North Atlantic storm track changes during the Last Glacial Maximum recorded by Alpine speleothems

The European Alps are an effective barrier for meridional moisture transport and are thus uniquely placed to record shifts in the North Atlantic storm track pattern associated with the waxing and waning of Late-Pleistocene Northern Hemisphere ice sheets. The lack of well-dated terrestrial proxy records spanning this time period, however, renders the reconstruction of past atmospheric patterns difficult. Here we present a precisely dated, continuous terrestrial record of meteoric precipitation in Europe between 30 and 14.7 ka. In contrast to present-day conditions, our speleothem data provide strong evidence for preferential advection of moisture from the South across the Alps supporting a southward shift of the storm track during the local Last Glacial Maximum (that is, 26.5–23.5 ka). Moreover, our age control indicates that this circulation pattern preceded the Northern Hemisphere precession maximum by ~3 ka, suggesting that obliquity may have played a considerable role in the Alpine ice aggradation.


Supplementary Figure 2 The 7H LGM-speleothems
The two stalagmites were found broken on a collapsed sediment bank. The samples show a similar candle-shaped morphology, which suggests a constant drip rate throughout the period of formation. The absence of significant morphological changes suggests that the mean saturation index (Sicc) must have remained constant throughout the growth history (ca. 0.35 to 0.45 for discharge rates of ca. 0.1 l/day; 5). The 568 mm-long 7H-2 stalagmite, ca 40 mm in diameter, is honey-coloured. Corrosion features are observed on the outer surface, including a small pool at the drip impact on top of the stalagmite. Approximately 57 mm distance from top (DFT), the growth axis shifted by ca. 10 mm. Age modelling (StalAge; 6) based on 29 multi-collector MC-ICP-MS U/Th ages associates this shift with a ca. 3ka hiatus but, otherwise, reveals a regular deposition between 30 and 17.2 ka with an average growth rate of 66±27 µm a -1 (Supplementary Figure 4). 7H-3 is 302 mm long and shows very similar features to 7H-2 (shape, colour, corrosion patterns). A shift in the growth axis is observed 43 mm DFT likely associated with the subsidence of the fine-grained sediment on which it grew. U/Th ages reveal a very regular growth of 39±11µm a -1 , between 22.9 and 14.6 ka, and no sign of a hiatus. Scale bar: 50 mm.

Supplementary Figure 3 Thin section photomicrographs of 7H stalagmites under crossed polars
The stalagmite petrography is characterized by two types of fabrics comprising bundles of compact and more porous fascicular-optic calcite crystals showing typically divergent optic axes (7). Both fabrics indicate that the calcite contains Mg in excess of 1000 ppm (8). Fabric (A) consists of close-packed bundles of translucent, elongated columnar, fascicular-optic calcite crystals while fabric (B) is characterized by secondary porosity (greenish arrow) where crystal tips show signs of dissolution followed by reprecipitation of single columnar calcite (red circles). This replacement is associated with more variable hydrology, explaining also the presence of impurities (yellow arrows). Since this early diagenesis occurred prior to, or simultaneous to the deposition of the subsequent layer of fascicular optic bundles, the process has not altered (significantly) the chemical properties of the crystals, and in particular the  18 O signal, which is believed to reflect that of the parent water. Similar fascicular-optic fabrics have also been observed in cryogenic carbonates (9; 10; 11), which could be associated with microbial oxidation of sulfides present in the host-rock. Microbial oxidation would have likely enhanced dissolution of the bedrock subglacially at low temperature (12) and generated most of the HCO 3 by the dissolution of the (marine) carbonate rocks, as indicated by the overall positive  Age modelling (StalAge; 6) reveals rather regular growth rates of 40 to 70 µm a -1 on average. Coeval deposition is observed between 23.4 and 20.6 ka as well as between 17.9 and 17.3 ka. A major hiatus identified in 7H-2 between 20.5 and 17.8 ka is supported by petrographical changes showing a shift in the growth axis. In contrast, the absence of clear petrographic evidence for a proper growth stop between 28.7 and 27.8 ka (i.e. during Stadial-4) rather suggests that this period was associated with slow deposition. Vertical error bars show the 2 analytical uncertainty on the U/Th ages. C analyses along individual growth layers (i.e. Hendy Tests) of 7H-2 and 7H-3. Vertical error bars show the analytical uncertainty (1) on stable isotope data. The more positive values observed in 7H-3 suggest that kinetic fractionation affects this sample more strongly. This result is consistent with a somewhat lower drip discharge as inferred from the speleothem growth rate. Overall, the elevated  13 C values reflect the isotopic composition of the host rock and strongly support the lack of soil-derived organic carbon in the karst system. The higher  13 C values observed in 7H-3 suggest an early CO 2 degassing associated with a low drip rate.  The measured trace elements all derive from the dissolution of the marine Schrattenkalk carbonates. The elevated magnesium concentration is consistent with the speleothem petrography, suggesting slow but regular drip rates. The presence of sulphur results from the oxidation of pyrite disseminated in the host rock. This interpretation is consistent with secondary gypsum deposits ( 34 S: -32 to -26.3 ‰; 17), which represent a common feature in the Sieben Hengste cave system. Because sulfide oxidation enhances the dissolution capacity of seepage water, it also represents a major process for speleothem deposition in periglacial karst areas (18). It is noteworthy that Y, an element which is commonly associated with colloidal transport from the soil, is largely below detection limit (~13 ppb). No chemical laminations were observed suggesting regular drip chemistry associated with a well-mixed aquifer. ; i.e. values in secular equilibrium with a bulk Earth 232 Th/ 238 U value of 3.8. The errors are arbitrarily assumed to be 50%. *** a b2k stands for before 2000 AD.

Supplementary Discussion
The modern annual precipitation at the study site averages 2000 mm, nearly 40% of which fall as snow between November and April. The  18 O of precipitation typically averages -13.1±0.6 ‰-VSMOW with a seasonal amplitude of ca. 10 ‰ (21). 10-day backward trajectories calculated from an ERA-Interim reanalysis data for the years 1995-2005 were started at 6 hours intervals on a 60x60 km horizontal grid with 30 hPa vertical intervals. A quantitative Lagrangian moisture source diagnostic (22, 23) reveals a dominant transport direction from the West, which is particularly pronounced during the winter season (Supplementary Figure 6). However, a seasonally varying fraction of moisture ranging on average between 10 and 30 % also reaches the 7H site from the South. Our method underestimates the seasonal variability since convective (summer) precipitation is less reliably captured. From climatological studies based on precipitation measurements, it is known that about 1/3 of the precipitation along the Northern Alps falls during the summer season, with the remaining 2/3 distributed equally over the other seasons (24).
While Mediterranean moisture sources can contribute substantial amounts of precipitation to the southern Alps during spring and autumn, cluster analyses of extreme rainfall events in northwestern Italy suggest that the strongest precipitation events are likely associated with significant moisture advection from the North Atlantic during periods of intense northward flow to the southern Alpine range (25). An increasing number of studies provide evidence that the North Atlantic is the dominant oceanic moisture source of alpine precipitation also in periods of strong southerly moisture advection (26). Considering drying ratios as low as 35 % (27) a substantial amount of North Atlantic sourced moisture may eventually reach the Sieben Hengste, also with southerly transport routes.
The 18 O depletion associated with orographic precipitation depends predominantly on thermodynamic effects (28) with topography being the dominant control on the isotope ratio of precipitation. Empirical observations from the Andes (13) confirm that   Figure 7). Considering all other parameters equal and assuming an average topographic barrier south of the study site at ca. 3000 m a.s.l., the potential difference in the  18 O signal associated with southerly advection reaches 5 to 7 ‰ as compared to the direct route from the North-West. Using the Rayleigh fractionation model a two endmember mixing model has been constructed for different surface temperatures. Given the approximate 50% threshold value at 16.3‰  18 O-VSMOV (Figure 4), a best matching mixing model was obtained with end-member values of 19.7‰ for the Alpine main crest and 12.8‰ for the direct approach to 7H, based on an initial surface temperature of 6.5ºC. This transfers the  18 O signal of between 1.5 and 3 ‰ as inferred from the peak glacial  18 O shift in the 7H-record to a southern humidity fraction ranging from 25 to 65 % of the average annual precipitation at Sieben Hengste during the LGM. Sensitivity of the estimate is about 10% less southerly contribution per degree in surface temperature change.
In contrast to the storm trajectories, local temperature effects are assumed to play only a second order role in the isotope fractionation affecting meteoric precipitation reaching the Sieben Hengste. Changing the temperature of condensation is expected to produce comparatively small amplitudes in isotope fractionation on palaeoclimate timescales. Typically, major precipitation events go along with substantial advection of warm moist air, which, due to its transient nature, is hardly reflected in the mean annual air temperature record. For example, temperature of precipitation in Greenland is up to 15 K warmer than the annual mean temperatures (22).
From present-day synoptic analysis it is well known that events of heavy precipitation in the Southern Alps are typically associated with intense moisture advection from the North Atlantic, in some cases tapping into the subtropical pool of high humidity (25). Such cases of meridional moisture advection are typically associated with a trough signature at mid-tropospheric levels, also known as narrow potential vorticity streamers. The jet stream bends around the outside of these troughs (streamers), and induces the dominant transport pattern in the troposphere below. When the Fennoscandian ice cap was present during the LGM, atmospheric motion was different as compared to today, (i) by a equatorward shift of the sea-ice margin, and thus the zone for baroclinic storm development, and (ii) by inducing a frequent cold high-pressure residing over the ice cap, blocking the flow and deflecting the jet stream, inducing pronounced undulations, known as Rossby wave breaking. With the tropics (and subtropics) cooling less than high-latitudes during the LGM, substantial moisture should have been available in proximity to the equatorward shifted baroclinic zone (storm track) and the overlying jet stream. As the disturbances due to the blocking anticyclone not just die off, but can continue to influence weather downstream, it is possible that additional wave breaking was induced over central and eastern Asia, triggering again strong meridional advection patterns underneath upper-level jet excursions. This time the systems would not transport moisture towards the Alps, but generate storms that could transport large amounts of dust into higher (Arctic) latitudes. Additional or alternative mechanisms may have included (i) increased deflation of aeolian material in the dust sources, (ii) increased wind speed due to changed land cover, (iii) larger deserts, (iv) less effective removal processes, (v) a more direct transport pathway towards the Arctic.