Marine Isotope Stage (MIS) 7 (ca. 246–186 thousand years before present (ka), where present is 1950 CE) stands distinctively apart from the last several interglacial cycles. Following the low-amplitude glacial Termination III (TIII), MIS 7e was a relatively weak interglacial that returned to glacial conditions (MIS 7d) within 20 thousand years (ky) (PAGES, 2016). Glacial conditions terminated with a second deglaciation (TIIIa), which gave way to a second interglacial with three distinct substages (MIS 7c, b, a). Although the inception of an interglacial is ultimately paced by astronomical forcing, the precise timing (millennial scale) of each glacial termination and the strength of the successive interglacial depends on various global climate parameters and feedbacks. Investigating these internal forcings requires the intercomparison of precisely dated climate records in order to identify leads, lags, and synchronicities across different climate zones. The role of forcings that led to the last two glacial terminations and the variability within the last two interglacial periods (MIS 5e and Holocene) have been well studied. Obtaining records from MIS 7 and the penultimate glaciation that contain sufficient resolution and chronological precision to make meaningful regional comparisons, however, has proved challenging. In the North Atlantic realm, marine records show that major meltwater pulse events punctuated a period of rapid sea surface warming associated with TIII and TIIIa (e.g., Channell et al., 2012; Hodell et al., 2008; Martrat et al., 2004, 2007). The chronologies of these marine records, however, are frequently dependent on orbital tuning or alignment to global benthic δ18O stacks, which feature large uncertainties of ±10 ka during MIS 7 (e.g., LR04 stack; Lisiecki and Raymo, 2005). In continental Europe, cave and lacustrine records indicate abrupt environmental shifts associated with the MIS 7 substages (e.g., Tzedakis et al., 2003; Roucoux et al., 2008; Despart et al., 2006) and TIII (Pérez-Mejías et al., 2017) accompanied by regional temperature and atmospheric circulation changes (e.g., Spötl et al., 2008; Badertscher et al., 2011; Columbu et al., 2019) often in step with North Atlantic climate change (Denniston et al., 2018). However very few terrestrial records span the entirety of MIS 7 and its glacial transitions. A lack of absolute-dated, continuous records hinders the ability to capture the timing and full duration of regional climate changes, resulting in critical gaps in knowledge.

This study aims to determine the onset of warming in continental Europe associated with TIII and TIIIa, as well as the regional climate variations during the five substages of MIS 7. To do so, we turn to the Austrian Alps (Fig. 1). Over the last century, the Alps experienced twice the amplitude of temperature change relative to the mean Northern Hemisphere (Auer et al., 2007), contributing to a 45 % reduction of glacier surface and roughly 50 % of ice volume between 1900 and 2011 (Huss, 2012). This high degree of sensitivity renders the Austrian Alps an ideal location to pinpoint the timing of continental temperature variations. Here, we focus on Spannagel Cave – a high-altitude marble cave located in the central Austrian Alps. Drip waters in this cave are fed by glacial and snowpack melt (Mangini et al., 2005). Past studies demonstrate that Spannagel speleothem δ18O represents a robust proxy of local winter temperatures and moisture sources, the latter being dominated by the North Atlantic Ocean (Mangini et al., 2005; see Sect. 2.1). Pioneering work by Holzkämper et al. (2005) and Spötl et al. (2007, 2008) presented multiple MIS 7 stalagmites and flowstones deposited in Spannagel Cave. Since these studies, new developments in the measurement precision of U and Th isotopes by Cheng et al. (2013) and Craig et al. (2016, 2017) have allowed for ultra-high age precision, with 2σ absolute age uncertainties averaging ±300 years for MIS 7. In addition, two newly discovered stalagmites from Spannagel Cave provide further insight into the penultimate glacial inception (MIS 7–6 transition). The resulting Spannagel δ18O record provides a high-precision chronology of abrupt climate change in Europe during MIS 7 while addressing the need for absolute-dated paleorecords from regions that are sensitive to the North Atlantic realm.

Spannagel Cave (Fig. 1; 47∘04′54′′ N, 11∘40′2′′ E; 2300 to 2530 ma.s.l.) is located above the timberline in the vicinity of a retreating glacier (Hintertux Glacier) close to the main Alpine crest. Previous publications provide details on the setting of this cave (Spötl et al., 2004; Spötl and Mangini 2007, 2010). A few salient features relevant for this study are summarized here. Spannagel Cave developed in a ca. 20 m thick Upper Jurassic marble (Hochstegen Formation) sandwiched between gneiss bedrock. An interesting feature of Spannagel Cave drip waters is that they are rich in U. Cave water U concentrations range from 5 ppb (small cave stream) to 33 ppb (stalactite drip water). As a result, calcitic speleothems from this cave are exceptionally high in U concentrations (3 to 399 ppm). The U is most likely sourced from the overlying gneiss which yields significantly higher U concentrations (3.4–12.0 ppm) than the marble (0.3–2.2 ppm).

Spannagel is a well-ventilated cave system with stable air temperatures in the interior of the cave that stay within a narrow range of +1.8 to +2.2 ∘C (measured at several sites using data loggers over a period of about 20 years). These values are about 2 ∘C higher than the mean annual air temperature at the elevation of the cave, which is a reflection of the small positive thermal anomaly due to the ascending geometry of the cave (chimney effect; Spötl and Pavuza, 2016). Relative humidity in the cave interior is invariably higher than 96 %. Today, the actively retreating Hintertux Glacier terminates ca. 500 m south of Spannagel Cave. During the last glacial maximum, the cave was completely covered by as much as 150–250 m of glacial ice (van Husen, 1987). The deposition of speleothems during several cold climate periods suggests that the Hintertux Glacier has remained largely temperate, or warm-based (Spötl and Mangini, 2007). Under glacial conditions, it is thought that the oxidation of pyrite in the host rock allowed karst dissolution beneath the glacier without the input of soil-derived carbon dioxide (Spötl and Mangini, 2007).

Stalagmite SPA121 (Fig. S1 in the Supplement) was found in the northern segment of the cave embedded in unconsolidated silty sand which was most likely transported into the cave by high-discharge streams during the last deglaciation. The stalagmite is 19.3 cm tall and was found attached to a platy and angular piece of calcite-cemented gravel. The surface of the sample is covered by a thin layer of medium gray, clay-rich calcite, but the interior preserves mostly transparent, inclusion-poor calcite, showing a striking pattern of cracks. The origin of these cracks is enigmatic but may be related to freezing of the cave or the sediments in which the sample was embedded during peak glacial periods. SPA183 is a 39 cm tall stalagmite collected from the central part of the cave and a slightly higher altitude than SPA 121. SPA183 revealed three distinct growth axes that are marked by variations in color (Fig. S1). SPA146 is a 24 cm tall stalagmite which grew in the same small chamber as SPA183. Both stalagmites were detached from their growth substrate and embedded in sand but likely transported less than a few meters. SPA146 shows two distinct growth axes (Fig. S1).

Stalagmites SPA121, 146, and 183 were halved and polished. Subsamples for U–Th dating (n = 40) were hand drilled along the growth axis of halved stalagmites. Subsample trenches were drilled no larger than 1 mm in width, such that the sampling error (∼ 175 years) remained within the analytical uncertainties (see Results). Target sample weights ranged from 200 to 20 mg in concordance with changes in U concentration. Subsamples were spiked with a mixed 233U–236U–229Th spike similar to that described in Edwards et al. (1987). Procedures for U and Th chemical separation and preparation of reagent solutions follow the methods described in Edwards et al. (1987) and Shen et al. (2002).

U and Th isotopic measurements were made on a Thermo Scientific Neptune Plus MC-ICP-MS following the instrument calibration and the Faraday cup measurement method described in Cheng et al. (2013). In addition, a 1013 Ω amplifier was installed within the detection system in order to collect low ion beam intensities (e.g., 234U and 230Th) (Craig et al., 2016, 2017, and references therein). The methods of gain calibration and dynamic time correction of the high resistor are largely based on Craig et al. (2016, 2017). Each sample was measured for 300 s or longer. Intensities of 234U and 230Th beams were on average 15 and 5 mV, respectively.

Stable isotope samples were micromilled continuously along the growth axis of each stalagmite at 0.15–0.20 mm increments. A total of 946 stable isotope measurements on stalagmite SPA121 were previously published in Spötl et al. (2008), 890 and 583 new stable isotope measurements from stalagmites SPA183 and SPA146, respectively, are reported here. A total of 13 Hendy tests were drilled along individual growth layers from stalagmites SPA146 and SPA 183 (Figs. S4 and S5 in the Supplement). All calcite powders were analyzed using a GasBench II coupled with a Delta V Plus isotope ratio mass spectrometer. The 1σ precision is 0.06 and 0.08 ‰ for δ13C and δ18O, respectively. Results are reported relative to Vienna Pee Dee Belemnite (VPDB).

Resulting U–Th ages and their respective replicates are in stratigraphic order within uncertainties (Table S1). This study focuses on the MIS 7 portion of stalagmite SPA121, which was deposited without interruption between 248.5 and 191.5 (±1) ka (see Spötl et al., 2008, for details on all SPA121 growth phases). New U–Th ages for this stalagmite fall within age uncertainties of the previously published ages (Spötl et al. 2008; Fig. 2). Similar δ234Ui values and 232Th and 238U concentrations further underscore a high degree of reproducibility between the two SPA121 data sets, which were measured in different laboratories using different instruments (MC-ICP-MS vs. TIMS). New SPA121 ages improve the precision of previously published age uncertainties by an order of magnitude (Fig. 2), from an average of 1.7 % to 0.17 %.

The late MIS 7 growth phase of stalagmites SPA183 and SPA146 occurred between 191.9–190.6 (±0.6) ka and 191.6–182.3 (±0.3) ka, respectively. The exact onset of growth is unknown, as both stalagmites show evidence of diagenesis spanning first 2–3 cm of the late MIS 7 growth phase (Fig. S2 in the Supplement; Table S1 in the Supplement). Evidence of diagenetic alteration includes a conspicuously white, milky calcite fabric and U–Th ages that are out of stratigraphic order and unable to be replicated. As a result, this study focuses only on the unaltered portion of the late MIS 7 growth phases of stalagmites SPA183 and SPA146 (Fig. S1). Relative age uncertainties of SPA146 and 183 average 0.09 % and 0.16 %, respectively. Significantly higher δ234Ui values from SPA146 and 183 relative to SPA121 suggest differing drip sources between the first two neighboring stalagmites and SPA121. Growth rates calculated from new SPA121 ages closely agree with previously published data in Spötl et al. (2008) (Fig. 2). The average growth rate is 5.6 µmyr-1, excluding one period of exceptionally low growth rate (0.8 µmyr-1) between 231.1 and 219.6 (±0.6) ka. In contrast, average growth rates of SPA183 and SPA146 are higher (63 and 140 µmyr-1, respectively). Differences in growth rates are likely due to differing drip sources.

SPA121 stable isotope values used in this study are from Spötl et al. (2008). New stable isotope data from SPA146 and 183 are reported in Table S2 in the Supplement. The range of SPA121 δ18Oc values (-8.1 ‰ to -14.7 ‰) and δ13C values (9.7 ‰ to 0.8 ‰) is in agreement with newly measured δ18Oc values from SPA146 and SPA183 (Fig. S3 in the Supplement). Slight (< 1 ‰) offsets in absolute values between stalagmites are observed, but do not influence the relative variations in stable isotopes which are the focus of this study. Hendy test results (Figs. S4 and S5; Spötl et al., 2008) show no evidence for significant kinetic fractionation during the deposition of all three stalagmites.

The δ18Oc signature of SPA stalagmites is depleted during the MIS 7 warm intervals (average -9.2 ‰) relative to the Holocene (average -7.8 ‰; Spötl et al., 2004), which suggests cooler winter temperatures during MIS 7. This is consistent with globally distributed evidence suggesting cooler Northern Hemisphere temperatures throughout MIS 7 (PAGES, 2016) in conjunction with lower sea levels (Robinson et al., 2002; Thompson and Goldstein, 2005; Dutton et al., 2009; Andersen et al., 2010; Murray-Wallace, 2002) and atmospheric pCO2 (Bazin et al., 2013) relative to the Holocene.

The δ13C signature of all three stalagmites is higher relative to modern and Holocene speleothems (-10 ‰ to -7 ‰), reflecting a signal that is buffered by the isotopic composition of the host rock (Spötl et al., 2004). High and commonly positive δ13C values indicate no significant input of organic C into the system, thereby indicating an absence of soil and vegetation above the cave throughout the duration of speleothem deposition (Spötl et al., 2008). Cooler surface temperatures and the absence of soil argue for a significantly larger Hintertux Glacier during MIS 7 relative to today, likely covering a large portion of the cave. The growth of speleothems during cold climate periods of MIS 7 is likely due to the warm-based nature of the glacier, which provides a supply of meltwater while preventing the cave from freezing.

The new Spannagel δ18O record spans the period of Northern Hemisphere warming associated with the later portion of TIII, the five substages of MIS 7, TIIIa, and the MIS 7–6 glacial inception (Fig. 3). Due to our unprecedented age control, we can determine the precise timing of regional changes associated with each MIS 7 substage, as well as the onset of warming in the Austrian Alps associated with TIII and TIIIa (Fig. 4). While terrestrial records cannot directly date changes to the ocean–cryosphere system during a glacial termination, the climatic excursions in high-sensitivity regions, such as the Austrian Alps, provide key temporal constraints on the climate events leading up to and during these transitions.

Following the start of speleothem growth at 247.3 ± 0.2 ka, an abrupt shift towards higher δ18O values at 242.5 ± 0.3 ka marks the onset of regional warming associated with the TIII deglaciation. The ensuing interglacial period (MIS 7e) is characterized by high δ18O values (greater than -10 ‰) and spanned 5 ky from 241.8 to 236.0 (±0.3) ka. Depleted δ18O values (less than -12 ‰) between 234.3 and 216.9 (±0.3) ka coincide with MIS 7d. Maximum regional cooling occurred between 231.3–228.6 (±0.2) ka. An abrupt shift towards higher δ18O values at 216.8 ± 0.3 ka marks the onset of regional warming associated with TIIIa. Two periods of high δ18O values between 215.7–212.9 (±0.4) ka (> -10 ‰) and 201.8–197.1 (±0.5) ka (-8.7 ‰) coincide with interglacial periods MIS 7c and 7a, respectively. A final shift towards lower δ18O values from 197.1 to 191.4 (±0.3) ka coincides with the MIS 7–6 transition, the latter portion of which is replicated by stalagmites SPA 146 and 183.

On millennial timescales, remarkable similarities are observed between Spannagel δ18O and paleorecords that are sensitive to the North Atlantic climate (Fig. 4). These similarities highlight the rapid climatic link between the European Alps and North Atlantic realm, as observed in later interglacial periods (e.g., Holzkämper et al., 2004; Mangini et al., 2007; Wilcox et al., 2020) and glacial periods (e.g., Moseley et al., 2014; Mayr et al., 2019). Most striking is the millennial-scale covariance of Spannagel δ18O and Chinese monsoon δ18O, which is explained through the following teleconnections: temperature anomalies in the North Atlantic region influence the intensity of heat transport by northern Hadley cell circulation, which triggers a latitudinal shift in its ascending branch, known as the Intertropical Convergence Zone (ITCZ). Latitudinal shifts in the ITCZ, in turn, influence the Chinese monsoon strength (see Cheng et al., 2016, for details). Thus, both Chinese and Spannagel speleothems respond to common climate forcings on millennial timescales. Identifying the mechanisms behind North Atlantic-forced excursions in Spannagel δ18Oc is challenging due to the complex array of processes that influence Spannagel δ18O but is likely linked to synchronous temperature changes and/or latitudinal shifts in the westerlies. In alignment with previous work, we interpret Spannagel δ18Oc as a faithful recorder of millennial-scale changes in the North Atlantic realm during MIS 7.

For the remainder of this discussion, we will examine the variations in Spannagel δ18O associated with MIS 7 in order to provide new temporal constraints on climate changes in the Austrian Alps, as well as new insights into millennial-scale changes in the North Atlantic leading up to TIII and TIIIa.

The response of climatically sensitive regions to glacial–interglacial cycles and their abrupt transitions provides key insight into the timing of global climate change. In this study, we present the first ever paleorecord of MIS 7 with relative age uncertainties < 2 ‰. Using this chronology, we can determine the precise timing and duration of climate variations in the Austrian Alps in response to the MIS 7 substages and associated glacial terminations. Following the start of speleothem growth, an abrupt increase of δ18O values at 242.5 ± 0.3 ka marks the onset of regional warming associated with TIII. The ensuing interglacial period (MIS 7e) is characterized by enriched δ18O values (> -10 ‰) from 241.8 to 236.0 (±0.3) ka. Depleted δ18O values (< -12 ‰) between 234.3 and 216.9 (±0.3) ka coincide with the Northern Hemisphere cool period. Similar to TIII, a brief negative excursion in Spannagel δ18Oc prior to the abrupt rise associated with TIIIa suggests that a major meltwater pulse occurred in the North Atlantic around 216.9 ka. An abrupt shift towards higher δ18O values at 216.8 ± 0.3 ka marks the onset of regional warming associated with TIIIa. Two periods of high δ18O values between 215.7–212.9 (±0.4) ka and 201.8–197.1 (±0.5) ka coincide with interglacial periods MIS 7c and 7a, respectively. A final shift towards lower δ18O values from 197.1 to 191.4 (±0.3) ka coincides with the MIS 7–6 transition. In total, this multi-stalagmite record provides important chronological constraints on climate shifts in the Austrian Alps associated with MIS 7 while providing new insight into the timing of millennial-scale changes in the North Atlantic realm.