4.1. Regional temperature change and mechanism
The climate in Europe is strongly governed by the westerly winds which transport relative cool (warm) air masses from the Atlantic in summer (winter). In Slovakia, the North Atlantic Oscillation (NAO), with its positive (negative) phase associated with an increased (decreased) strength contrast of the Icelandic Low (IL) and the Azores High (AH), controls the westerly variability during October-March (Hurrell, 1995) and hence the temperature at the study site. The East Atlantic (EA) pattern (Barnston and Livezey, 1987) with its positive (negative) phase corresponding to a low (high) pressure anomaly in western Europe (centered at 55˚N and 20˚W), can also influence the position of the IL and the AH and the associated westerly routes. The October-March temperature at the Tatry meteorological station (1950-2012 AD) shows a strong correlation with sea-level pressure (SLP) anomalies at 60˚N and 10˚W (Fig. 3A, shades), resembling the correlation between the EA index (data from the Climate Prediction Center) and SLP (Fig. 3A, contours). The dipole correlation regions in the subtropics and the subpolar realm are also similar to the NAO pattern. The Tatry temperature time series (October-March) reveals some similarity to the NAO index (Fig. 3B; r = 0.31, n = 0.62, p < 0.05, 1950-2012; Jones et al., 1997) and the EA index (Fig. 3C; r = 0.41, n = 62, p < 0.01, 1950-2012; data from the Climate Prediction Center), corroborating that the NAO and EA modes both control October-March temperature at Zlomiska cave.
A negative (positive) NAO with reduced (enhanced) contrast between the IL and the AH lead to weak (strong) westerlies and less (more) warm air masses reaching the study area (Woollings et al., 2012). A negative (positive) EA, with a high (low) pressure anomaly over western Europe, reflects an increase (decrease) in atmospheric blocking frequency, which prevents (enables) the westerly winds to enter mainland Europe (Woollings et al., 2012; Madonna et al., 2017). The similarities between the ZL01 δ18O record and instrumental NAO and EA indexes (Mellado-Cano et al., 2019; Comas-Bru and Hernández, 2018) back to 1685 AD (Fig. 3E and F) further confirm that the NAO and EA are important drivers of temperature changes in the study area on multi-decadal to centennial time scales. The three multi-centennial cold periods registered by the minima in the ZL01 δ18O record (Fig. 2A) could hence reflect weak or absent westerly winds.
Correlation analysis between the EA index and European precipitation (Fig. 4A) shows that a low (high) EA index, with a high (low) frequency of atmosphere blocking over western Europe, can lead to a dry (wet) climate on the British Isles and western/central Europe. Periods of cold winters inferred from the ZL01 δ18O record match those with a dry climate inferred from low peat water-levels in Britain (Fig. 4B; Charman, 2010), high stalagmite Mg/Ca ratios in Bunker cave, NW Germany (Fig. 4C; Fohlmeister et al., 2012), high stalagmite growth rate in Roaring cave, Scotland (Fig. S8A; Baker et al., 2015), high Mg values (Fig. S8B) and high δ18O values in stalagmites of Herbstlabyrith cave, central Germany (Waltgenbach et al., 2021) (Fig. S8C). The dry periods could have been also affected by a negative NAO phase, which impacts precipitation on the British Isles (Charman, 2010) and in Germany (Fohlmeister et al., 2012) (Fig. S9). The coherency between the ZL01 δ18O time series and proxy records from Britain and Germany suggests that, during the intervals of ZL01 δ18O minima, weak westerlies resulted in reduced moisture transport to the British Isles and western Europe and low temperatures in Slovakia. This climate pattern likely resulted from strong atmospheric blockings over western Europe, similar to a negative EA or NAO phase.
4.2. Climate change and tribal migrations
The periods of three main tribal invasions into Europe, the Celtic invasion (ca. 4th-2nd century BC), the Great Migration (ca. 4th-6th century AD), and the invasion of the Magyars (ca. 9th-11th century AD) coincide with distinct intervals of anomalously cold winters as inferred from the ZL01 δ18O record (Figs. 2A), suggesting a possible link between of human migrations and climate. Severe cold winters are known to delay germination, lower the resistance to diseases, and reduce the yield and quality of crops (Hussain et al. 2018; Liu et al., 2020), eventually forcing humans to migrate. ZL01 Sr/Ca and Ba/Ca records suggest humid conditions during the Celtic invasion, the Great Migration, and the invasion of the Magyars (Fig. 2B). These data argue against the hypothesis of the drought-driven migrations (e.g., Drake, 2017) and instead emphasize the importance of temperature changes on human activities in central-east Europe. In addition to anomalously cold winters, low summer temperatures as recorded by tree-ring data (Büntgen et al. 2011) (Fig. S7) could also be mportant in triggering these tribal movements.
In central-eastern Europe, variations in the intensity of the westerlies might have had less impact on precipitation compared to temperature, as the atmospheric moisture decreases along the westerly tracks from the Atlantic Ocean. This is supported by the lack of a correlation between the NAO index and precipitation amount in the study area (Fig. S9). Accordingly, changes in the orientation and intensity of the westerlies led to temperature rather than precipitation changes in central-eastern Europe, different from the northern fringe of the Mediterranean Sea, which is more sensitive to hydroclimate changes (e.g., Cullen et al., 2000; Drysdale et al., 2006).
4.3. Possible mechanism
Cold periods with enhanced blockings over western Europe on decadal to centennial time scales suggest that the positions of the AH and the IL shifted towards high latitudes (Moore et al., 2013; Comas-Bru and Mcdermott, 2014). This poleward displacement of the AH and the IL was suggested to be induced by cooling of the high-latitude Atlantic Ocean, which shifts the region of transient eddy heat flux at 850 hPa northward (Rivière, 2009; Kown et al., 2020). This hypothesis is supported by the coherency between the ZL01 δ18O record and sea-surface temperatures (SST) along the path of the North Atlantic Current (Saenger et al., 2011) (Fig. S10B), with low SSTs corresponding to a weakened warm water transport from the tropical Atlantic to the polar region. Minima in ZL01 δ18O data are also in accordance with periods of weak Atlantic Multidecadal Variability (AMV) (Lapointe et al., 2020) (Fig. S10C). This suggests that a weak Atlantic Meridional Ocean Circulation (AMOC) reduces warm water transport from the equator to the polar regions, in turn leading to a cooling of the North Atlantic.
The NAO could also affect the (winter) temperature in the study area, but periods of δ18O minima are unrelated to negative NAO phases (Fig. S11) (Olsen et al., 2012; Ortega et al., 2015), possibly because (i) the NAO phase change was not the dominate forcing during the past three millennia in this region, and/or (ii) limitations of NAO reconstructions (Ortega et al., 2015) hamper such a comparison. Finally, intervals of high volcanic activity (Fig. S10A; Kobashi et al., 2017) show no relationship with minima in ZL01 δ18O, arguing against a volcanic forcing of these cold intervals.