Imprints of Holocene aridity variability in the Aegean Sea and interconnections with north-latitude areas

Factor analysis

To investigate the distribution of the elemental concentrations along the core, we ran a factor analysis (FA) on the XRF counts and compared these results with specific elementary ratios and the XRD data (Figures 2, Supplemental Figure S2). The first four factors are considered to be significant and explain ~83% of the dataset variability (Figures 2, S1, Supplemental Table S1). The first factor (FS1; explained the 54.1% of the total variance, Supplemental Table S1) includes Al, Si, K, Ti, Cr, Ba, Fe, Co, Cu, Zn, Ga, Rb, Y, Zr, Rb, and Pb which are commonly considered as terrigenous indicators (Zabel et al., 2003) with positive loadings, and Ca, Sr and As with negative loadings. Ca and Sr are chemically similar and are major constituters in carbonate sedimentation (Cuellar-Martinez et al., 2017, and references therein), while As is probably associated with volcanic formations surrounding the study area (Nordstrom, 2002) or pyrite authigenesis (Thomson et al., 2006). Therefore, FS1 depicts the intervals of the core sediments where the presence/absence of carbonates is significant. The downcore variations of the FS1 scores show that the high concentrations of carbonates (negative scores) occur within the LU-IIIA and IIIB and LU-I and the lowest values (positive scores) in LU-II and IIIC (Figure 2). In an attempt to investigate the nature of the carbonate sedimentation, we used the following elemental ratios: the Ca/K ratio, as an index for the biogenic carbonate contribution, since K is mainly related to illite; Sr/K and S/Ca ratios, as indexes for the authigenic carbonate sedimentation, since aragonite is rich in strontium while gypsum rich in S; and the Ca/(S+Sr) ratio to depict the detrital carbonate fraction, since aragonite and gypsum are removed (Neugebauer et al., 2015, 2016). Within the LU-III, the values of these ratios fluctuate in accordance with the subunits and show high values of S/Ca in the subunit LU-IIIC and frequent alternations of the ratios S/Ca and Sr/K in the LU-IIIA (Figure 2). These observations coincide with the XRD measurements and suggest the presence of gypsum and alternations of gypsum and aragonite within LU-IIIC and LU-IIIA, respectively (Figure 2). On the other hand, the detrital carbonate fraction (high Ca/(S+Sr)) characterizes the subunit LU-IIIB. The biogenic carbonate fraction also contributes to the LU-IIIA and LU-IIIB, and partially coincides with brief intervals where microfauna shells were observed (suppl. Table S1: Noti et al., 2022). However, the most significant contribution of biogenic carbonate sedimentation is found within LU-I where the record presents constantly high values of Ca/K and Sr/K (Figure 2). The XRD measurements indicate that these high Sr/Ca values are linked to the presence of aragonite. The high biogenic carbonate fraction within this interval can be explained by the high abundances of microfauna assemblages within LU-I of which the dominance of benthic species related to dysoxia or seasonal anoxia may explain the aragonite precipitation (Noti et al., 2022).

The second factor (FS2; explained the 12.6% of the total variance, Supplemental Table ST1) includes as significant variables the XRF counts of Cl, Br, and I. All these elements are considered as salinity indicators. The downcore variations of the FS2 are high within LU-II and LU-I (Figure 2), suggesting a marine influence at the core site at that time. This observation correlates well with the microfaunal assemblages which suggest the presence of a lagoon at the core site within LU-II which later on within LU-I, developed into a marine bay (Noti et al., 2022). The third factor (FS3; explained the 6.5% of the total variance, Table ST1) included the XRF counts of Mg, Mo, and S. All these elements are associated with sediments rich in organic material (Algeo and Lyons, 2006; Nieto-Moreno et al., 2011). The high values of FS3 correlate well with the enhanced presence of pyrite (Huerta-Diaz and Morse, 1992; Łukawska-Matuszewska et al., 2019) and the increased Corg content (Noti et al., 2022) within LUIII A & B (Figure 2), suggesting enhanced accumulation of organic matter. The fourth factor (FS4; explained the 4.5% of the total variance, Supplemental Table ST1) included only the XRF counts of Mn. Although, Mn is considered as a redox-sensitive element and usually is involved to depict diagenetic processes in the marine sediments (Raiswell and Canfield, 2012), it is also related to detrital sediments since Mn oxyhydroxide precipitates are dispersed within sediments in the form of coatings of detrital particles (i.e. Bayon et al., 2004). High values of FS4 observed mostly within LU-IIIB (Figure 2) and correlate well with strong presence of quartz, suggesting a detrital source for this element.

Elements and elemental ratios

The main patterns in the elemental composition of the Astypalea core (Figure 3) indicate that the sediments are dominated by an opposing decreasing trend of the terrigenous sourced elements, such as Aluminum (Al), Silicium (Si), Titanium (Ti) and Zircon (Zr), and increasing trend of carbonate elements, such as Strontium (Sr) and Calcium (Ca). A sharp transition between the terrestrial and carbonate elemental concentrations occurs around 6.0 ka, which marks the boundary between LU-II and LU-III of which the latter is characterized by a laminated interval with up to 2% Corg concentrations.

OPEN IN VIEWER

Figure 3.

Elemental concentrations in counts per second derived by X-Ray Fluorescence scanning of core ASTC1. The age model of ASTC1 is based on radiometric 14C dates (see Noti et al., 2022). Zones I to IIIC indicate the main lithological units and subunits of the ASTC1 sediment core described by Noti et al. (2022). Horizontal dashed lines indicate the boundaries between Lithological Units (LU) I-IIIC (Noti et al., 2022).

Sr and Ca show a high similarity along the record (Figure 3) except for two brief periods between 8.0 and 7.4 ka, where probably an additional detrital source of Calcium may have enriched the Ca values with respect to Sr, which is usually used as marker for biogenic origin as it is incorporated into aragonite (Foubert and Henriet, 2009; Richter et al., 2006). The close resemblance between the Si and Al records indicates furthermore that there is no clear evidence for large changes in biogenic silica production.

Ti and Zr are known to be enriched in aeolian dust from the Sahara Desert (Guieu and Thomas, 1996; Wehausen and Brumsack, 2000) and both have been used as a dust proxy in the Mediterranean region when they are divided by Al (Jimenez-Espejo et al., 2008; Lourens et al., 2001; Rodrigo-Gámiz et al., 2011). Accordingly, we chose to compare the trends of two elemental ratios, the Ti/Al and Zr/Si to investigate possible changes in dust input at the site (Figure 3). Clearly, both ratios show a high similarity and a general increasing trend toward the present with maximum values between 3.5 and 2.5 ka (Figure 3).

Spectral analysis

We applied spectral analysis on the detrended Ti/Al and Zr/Si time series to evaluate if they are determined by cyclic variations (Figure 4). This analysis revealed the prevalence of three major frequencies within our record, of which a 2493-2685-year cycle dominates the signal (highest power; 99% c.l.), followed by a 1091–1396-year cycle (90% c.l.), and 521–553-year cycle (95–99% c.l.; Figure 4).

OPEN IN VIEWER

Figure 4.

Spectral analysis results of the detrended time series of: (left top and bottom) Ti/Al and Zr/Si detrended record of ASTC1 sediment core, and (right top and bottom) the 20-pt moving averaged GISP2-K record (Mayewski et al., 2004) and stacked North Atlantic Ice indices (Bond et al., 1997) between 0 and 9 ka. The brown shaded areas indicating the spectral results obtained from the Multi Taper Method (and Lomb Scargle for the North Atlantic Ice Indices) in ACycle (Li et al., 2019) and the solid black lines are those obtained from the standard Blackman-Tukey power estimates in AnalySeries (Paillard et al., 1996). The red solid curves indicate the MTM (LS) (AR1) confidence levels at 99%, the red dashed lines at 95% and red dotted lines at 90%.

To further investigate the cyclic patterns in the Ti/Al and Zr/Si time series, we extracted their main spectral components using a Gaussian filter centered at their peak frequency (Figure 5). In the first panel of each figure, we have plotted the fourth order polynomial fit used for detrending. In the Second–Forth panels, we have subsequently plotted the extracted ~2500-year, ~1200-year and ~525-year components (red dashed lines). In both Zr/Si and Ti/Al time series, all filtered components reveal an increasing amplitude toward the present (red lines of the first panel), indicating that their variability increased during the Holocene. The lowest variability is found in LU-III between 6 and 8 ka (Figure 5). The two proxy records show an almost in-phase relationship for all three filtered signals.

OPEN IN VIEWER

Figure 5.

Comparison between the major cycles (dashed red lines) identified in the Zr/Si and Ti/Al time series as overlays to their original detrended series (black solid lines). The original Zr/Si (gray shaded area) and Ti/Al (black solid line) records are shown in the left panel (first panel) including their fourth order polynomial fit as indicated by red dotted and solid lines, respectively. The second, third, and fourth panel refer to the 2500-yr, 1200-yr, and 525-yr filtered components of the two proxies respectively.

Holocene increases in aridity in the Eastern Mediterranean

The geochemical data derived from the FA and the ratios of the elements together with the mineral data from the XRD measurements present similar trends and indicate fluctuations in the sedimentation at the core site since 8.7 ka. The alteration in the site’s evolution was both accompanied by sea-level rise and climate change (Noti et al., 2022). During the deposition of LU-IIIC (8.7–8 ka), under lower sea level (−27 to −13 m below present), the core site was isolated from the open sea and authigenic carbonate sedimentation was dominant. Based on all proxy evidence (section 3.2; Figure 3), the accumulation of detrital material increased between 7.3–8 ka (LU-IIIB). This interval is associated with increased humidity in Aegean region as pointed out by studies in marine and terrestrial records (i.e. Geraga et al., 2010; Schemmel et al., 2017) and corresponds to the later phases of the African Humid period. This shift to wetter conditions is attributed to increased boreal insolation and the northward shift of the Intertropical Convergence Zone (ITCZ, i.e. deMenocal et al., 2000). The high aragonite and gypsum percentages found in LU-IIIA between 7.3 and 6 ka are most likely associated with increases in salinity under enhanced evaporation in isolated and restricted water bodies (Koutsodendris et al., 2015), when the connection between the Vathy basin and the open sea was gradually being developed (Noti et al., 2022). On the other hand, the occurrence of pyrite together with the increased values of organic material and biogenic fraction indicators, may also provide evidence for a biological origin of the carbonate precipitates (Sondi and Juračić, 2010) in this subunit. After 6 ka, the salinity indicators and the biogenic fraction present increasing trends and reflect the permanent connection of the study area to the open sea, forming at the initial phases a lagoon (LU-II; 6–4.1 ka) and then a bay (LU-I, 4.1–0 ka, Noti et al., 2022).

Beyond the impact of the sea level rise, the examination of the retrieved sediment record suggests that climatic variability also resulted in changes on the geochemistry record of the study area. In order to further investigate the climatic variability, we focused on the records of the Ti/Al and Zr/Si ratios whose fluctuations as discussed previously (section 3.2) represent long- and short-term changes in dust supply and thus could be associated with aridity intensification. During LU-I & LU-II, a general increasing trend of both ratios, together with a general enrichment in the δ¹⁸O values (Noti et al., 2022) suggest an increase in dust supply and in aridity at the core site, over the last 6 ka. Around that time, drier conditions and/or increases in dust flux started to be recorded in the Balkans (Finné et al., 2019; Longman et al., 2017 and references therein), in Anatolia region (on speleothems, Cheng et al., 2015), and marine sediment data, that is, Hennekam et al., 2014; Schmiedl et al., 2010) and in North Africa region (Palchan and Torfstein, 2019). This shift to drier conditions in the eastern Mediterranean has been attributed to the southward migration of the ITCZ during the middle and Late-Holocene (i.e. Fleitmann et al., 2003). The retreat of monsoonal rains by the end of the African Humid Period resulted into the onset of Saharan desertification (deMenocal et al., 2000). From about that time (Ehrmann et al., 2017) and up to present, the Sahara constitutes a major source of dust influx to the eastern Mediterranean region (Avila et al., 1998; Beuscher et al., 2020; Ginoux et al., 2012).

In the Astypalea record, though, short term fluctuations in both Ti/Al and Zr/Si ratios, suggest brief oscillations of increased terrigenous supply associated most probably with increases in aridity. High values of both Ti/Al and Zr/Si ratios cluster at around 8.4, 8.1, 7.7, 4.9, 4.6–4.2, 3.8, 3.3–2.4, 1.8–1.5, and 0.6–0.3 ka (Figure 6). Previous studies of high resolution datasets from lake sediments (Peloponnese: Emmanouilidis et al., 2022; Katrantsiotis et al., 2019; Unkel et al., 2014; Turkey: Eastwood et al., 2007), speleothems (Peloponnese: Finné et al., 2014, 2019; Leb-anon: H. Cheng et al., 2015) and marine sediment cores (North Aegean Sea: Gogou et al., 2016; Hamann et al., 2008; Western Greece: Koutsodendris et al., 2017; Israel: Hennekam et al., 2014) have shown numerous arid events, in the EM within the Holocene, and several of these events appear at intervals similar to our record, although the exact time of their occurrence is variable among the records probably due to small dating uncertainties. Among them, the most widely documented arid events are those centered at 8.2, 6–5, 4.2, 3.5–2.5, 1.2–1 ka, and the 0.65–0.15 ka and some of them are globally referred as the Holocene “Rapid Climate Change” events (RCCs, Mayewski et al., 2004, and references therein). However, the main RCCs expressed in the EM, and specifically in the Aegean Sea records are those between 8.6–8.0, 6–5.2, and 3.1–2.9 ka, which are relevant with the increases suggested by our record. Other severe millennial events in the EM associated with increased aridity are centered at 5.3–4.2 and 2.8–1.4 ka as shown by the high resolution δ¹⁸O record on cave speleothems from the south-eastern Mediterranean region (Jeita cave, Cheng et al., 2015) and the events around 7.7 and 1.8 ka which coincide with aridity intensification mostly evidenced in cores from the north-eastern Mediterranean region (Aegean Sea, Carpathians and Turkey; Finné et al., 2019; Longman et al., 2017; Triantaphyllou et al., 2016)

OPEN IN VIEWER

Figure 6.

From left to right, the GRIP δ18Ο from the Greenland ice core (blue line), the Sanbao cave (central China) δ18Ο record (red line), the Jeita cave δ18Ο record (black line), the Ti/Al (orange line) downcore variation of the PS009PC Levantine marine sediment core, the Chew Bahir PC1 in the eastern Africa showing the dust variability (green line), and the Ti/Al (red line) and Zr/Si (black line) downcore variation from the ASTC1 marine sediment core in Astypalea, South Aegean Sea. The gray shaded areas represent the arid intervals found in ASTC1 record.

These arid events appear to have impact on civilizations developed in Balkans (i.e. Berger et al., 2016; Finné et al., 2017) as well as in Anatolia and Middle East (i.e. Berger et al., 2016; Clarke et al., 2016). In our record, the time occurrence of the changes between arid and wet climatic conditions appear to almost coincide with main archaeological periods in Greece: the Late Neolithic period (~9–5 ka) coincides with a general humid period (~until 7 ka) followed by a transitional period (wet to dry, 7–5.5 ka BP (Desprat et al., 2013; Finné et al., 2011); the Early (5–3.9 ka) and the Late Bronze (3.5–3 ka) period which were more arid than the Middle Bronze period (3.9–3.5 ka, Finné et al., 2011). Enhanced aridity also characterized the Greek Dark Ages (3-2.6 ka, (Langgut et al., 2013; Ina Neugebauer et al., 2015; Roberts et al., 2011) while wetter conditions prevailed during the classical period (2.4–2.3 ka, Dean et al., 2015). The Roman period (2–1.6 ka) coincides with relative high levels of aridity which continued up to the early phases of the Byzantine period with several reported extremely cold seasons (1.6–0.7 ka) (Telelis, 2008; Xoplaki et al., 2016). In the North Aegean though, the latter period is characterized by fluctuations in the SSTs and relative high river discharge (Gogou et al., 2016).

The most recent increase in aridity in our record occurs between 0.6 and 0.3 ka and probably corresponds to Little Ice Age (LIA; Figure 6). Oscillations of humidity/aridity indicators corresponding to Medieval and LIA periods have been observed in the paleo-climatic proxies from Greece (Finné et al., 2017; Gogou et al., 2016; Hamann et al., 2008) and Levant region (Cheng et al., 2015). These studies suggest that the EM experienced reduced precipitation and drier conditions during the LIA, while the Medieval Climate Anomaly (MCA) is associated with relatively wetter conditions and increased humidity in certain parts of the EM. Our record also indicates decreased aeolian activity which may reflect wet climate conditions in the south Aegean Sea within the Medieval period (1.1–0.7 ka) and increased aridity within the LIA. In addition, our records suggest climatic fluctuation within the LIA with the aridity being interrupted for a short period around 0.5 ka. This interruption has been linked to an increase in humidity within the LIA, which was also evidenced in the North Aegean Sea at around 0.5 ka, pointing out the high hydroclimate variability within this period (Gogou et al., 2016). Within the LIA, glacier development/expansion occurred also in many Mediterranean mountain locations, while semi-permanent snowfields became established in the mountains of Greece, which are too dry for glacier development (Hughes, 2014). Drop in mean summer temperatures by 2°C–3°C at around 0.35 ka in the Pirin Mountains, SW Bulgaria (Grunewald and Scheithauer, 2010), support further the dry scenario within the LIA across these regions. Increased aridity is also inferred from central Anatolian Lake Nar Gooülu data (Dean et al., 2013), while frequent widespread freezing events between 0.35 and 0.02 ka also affected the Black Sea, Bosporus, Golden Horn, and Istanbul region (Yavuz et al., 2007). The main factors responsible for the winter cold and snow-falls within this region involved northerly airflow, with high pressure over northern Europe and lower pressure over the central or eastern Mediterranean (Xoplaki et al., 2001).

Global arid traces within the Holocene

To better understand the phase relations between the northern-, mid- and south- northern latitude sub-systems, we compared our record to those from five other sites with high temporal resolution and precise age control (Figure 6). These records are from Greenland (GRIP δ18Ο ice core record, Johnsen et al., 2001), from eastern China (speleothem δ18Ο record of the Sanbao cave, Wang et al., 2008), from central Africa (Chew Bahir PC1, Trauth et al., 2018), Ti/Al record from southeast Levantine Basin in the eastern Mediterranean (marine sediment core PS009PC, Hennekam et al., 2014), and from Jeita cave in Lebanon (δ¹⁸Ο on cave speleothems, Cheng et al., 2015), Figure 6). The overall trend shown by the geochemical ratios Ti/Al and Zr/Si points to dust increase over the last 6–5 ka which agrees well with other trends seen globally, like in record from Africa (Chew Bahir), Greenland ice core record (GRIP δ18Ο) and the monsoon-related speleothem δ18Ο record of the Sanbao cave in central China (Figure 6). Prior to the aridification trend, all datasets suggest a wetter trend for the Early to Middle Holocene compared to the Middle-Late-Holocene, suggesting a coherent climatic response across these regions (Figure 6), although decoupling between GRIP and Sanbao d18O record is due to the stronger response of the latter to insolation forcing than the first (Dong et al., 2010). The increases in aridity though noticed between 8.7 and 8 ka in our record, couple with the cooler trend shown by the GRIP record and the Sanbao cave speleothems. This increase of aridity is possibly linked to the 8.2 cooling event, an event which had a global impact and affected various parts of the globe, such as North Atlantic, Europe, Asia, Africa, and South America (Blockley et al., 2012; Bond et al., 1997; Guo et al., 2000; Haug et al., 2001; Magny et al., 2003; Rein et al., 2005; Shanahan et al., 2015; Yao et al., 2017). There is ongoing debate among scientists regarding the exact timing and causes of the 8.2 ka cooling event, as well as its duration. Some suggest that the event began abruptly around 8.2 ka (Alley et al., 1997; Grootes et al., 1993; Haug et al., 2001; Kennett et al., 2000; Rasmussen et al., 2006), while others propose that it may have started earlier with the oldest scenario setting the initiation at 8.6 ka BP (Rohling and Pälike, 2005; Wang et al., 2010). There is also debate about the primary drivers of the cooling, with some studies suggesting changes in ocean circulation and freshwater input and others highlighting changes in solar radiation or volcanic activity. The climatic response of this cooling event in other records from the EM is marked by drier conditions and reduced precipitation (Develle et al., 2011; Staubwasser et al., 2003), increased aeolian activity and dust deposition (Develle et al., 2011; Torfstein et al., 2013). Thus, this shift appears to have affected synchronously all cross-correlated dataset used in this study, showing the teleconnection of the northern and southern latitudinal climatic systems. Next to this arid interval, a prominent increase of aridity seen in the ASTC1 around 7.7 ka seems to agree with a relative signal from south-eastern EM (Jeita Cave; H. Cheng et al., 2015). As previously mentioned above a similar trend has been recorded in records northern of the examined area (up to Carpathian region; Longman et al., 2017). This may reflect a northern origin of this event related to the intensification of the Siberian High.

The exact timing, however, of the various dry periods addressed by the global records may vary due to differences among the specific datasets, the location, as well as the methods used to reconstruct past climate and their resolution (Bond et al., 2001; Mayewski et al., 2004; Rasmussen et al., 2014). For example, some suggest that the Holocene Climatic Optimum came to end approximately at 6–5 ka when the climate conditions turned into more arid (Cullen et al., 2000; Kaniewski et al., 2010; Roberts et al., 2011), while others suggest that the onset of aridity occurred later, around 4–3 ka (Koutsodendris et al., 2013; Roberts, 2014) (Figure 4). The compared records here show that the increase in aridity started at around 6 ka or earlier and continued until full arid conditions were reached at around 5 ka (Figure 6). This shift from wetter conditions within the Holocene toward aridity marks the most prominent ecological changes during the Holocene, with the termination of the African Humid Period (AHP). Some records reveal a more gradual transition, like in the ASTC1, Chew Bahir, Sanbao, and the GRIP record, and others support a more abrupt termination of the AHP, like in the PS009PC record and other marine archives (Hennekam et al., 2014; ODP 658C sediment core: deMenocal et al., 2000). The exact timing of the AHP termination is debated, basically due to the different location and their proximity to the governing climatic system. deMenocal et al. (2000) differentiate between two feedback mechanisms that could have amplified such a transition, a coupled vegetation-albedo feedback and ocean surface temperature-moisture feedback, as well as declining summer insolation (Renssen et al., 2006), and the southward migration of the ITCZ (Sachs et al., 2018; Zhang et al., 2021).

The drying trend continues until around 2 ka, with the most prominent increase in aeolian activity seen in ASTC1 record centered around 2.8 ka. The period around 3 ka has been suggested to be the driest time of the Holocene in the EM (Finné et al., 2019) but, there is also multiple evidence, of draught and increased aridity during this period in several regions around the globe, including EM (Soreq Cave in Israel, Lake Van in Turkey, Bar-Matthews et al., 2003; Lamb et al., 2007), Central Asia (Shihua Cave in northeast China, Ku and Li, 1998), Central America (Lake Chichancanab in the Yucatan Peninsula, Mexico, Hodell et al., 2005), and South America (Cueva Larga in Peru, Rein et al., 2005), and in other sediment cores from southeastern Greece (Gogou et al., 2016; Zanchetta et al., 2013). Changes in the North Atlantic due to solar activity and/or freshwater input from melting ice sheets around 4 ka is suggested to be responsible for this long-term aridity and the subsequent weakening of the Indian monsoon system (Kathayat et al., 2018) following the 4.2 ka cold event. Some studies suggest that the decrease in precipitation during this period may be related to changes in large-scale atmospheric circulation patterns, such as the Intertropical Convergence Zone (ITCZ) and the North Atlantic Oscillation (NAO), as well as solar variability (Izdebski et al., 2016; Magny, 2013; Mayewski et al., 2004) and increased volcanic activity (Kobashi et al., 2017).

After that, the most recent aridity increases at 0.7 and 0.3 ka show a coupling with the global records. This most recent RCC (<0.6 ka) features bipolar cooling but a more variable response in humidity at low latitudes (Mayewski et al., 2004), with this interval appearing to be more complex than the classic “cool poles, dry tropics” pattern that typified the Pleistocene and most of earlier Holocene RCCs. However, here it seems that the climate response to this RCC was almost synchronous across the compared regions with cooler and drier conditions implied by the proxies.

Holocene climatic cyclicity

The Holocene has seen a complex interplay of natural factors that have driven cyclical patterns of climate change over time. Several cycles of climate change have been identified, including periods of warming and cooling in ice and sediment records. Here the application of the spectral analysis suggests that the Ti/Al and Zr/Si ratios of the ASTC1 sediment core are non-stationary data, but present cyclicity referring to recurrence of dry/cold climatic conditions with main periodicities at ~2500, 1200 and 525 years within the Holocene (Figure 3). The frequency of the variability increases after 6 ka which is coincident with previous studies which show increase in the frequency of aridity events after the end of the African Humidity Period (Balkans: Finné et al., 2019; Anatolia and Middle east: Clarke et al., 2016). An oscillation with a period of about 2100–2500 years in ¹⁴C concentration, known in the literature as the Hallstatt cycle (Vasiliev and Dergachev, 2002), has been found in various paleoclimatic records spanning the Holocene (Scafetta et al., 2016, and references therein). This 2100–2500-year oscillation both in the cosmogenic radioisotope and in the climate, records have been suggested to have an origin of three kinds: astronomical, solar, and Earth’s endogenous. The comparison with the GISP2 record and the stacked drift ice indices shows an in-phase relationship within the 9 ka with a small offset though (Figure 7). The variability however within the ASTC1 proxies is small prior to 5 ka, and thus better describes this section of our record showing a positive relation of the cooler conditions in the northern hemisphere with the intensification of the aeolian activity and/ or drier conditions in the South Aegean Sea with a 2500-yr periodicity.

OPEN IN VIEWER

Figure 7.

Gaussian Filter of ~2500 years applied on the detrended data from Drift ice indices in North Atlantic, GISP2 record and the ASTC1 (Ti/Al and Zr/Si) record. The green and blue bars refer to the glacial advance in Scandinavia and North America respectively. The numbered sections refer to the Bond events within the Holocene.

A second frequency of 1200-years was also recognized in the ASTC1 record. This cyclical pattern seen in the two proxies seems to be in phase with the GISP record especially for the last 3 ka, while for the older Holocene time interval an anti-phase relationship is displayed. This periodicity is close to ~1500-yr cyclicity as seen in the drift ice indices time series (Bond et al., 1997). This cyclicity is described by cooling and increased ice-rafting in the North Atlantic that occurs every 1500 years on average and is suggested to be driven by changes in ocean circulation and the freshwater release from ice sheets (Bond et al., 1997). Thus, implying that the cooler conditions described by Bond cycles and the Greenland ice record within the last 3 ka may be linked with an expression of drier conditions in the eastern Mediterranean region, as shown by the ASTC1 record (Figure 8).

OPEN IN VIEWER

Figure 8.

Gaussian Filter of ~1200 years applied on the detrended data from Drift ice indices in North Atlantic, GISP2 record and the ASTC1 (Ti/Al and Zr/Si) record. The green and blue bars refer to the glacial advance in Scandinavia and North America respectively. The numbered sections refer to the Bond events within the Holocene.

The spectral analysis of ASTC1 Zr/Si and Ti/Al time series also reveals a significant period of ~525 years over the Holocene, which may be linked with the reoccurrence of the Bond events or other events of similar frequency (Figure 9). A common periodicity of ~500 years during the Holocene is also found in solar activity (Cheng et al., 2015) and references therein), Greenland temperature (Stuiver et al., 1995), the North Atlantic Deep Water (NADW) circulation (Chapman and Shackleton, 2000), as well as temperature (Xu et al., 2014) and monsoon variability (Wang et al., 2005) in East Asia. Cave speleothems from Eastern Mediterranean (Cheng et al., 2015) and in China (Li et al., 2023), also show a significant ∼500-yr and ~550-yr periodicity respectively in-phased with our ASTC1 ~525-yr cycle, suggesting that the ∼525-yr cycle may be a large spatial-scale phenomenon. The above-mentioned observations imply teleconnections between a common forcing and various responses or feedback (Chapman and Shackleton, 2000). It is likely that atmospheric response to reduced solar irradiance may lead to coincident increase in North Atlantic drift ice, reducing the NADW intensity, cooling of both the ocean surface and high-latitude continent around North Atlantic and North Pacific, and weakening of the Asian monsoon (e.g. Cheng et al., 2012; Emile-Geay et al., 2007; Xu et al., 2014), triggering cyclical patterns of higher dust supply in the South Aegean region. The potential mechanism might be attributed to the Atlantic Meridional Overturning Circulations (AMOC) changes. This is because a strengthened (weakened) AMOC will not only result in a temperature increase (decrease) in China, but also lead to a weakened (strengthened) anticyclonic circulation over the Mediterranean via changing the surface temperature, contributing to a long-term increase (decrease) in the wet season Mediterranean precipitation (e.g. Delworth et al., 2022; Stockhecke et al., 2016).

OPEN IN VIEWER

Figure 9.

The Gaussian Filter of ~550 years applied on the detrended data from Drift ice indices in North Atlantic, GISP2 record and the ASTC1 (Ti/Al and Zr/Si) record. The green and blue bars refer to the glacial advance in Scandinavia and North America respectively. The numbered sections refer to the Bond events within the Holocene.