the annual averaged global SST is 16.1 °C (years: 1901–2000). The arrow illustrates a deep Cyprus Low (CL) mid-latitude cyclone trajectory after ref. . C Relief map with mean annual rainfall isohyets after ref. ^,. (mm/year; alternating purple regular lines and dashed lines at 100 mm increment’s). Locations of Soreq and Peqi’in, and rainfall sampling sites. Disks coloured according to the multiannual accumulated rainfall d-excess after ref. ^,.

Fig. 2. Last interglacial temperature and marine records.

A Pollen reconstructed July temperatures (T) (Sokli, Finland) after ref. ^,. B δ¹³C(Cibicidoides weullerstorfi) (red line) with 3 pt. moving median curve (black line) and Neogloboquadrina Pachyderma (s) coiling ratio (blue line) after ref. ^,. (MD03-2664, North Atlantic). C Speleothem fluid inclusion (FI) derived temperatures (Schrattenkarst, Switzerland) after ref. . Calculated using various δ²H/°C calibrations: 0.65‰/°C (purple disks and red dashed line), 0.6‰/°C and 0.7‰/°C (purple dashed lines). Temperature error bars reflect isotope measurement errors, δ¹⁸O/annual air temperature slope error, and 1σ of repeated measurements. D Temperatures calculated using FI (red hexagons), speleothem δ¹³C (green triple-cross’s) and estimated CaCO₃ saturation index (SI; orange diamonds) (Cesare Battisti Cave, Italy) after ref. . Temperature error bars consider measurement error, repeated measurement error, and model/calibration equation error. E Fluid inclusion δ²H derived T (cyan triangles and brown polygons for respective BAR-II#L and BAR-II#B speleothems) with 3 pt. moving median (black line) (Baradla, Hungary) after ref. ^,. Temperature errors calculated from error propagation of analytical and δ²H/T gradient calibration uncertainties. Sample age uncertainty reflected by black error bar in the lower left. F Peqi’in TEX₈₆ temperatures from discrete samples (green disks), accumulated samples (blue rectangles), a high methane index sample (grey diamond), and 2 pt. median running curve (red line). Temperature errors were estimated from repeated measurements of a speleothem standard and, for accumulated samples, error propagation. Also shown are clumped isotope (Δ₄₇) temperatures from Soreq Cave speleothems (purple pentagons). The U-Th ages (± 2σ) for the Peqi’in PEK-9 speleothem are shown in the panel below (black disks). Yellow region highlights interval where cooling is evident in the temperature records.

Fig. 3. Peqi’in Cave and Mediterranean records.

A Peqi’in fluid inclusion (FI) water δ¹⁸O (grey disks) and 2 pt. median (black line). For FI isotopes, y-axis error bars represent the ±2σ of standard measurement replicates. B Speleothem δ¹⁸O_calcite from Soreq (green line) and (revised) Peqi’in (brown line)^,. C δ¹⁸O_calcite of Globigerinoides ruber (albus) from Eastern Mediterranean LC21 (orange line). Black arrow illustrates when the Intertropical Convergence Zone shifted southward. D Speleothem δ¹⁸O_calcite from Corchia (red line)^,. E Peqi’in FI δ²H (purple disks) and 2 pt. median (black line); F Peqi’in FI d-excess (blue disks), 2 pt. median (black line), and global meteoric water line (GMWL; dashed line). d-excess_calc calculated from FI δ²H, speleothem δ¹⁸O_calcite, TEX₈₆ and equation by ref. . (cyan dashed line). G Peqi’in TEX₈₆ temperatures from discrete samples (green disks), accumulated samples (blue rectangles), a high methane index sample (grey diamond), 2 pt. median curve (red line) and U-Th ages (lower pane; black disks; ±2σ). Yellow region highlights cooling interval.

Fig. 4. Measured rainfall d-excess at Soreq vs. reanalysis derived relative humidity at 2 m calculated using sea surface temperature (RH_SST).

A Smoothed density histogram of >15 mm rainfall events. Light (dark) regions indicate high (low) data density, respectively. B Accumulated rainfall values of clusters (Supplementary Fig. 2; red diamond= N-cluster, yellow star = NW.-cluster; green square = NW(short)-cluster; blue triangle= W-cluster) and total accumulated rainfall (black disk). The linear regression calculated for these points is shown (black line) as is the comparable empirically-derived linear regression after ref. . (green dashed line).

Fig. 5. Measured rainfall d-excess at Soreq vs. reanalysis derived 2 m air temperature minus sea surface temperature (T_2M – SST).

A Smoothed density histogram of >15 mm rainfall events. Light (dark) regions indicate high (low) data density, respectively. B Accumulated rainfall values of clusters (Supplementary Fig. 2; red diamond= N-cluster, yellow star = NW.-cluster; green square=NW(short)-cluster; blue triangle= W-cluster) and total accumulated rainfall (black disk). Additionally shown are the linear regression of these points (black line and equation) and the linear regression of near surface moisture measurements at Rehovot, Israel after ref. . (purple dashed line). C The two extreme d-excess clusters (meridional N-cluster and zonal W-cluster) and their characteristic trajectories. Specific humidity changes are illustrated by colour changes along the trajectories and emphasize the location of dominant moisture uptake.

Fig. 6. Recent Eastern Mediterranean rainfall d-excess, Mediterranean Oscillation Index and Atlantic Meridional Overturning Circulation index trends.

A Accumulated rainfall d-excess values calculated over three-year intervals for northern rainfall collection sites (yellow disks and line), central rainfall collection sites (red squares and line), and Soreq (black diamonds and line). Superimposed on the plot is the North Atlantic cluster frequency ratio derived from rainfall at Soreq (blue dashed line). B The December to February (DJF) annual mean (grey line) and three-year mean (red and orange lines) of the Mediterranean Oscillation Index (MOI₂). A positive MOI₂ was shown to correlate with a southward shift of the CL tracks reaching Israel^,. C Atlantic meridional overturning circulation (AMOC) indices based on mean sea surface temperature (SST) differences^,. Negative (positive) trends infer a weakening (strengthening) AMOC. SST_SG-GM (blue line): defined as the difference between the mean SST of the subpolar gyre region (SG) and the whole globe. SST_SG-GM-AMO (green line): based on the same spatial regions as SST_SG-GM but with the contribution of the Atlantic Multidecadal Oscillation (AMO) removed. SST_SG-NH (purple line): defined as the difference between the mean SST of SG and the Northern Hemisphere. SST_DIPOLE (yellow line): defined as the difference between South-Atlantic and North-Atlantic SST’s. These indices correlate well with actual AMOC strength in simulations with freshwater hosing and gradual CO₂ increase models.