Introduction

Mediterranean climate is characterized by mild wet winters and warm to hot, dry summers. In the Mediterranean basin, summer droughts are secured by the northward migration of North African and Azorean highs, hampering the influx of Atlantic storm tracks. As anticyclones retreat in the fall, westerlies and polar-continental air masses yield abundant winter precipitation1,2,3. Regional rainfall rates also respond to the North Atlantic Oscillation (NAO)4,5. Previous studies emphasized the importance of winter precipitation in shaping the long-term environmental evolution of the northern Mediterranean and Balkan borderlands in the recent geological past. It is assumed that the magnitude of interglacial winter precipitation over the last 1.36 Myr varied in phase with the African monsoon, as the low-latitude insolation forcing on Mediterranean sea-surface temperatures would strengthen the local cyclogenesis and fuel winter storm tracks from North Atlantic low-pressure systems6,7,8. However, small-scale temporal relationships between African Monsoon and Mediterranean precipitation are still poorly understood. Holocene studies question the synchronicity between the two, because the advection of humid Atlantic air over the southern Mediterranean increased regional precipitation since the middle Holocene, considerably after the beginning of the Green Sahara event9,10,11,12,13.

Our study record, which we refer to as the Blatta section (39° 0′ 37.07″ N, 16° 50′ 31.57″ E), is a ca. 23 m-long mid- to outer-shelf Lower to Middle Pleistocene marine succession exposed in the northwestern sector of the Crotone Basin (Southern Italy). It is part of an expanded, shallowing-upward stack of slope to inner shelf sediments encompassing continuously the MIS 26-MIS 16 interval (Supplementary Material 1)14,15. The Blatta section straddles MIS 17, a poorly investigated interglacial if compared to the contiguous MIS 19 and MIS 15-MIS 13, or other recent integlacials16,17. MIS 17 investigation is especially relevant considering that the establishment of 100-kyr glacial cycles took place shortly after18,19. A steady sampling pace of 10 cm was followed across the MIS 17 interval (Supplementary Material 1). The very high resolution and sensitivity of land-sea geochemical and micropaleontological proxies employed for this work shed light on past seasonal hydroclimate variability in the central Mediterranean, which is a crucial information for understanding present and future environmental changes3.

Chronology of the Blatta section

Our age model is based on a linear interpolation between 11 tie-points identified as ‘cold spells’ in the Blatta section, corresponding to “heavy” δ18O values for U. peregrina and/or spikes of N. pachyderma sinistral coiling (sx) (Supplementary Material S2, Fig. S2). Short-lived abundance fluctuations of other planktonic foraminifera species, such as the warm water species Globigerinoides ruber20,21, confirm the presence of a pervasive millennial-scale variability throughout the record (Supplementary Material S3). Target record is the coeval succession of stadial events at the Iberian Margin22 (8-IMS1 and 7-IMS10 to 7-IMS19), where the signal of sub-orbital oscillations in Atlantic Meridional Overturning Circulation (AMOC) strength is sharp23,24,25,26,27,28. It is thus assumed that AMOC slowdown led to a southward shift of the Polar Front and increases in the strength of the Northern Hemisphere atmospheric activity, analogous to the stadials and Heinrich events of the last glacial that are associated with a drop in Mediterranean SSTs, incursions of N. pachyderma through the Gibraltar Strait and increased ventilation of the Mediterranean seafloor24,25,26,28.

Correlation is further validated by the overall match between our benthic δ18O record and the δ18O profiles of the Iberian Margin29 and the LR04 benthic stack16 (Supplementary Material S2, Fig. S2). An age of 715.48 ± 4 ka is obtained for the Parmenide tephra, in excellent agreement with the previous 40Ar/39Ar dating of 710 ± 5 ka15. Calcareous nannofossils cannot provide chronology, as the identified MNN 19f. (concurrence of medium-sized Gephyrocapsa omega and Pseudoemiliana lacunosa) is a long-range Zone30. The calculated resolution for the main sampling pace (10 cm) is between 0.13 and 0.52 kyr, 0.38 kyr on average (Supplementary Material 2). Wide excursions in the local estimated sediment accumulation rates are consistent with the inferred mid- to outer-shelf depositional setting, where sedimentation responds to a complex interplay between sediment supply, climate, eustasy and tectonics13,14.

The hydroclimate regime during MIS 17 in the Mediterranean

The magnitude of monsoonal precipitation responds to summer insolation levels in the tropics, with maximum values under conditions of minimum orbital precession and maximum eccentricity2,31. Although restricted to low-latitude regions, monsoonal precipitation extend their effects to the eastern Mediterranean in the form of riverine runoff, the Nile River being the largest contributor in this regard32,33,34. Extreme precession-related Nile floods are believed to cause a density-driven stratification of the eastern Mediterranean, with deposition of organic-rich layers (sapropels) in the deep domain2,31,35. In shallower depositional settings inadequate to sapropel formation, a geochemical signature of these events is still preserved as negative δ18Oplanktonic and δ13Cbenthic excursions and/or spikes in the Ba/Al ratio2,36,37,38 (Supplementary Material S4, Fig. S4). In the Blatta section, where the lithological evidence of sapropel layers is missing15 (see Supplementary material S5, Fig. S5), two well-defined minima in the δ13C record of U. peregrina (Fig. 1) are in chronological agreement with the peaks in Ba/Al ratio (productivity) and elemental proxy PC2 (river runoff) found in the deep eastern Mediterranean, correlative to sapropel layers S17 (around 714 ka) and S16 (around 692 ka)35, and with indications of increased precipitation at Lake Ohrid7 (Fig. 1). Concomitant increases in benthic foraminifer oxygen-deficiency stress (ODS) species (Fig. 1) confirm that the zonal vertical circulation of the northern Ionian Sea suffered from episodic slowdowns close in time to periods of precession-related maximum African monsoon activity and sapropel deposition in the eastern Mediterranean.

Figure 1
figure 1

Plot of benthic δ13C and benthic foraminifera data at Blatta section and comparison with selected records. From the left, Ba/Al ratio and principal component 2 (PC2, river runoff) at ODP Site 967, eastern Mediterranean Sea32. Insolation at 65°N summer solstice variations (Laskar et al. 2004)39. Simulated precipitations at Lake Ohrid, Balkan region7. Downcore variations of U. peregrina δ13C values and ODS benthic foraminifera species percentage (this study). Non-Arboreal Pollen (NAP) (this study), note reverse axis. Mediterranean-type vegetation from Blatta section, central Mediterranean (pink line, this study) and the IODP Site U1385 (red line18). Downcore variations of U. peregrina δ18O values (this study). Obliquity at 65° N summer solstice variations40. Thick lines in ODP 967, Ohrid Lake and Blatta records are 3-pt running averages. Horizontal dotted black lines indicate MIS boundaries from ref15. Vertical grey boxes indicate stadial phases, progressively labeled, following the procedure by ref19. The timing extent of sapropel S16 follows ref32. The black asterisk marks the position of the Parmenide ash layer.

Full size image

Terrestrial pollen from the Blatta section reveal a long-term vegetational trend that follow very closely the benthic δ18O record, suggesting that central Mediterranean climates responded primarily to a 41-kyr glacio-eustatic (obliquity) forcing (Fig. 1). Closed mesothermal Mediterranean-type forests were dominant during full MIS 17, pointing to present-day climatic conditions in the area, with rainy winters and dry summers18,41,42. Changes in the Mediterranean forest pollen record at the Blatta section follow very closely those found at the Iberian Margin IODP Site U1385, thus supporting a scenario of increased westerlies penetration into southern Europe around 695 ka18, with winter storm tracks reaching as far as central Mediterranean.

Non-arboreal plants (NAP) communities, evocative of dry to sub-desertic climates, characterize late MIS 18 and early MIS 16 glacials (Fig. 1). Individual peaks of water-demanding conifers (Mountain Forest in Supplementary material S6, Fig. S6), suggestive of increased annual precipitation with abundant summer rainfalls, are documented within MIS 17. The main pluvial event occurs at the termination of 7-IMS15 stadial, in late full MIS 17, where water-demanding conifers attain to ca. 70% of the total pollen assemblage (Fig. 2). This episode coincides with a major decrease in G. ruber δ18O, increased G. ruber–U. peregrina δ18O offset (Δδ18OUper-Grub) and abundance of G. sacculifer (Fig. 2), pointing to the persistence of a surface freshwater lens throughout the summer with increased density-driven stratification of the local water column20,28. Notably (Fig. 2), this event occurs as the Mediterranean forest declines and reconstructed summer precipitation increases in the Iberian Margin record18.

Figure 2
figure 2

Geochemical and micropaleontological data collected at the Blatta section and comparison with selected records. From the left: Ba/Al ratio at ODP Site 967, eastern Mediterranean Sea32. Downcore variations of G. ruber δ18O values (this study). Downcore variations of U. peregrina δ13C values (this study). Downcore variations of Δδ18O values between U. peregrina and G. ruber (this study). Downcore percentage values of G. sacculifer (this study). Insolation at 39° N in August variations40. Downcore percentage values of mountain forest species (this study). Mediterranean forest from IODP Site U138518. Except for the insolation record and for the Mediterranean-type vegetation, thick lines are 3-pt running averages. Horizontal dotted black lines indicate MIS boundaries from ref15. Vertical grey boxes indicate stadial phases, progressively labeled, following the procedure by ref19. The timing extent of sapropel S16 follows ref32.

Full size image

According to our age model, regional precipitation peaked ca. 3 kyr after the precession-driven monsoonal maximum and sapropel S16 deposition, as marked in our record by the negative δ13C excursion of U. peregrina (Fig. 2). A 3-kyr lag is fully consistent with the expected delay in insolation maxima between the central Mediterranean and the tropics (see August insolation at 39° N in Fig. 2). We conclude that central Mediterranean precipitation during MIS 17 were primarily controlled by obliquity but also responded to local insolation changes, which is not perfectly in phase with the precession-related African monsoon signal. Regional precipitation may have been supplied by atmospheric moisture reservoirs that changed seasonally, according to the prevailing orbital forcing at the time. Obliquity-forced winter rainfalls would exploit both Mediterranean and Atlantic sources, while precession-related precipitation would be fueled by Mediterranean yields in the late summer-early fall and by Atlantic advection in the winter43.

Processes and mechanisms for increased Mediterranean summer precipitation during MIS 17

Two mechanisms may sustain a scenario of augmented summer precipitation in the central Mediterranean after sapropel S16 deposition. The first relies on the variability in local summer insolation rates, which attain a maximum during August in the study area (Fig. 2). Increased insolation rates favor the development of Convective Systems with Local Effects (CSLE) by increasing evaporation, vertical rise and condensation of humid air masses44. Nowadays, CSLE may cause intense precipitation in the very same area as the Crotone Basin, where steep coastal orographic barriers promote the rapid adiabatic rise of warm, humid air masses. CLSE magnitude is further increased by the periodic influx of cool air advected from the north. By comparison to recent interglacials, similar conditions may have occurred during stadials 7-IMS15 to 7-IMS13 (late full MIS 17; Fig. 2), when cooling over the Atlantic increased the frequency and magnitude of summer continental and polar outbreaks to the Mediterranean, as evidenced by drops in summer SSTs and erosion of the thermocline17,27,45,46,47,48.

An alternative explanation implicates the development of extra-tropical (Mediterranean) cyclones during stadial 7-IMS15. As for the previous, their magnitude is expected to increase during stadials in response to AMOC weakening. Indeed, modern summer extra-tropical cyclones are fueled by unstable air masses with high potential vorticity conveyed to the Mediterranean by the southward migration of the Polar Jet49,50. As documented for the Younger Dryas40,50,,51,52, a similar scenario is associated with conditions of AMOC slowdown and southward expansion of sea ice, since a stronger E-W temperature gradient in the mid-latitude North Atlantic would favor the incursion of westerlies and moisture over the Mediterranean52.

Lake level fluctuations, vegetation biomes and speleothem isotopes in southern Europe borderlands point to an increased summer precipitation regime since the 8.2 ka event9,10,11, when the interglacial stability declined38,53 and continental Europe was still experiencing dryness11,12. CSLE described for the MIS 17 may also provide a suitable explanation for contrasting precipitation levels in Europe since the middle Holocene. By analogy with MIS 17 hydroclimate, contrasting precipitation levels may be developed between continental and southern Europe once Holocene climatic instability was established, and well later than the beginning of sapropel S1 deposition in the eastern Mediterranean (10.8 ka2,36). Local convective processes would explain the limited geographical (coastal) extent of the observed summer precipitation, in contrast with inner continental areas where the mechanism is ineffective9,10,11,12,35.

MIS 17 Precipitation and increased stadial moisture

CSLE and extra-tropical cyclones are not mutually exclusive, as they both comply with the deployment and persistence of North Atlantic storm tracks and Mediterranean cyclogenesis3,6,7,8. However, the CSLE model better suits the overall scenario found at the Blatta section. In our record, intervals of “light” δ18O spikes for G. inflata (Supplementary Material S7, Fig. S7), higher abundances of G. inflata and G. truncatulinoides in early and late MIS 17 (Supplementary Material S3, Fig. S3) point to the development of a local cool mixed-layer20 in response to seasonal atmospheric low-pressure conditions. These conditions are especially well represented in stadials 7-IMS18, 7-IMS 15 and 7-IMS11, where the concomitant decline of NAP and expansion of water-demanding forest elements (Fig. S6) is evocative of increased precipitation (Fig. 1).

Increased regional precipitation during stadials seem inconsistent with other circum-Mediterranean sites, where cold spells are associated with arid climates23. However, we stress that the regression to semi-steppe documented during stadial periods at Lago Grande di Monticchio (southern Italy) or in the Anatolian Peninsula23,54, as well as the context of weaker precipitation suggested by models55,56, took place during a full glacial, when local convective systems were hindered by minimal local insolation rates. Our results suggest that interglacial stadials may be associated with wetter climates, differently from the classical model of dry stadials and wet interstadials that are believed to characterize glacial periods.

Conclusions

The multi-proxy terrestrial and marine records obtained from the Blatta section point to the occurrence of suborbital climatic oscillations during MIS 17 that are consistent with the AMOC variability reconstructed at the Iberian Margin. Timing, rates and regimes of regional precipitation were out of phase with the precession-related African monsoon activity, which triggered the sedimentation of sapropel S16 in the deep eastern Mediterranean. Local summer precipitation rates increased dramatically by the end of sapropel S16 deposition, when the Mediterranean thermohaline circulation started recovering (Fig. 2), in agreement with the climatic and oceanographic evolution reconstructed across sapropel S111,12. Increased summer precipitation in the central Mediterranean can be explained by the development of CSLEs and extra-tropical cyclones under the joint effects of increased local evaporation rates, stronger convection44 and southward shifts of the Polar Jet during stadials, which would favor the advection of cool and humid Atlantic air52.

Our study illustrates the importance of global climatic drivers like insolation and AMOC in impacting regional-scale hydrological systems, with amplifying effects that may have impacted the Mediterranean area during MIS 17, and possibly the early-middle Holocene. Unravelling the mechanisms and processes in place, like those here referred to as CSLEs, is a basic requisite for a better understanding and effective forecasting of extreme summer precipitation events, such as those increasingly afflicting densely populated coastal areas of the Mediterranean under the ongoing climate change.

Methods

Stable isotopes

Oxygen and carbon isotope analyses were run on 5/7 individuals of the planktonic foraminifera Globigerinoides ruber (220 samples), Globorotalia inflata (100 samples) and of the benthic foraminifer U. peregrina (250 samples). Specimens were handpicked from samples collected each 20 cm. Samples were measured by automated continuous flow carbonate preparation GasBenchII device57 and a ThermoElectron Delta Plus XP mass spectrometer at the IAMC-CNR (Naples) isotope geochemistry laboratory. Acidification of samples was performed at 50 °C. An internal standard (Carrara Marble with δ18O = − 2.43 vs. VPDB and δ13C = 2.43 vs. VPDB) was run every 6 samples and the NBS19 international standard was measured every 30 samples. Standard deviations of carbon and oxygen isotope measures were estimated at 0.1 and 0.08‰, respectively, on the basis of ~ 70 repeated samples. All isotope data are reported in per mil (‰) relative to the VPDB standard.

Pollen

190 samples have been subjected to palynological investigations Rock fragments with a dry weight of 10 g were treated according to standard procedures, namely: removal of the calcite content with concentrated HCl; elimination of silicates by means of concentrated HF; digestion of the organic matter (whenever necessary) with hot diluted KOH; separation of the pollen content by means of gravitative separation with ZnCl2 at d = 2.004; ultrasonic disintegration of the residual inorganic fraction. Residues have been stored in glycerin and mounted at the spot on disposable slides. Pollen were analyzed under a light microscope with 100 × and 430 × magnifications. For each sample, a minimum of 195 and a maximum of 991 (average 359) grains have been counted based on the specific pollen abundance.

Planktonic foraminifera

A total of 188 samples were prepared for planktonic foraminifera analysis. Samples were washed using a 63 µm mesh sieve and were oven-dried at 40 °C. Quantitative analysis was carried out counting all the specimens occurring on the total residue from the fraction > 125 µm.

Planktonic foraminifera species were identified following taxonomic concepts by58,59. The Globigerinoides ruber white group includes Globigerinoides elongatus and Globigerinoides conglobatus. The Globigerinoides sacculifer group includes Globigerinoides trilobus and Globigerinoides quadrilobatus. The Globigerina bulloides group includes Globigerina falconensis.

The planktonic/benthic (P/B) ratio, as a proxy for quantitative paleo-depth estimates, follows the calibration by ref60:

$${\text{D}} = {\text{e}}^{{\left( {{3}.{58718 } + \, \left( {0.{\text{3534 x }}\% {\text{P}}} \right)} \right)}}$$

where D is the seafloor depth in meters below sea level and %P is the percentage value of planktonic foraminifera specimens.

Benthic foraminifera

A total of 188 samples were prepared for benthic foraminifera analysis. Samples were washed using a 63 µm mesh sieve and were oven-dried at 40 °C. Quantitative analysis was carried out counting all the specimens occurring on the total residue from the fraction > 125 µm.

Benthic foraminifera species were identified following taxonomic concepts by ref61. Bolivina dilatata includes Bolivina spathulata. Uvigerina mediterranea includes U. peregrina. Specimens of the genus Quinqueloculina, Triloculina, Cruciloculina, Biloculinella, Miliolinella, Pyrgo and Sigmoilina were grouped as Miliolids. The deep infaunal taxa Chilostomella spp. and Globobulimina affinis were grouped as oxygen deficiency stress (ODS) species62.

Nannofossils

A total of 71 samples were investigated for calcareous nannofossils. Samples were prepared from unprocessed material as smear slides and examined using a light microscope at 1250 × magnification. Quantitative analysis was carried out counting the relative abundance of selected taxa with respect to a pristine population of at least 500 specimens30 with sizes ≥ 3.5 µm. Taxonomic concepts follow those of refs63 except for species of genus Gephyrocapsa, which are those of ref64. The adopted zonal scheme is from ref30.

Ar/Ar dating

Sanidine crystals were extracted by simple disaggregation of the ash layer, with further handpicking under a binocular microscope. They were cleaned by ultrasonic etching in a dilute (2–5%) hydrofluoric solution for 3–5 min, followed by ultrasonic rinse in acetone, ethanol, and deionized water. Single crystals were individually loaded in three 4 mm I.D. holes machined into a 11 mm O.D., 3 mm thick, Al-irradiation disks. Samples were co-irradiated with the irradiation monitors ACR-2 and TCR loaded in smaller adjacent pits bracketing the three sample locations. Corrections for isotopic interferences from K, Ca, and Cl were applied using production ratios listed in Scaillet et al. (2013)65. After baking overnight at 180 °C, single crystals were individually analyzed with a continuous 20 W Synrad ® CO2 laser source coupled to a noble gas mass-spectrometer operated in pulse-counting mode65. Released gases were purified prior to gas admission into the mass-spectrometer by exposure for 10 min on two air-cooled GP50 S.A.E.S.® getter cartridges featuring a Zr–Al St101® alloy held at 250 °C. Age were calculated using in-house software66 based on conventional isotope abundances and monitor and decay constants65. Monitoring of the instrumental mass-fractionation is achieved by daily calibration of the atmospheric 40Ar/36Ar isotope ratio on air shots interspersed with the unknowns at different peak intensities to correct for nonlinearity and counting dead-time (for every isotope) using in-house software. Age errors are plotted and tabulated at 2σ and include corrections for (1) counting dead-time for every isotope, (2) system blanks, (3) mass-discrimination, (4) post-irradiation decay of 39Ar, 37Ar, and 36Cl, (5) isotope interference corrections from K, Ca and Cl, (6) atmospheric contamination, (7) neutron-flux gradients, and (8) monitor’s age error.