Early last glacial maximum in the southern Central Andes reveals northward shift of the westerlies at ∼ 39 ka

. The latitudinal position of the southern westerlies has been suggested to be a key parameter for the climate on Earth. According to the general notion, the southern westerlies were shifted equatorward during the global Last Glacial Maximum (LGM: ∼ 24–18 ka), resulting in reduced deep ocean ventilation, accumulation of old dissolved carbon, and low atmospheric CO 2 concentrations. In order to test this notion, we applied surface exposure dating on moraines in the southern Central Andes, where glacial mass balances are particularly sensitive to changes in precipitation, i.e. to the latitudinal position of the westerlies. Our results provide robust evidence that the maximum glaciation occurred already at ∼ 39 ka, signiﬁcantly predating the global LGM. This ques-tions the role of the westerlies for atmospheric CO 2 , and it highlights our limited understanding of the forcings of atmospheric circulation.


Introduction
The southern westerlies are an important driver for upwelling around Antarctica and deep ocean ventilation (Toggweiler et al., 2006). As atmospheric CO 2 is constantly and naturally removed in large quantities by marine organisms, and the respired CO 2 is accumulating in the deep ocean, changes in upwelling might have a substantial affect on the concentration of atmospheric CO 2 and climate on Earth. It has been suggested that a more equatorward position of the westerlies during glacials resulted in a weakening of the northward Ekman transport of surface waters, reduced deep ocean ventilation, and low levels of atmospheric CO 2 (Toggweiler et al., Correspondence to: R. Zech (godotz@gmx.de) 2006). This hypothesis builds on the general notion that the westerlies were shifted northward during the global LGM, which has been inferred from terrestrial paleoecological and marine records off Chile (Heusser, 1989;Stuut and Lamy, 2004). However, there is a long-standing controversy regarding the interpretation of pollen records in Patagonia (Markgraf, 1989;Markgraf et al., 1992), and marine records may reflect a complex mixture of regional and local up-welling and environmental signals (Stuut et al., 2006). Additionally, climate modeling results do not show significant shifts of the westerlies during the global LGM (Rojas et al., 2008).
In order to contribute to the reconstruction of the position of the southern westerlies, we applied 10 Be surface exposure dating on moraines in the southern Central Andes. Whereas glaciers in Patagonia always received plenty of precipitation and reached their maximum extents at times of low temperatures, i.e. in-phase with the global LGM (Douglass et al., 2006;Kaplan et al., 2008), glacier mass-balances in more arid environments, such as the southern Central Andes, become more precipitation-sensitive, and the glacial chronologies there thus also reflect changes in the position of the westerlies (Kull et al., 2002.

Geographical setting and surface exposure dating
The southern Central Andes north of ∼40 • S are characterized by strong precipitation gradients in north-south and eastwest directions (Fig. 1). They are therefore an ideal location to record latitudinal shifts of the westerlies. We applied 10 Be surface exposure dating in the Rucachoroi Valley, Central Argentina (39 • 14 S, 71 • 11 W) (Fig. 2). The valley is located east of the Andean divide, but receives most of its precipitation from the Pacific related to the seasonal northward shift of the westerlies in austral winter (Garreaud and Aceituno, 2007). Weather station data from Bariloche (41 • 15 S) indicate ∼800 mm mean annual precipitation with a pronounced austral winter maximum and 8 • C mean annual temperature at 840 m above sea level (a.s.l.). No glaciers exist today in the Rucachoroi Valley despite summits reaching >2000 m a.s.l. Extensive glaciation in the past is documented by the U-shaped valley form, glacial over-deepening (formation of Lake Rucachoroi at 1230 m a.s.l.), and a prominent right-lateral moraine merging into a lateral-frontal moraine at ∼1200 m a.s.l. (Fig. 2).
We sampled and analyzed seven boulders from the prominent lateral and lateral-frontal moraine in order to determine the timing of the maximum glaciation, as well as six additional boulders and polished bedrock from further up-stream to constrain the deglaciation history (Fig. 2). The sampling strategy followed standard guidelines, collecting ∼0.5 kg of rock material from the flat top of large, stable boulders or polished bedrock surfaces. Exact sample locations were determined using a handheld GPS and documented with photographs. Sample preparations followed standard laboratory techniques including separation of quartz, addition of a 9 Be carrier, dissolution in HF, and chromatographical purification of beryllium. The 10 Be/ 9 Be AMS measurements were conducted at the ETH AMS facility in Zurich. We used the CRONUS-Earth online calculator (http://hess.ess. washington.edu) applying the scaling model of Lifton et al. (2005) to calculate the surface exposure ages (all sample data are provided in the supplementary Table 1). Note that our conclusions are independent of the choice of the scaling model (see Supplement Table 2).

Results
The exposure ages from the moraines range from 30.2 ± 3.0 ka to 38.8 ± 3.9 ka (Fig. 2). The observed scatter mainly reflects geological uncertainties due to post-depositional processes, and the generally low probability of pre-exposure justifies the application of the "oldest age model" (Putkonen and Swanson, 2003;Zech et al., 2005). Following this concept, the oldest boulder from a moraine is the best available minimum estimate for the true deposition age, unless it is a statistical or stratigraphic outlier. Accordingly, we can infer that the maximum ice extent in the Rucachoroi Valley occurred at ∼39 ka, clearly pre-dating the global LGM.
The exposure ages of 17.9 ± 1.8 ka and 14.8 ± 1.5 ka from the samples RU31 and RU32 (Fig. 2) show that the glacier continued to occupy the upper part of the Rucachoroi Valley and started to retreat beyond the Rucachoroi Lake only after the global LGM. The absence of moraines other than the prominent ones dated to ∼39 ka likely indicates that no prominent glacial re-advances occurred between ∼39 ka and ∼18 ka. Deglaciation of the upper part of the Rucachoroi Valley occurred after ∼18 ka, and most of the valley became ice-free by ∼15.5 ka (RU11: 15.5 ± 1.5 ka, RU21: 15.5 ± 1.5 ka). Only small, sporadic cirque moraines provide geomorphological evidence for minor glacial readvances and climate reversals during the Lateglacial. One respective exposure age of 12.2 ± 1.2 ka (RC12) has been obtained so far, but will have to be corroborated in future studies.

Comparison with other paleoclimate proxies
The early local LGM in the Rucachoroi Valley and deglaciation after ∼18 ka is consistent with findings from the Chilean Lake District (CLD) (Fig. 1). There, extensive radiocarbon dating showed that the piedmont glaciers reached maxima at ca. 29.6, 26.9, 23.1, 21.0, 14.9 and 13.9 14 C ka BP (∼35, 31, 28, 25, 18 and 17 cal ka BP), with earlier advances being notably more extensive in the northern CLD (Denton et al., 1999;Lowell et al., 1995). Further north, in the Mendoza Andes (∼33 • S), Argentina, the 'Penitentes Till' is constrained by a minimum thermo-luminescence age of 31.0 ± 3.1 ka, as well as a U/Th age of 38.3 ± 5.3 ka from travertine (Espizua, 2004). Ultimately, exposure ages from the Dona Rosa Valley (∼31 • S), northern Chile, have shown that the maximum datable advance there occurred at ∼39.0 ± 4.1 ka (Zech et al., 2007. Thus, in summary, there is inevitable evidence that the local LGM in the southern Central Andes from 30 to 40 • S significantly predated the global LGM on both the western and eastern side of the divide. We argue that this early LGM reflects increased precipitation and a northward shift of the westerlies, because glacier mass balance studies show that glaciers become very sensitive to changes in precipitation (rather than temperature) in arid and semiarid environments (Kull et al., 2002. This precipitation-sensitivity is evident today from the pronounced increases in equilibrium line altitudes from Patagonia to the Central Andes, as well as from contrasting glacial histories during the Little Ice Age, which have been explained with varying dominance of precipitation and temperature controls on mass balance (Luckman and Villalba, 2001). Analogously, differences in the timing of the glacial maxima during the last glacial have to be explained: Glaciers in Patagonia always received plenty of precipitation, are therefore mainly sensitive to temperature changes and reached their maximum extents in-phase with the global LGM at ∼25 ka (Douglass et al., 2006;Kaplan et al., 2008). An early LGM in the more arid southern Central Andes, on the other hand, requires increased precipitation compared to the global LGM.
Our interpretation is corroborated by glacier-climate modeling results from the nearby Las Leñas Valley (35 • S), where the most extensive moraines (tentatively dated to ∼42 ka based on preliminary exposure ages) were deposited under climate conditions much more humid than today (60-150% increase in mean annual precipitation, 5.9-8 • C decrease in mean annual temperature, Wäger et al., unpublished manuscript). Other, non-glaciological findings are also in agreement with this notion. Pollen records from ∼25 • S, for example, indicate increased winter precipitation between 40-33 ka and between 24-17 ka (Maldonado et al., 2005), and lake sediments from Central Chile show high lake levels of Laguna Tagua Tagua (34.5 • S) between ∼40 and 17 ka (Valero-Garcés et al., 2005). The circumpolar nature of the westerlies' shift should be investigated in more detail in the future, but extensive glacial advances, high lake levels and river runoff maxima are documented, for instance, also in southern Australia at ∼35 ka (Barrows et al., 2001;Kemp and Spooner, 2007). There, an even more extensive glacial advance has been dated to ∼60 ka, which has not yet been identified in the southern Central Andes.

Discussion of the forcings of the westerlies
What does the proposed northward shift of the westerlies at ∼39 ka mean in terms of our understanding of the forcings responsible for the latitudinal position of the southern westerlies? Traditionally, low Antarctic temperatures and extensive sea ice have been invoked to push the westerlies northward during the global LGM (Heusser, 1989;Stuut and Lamy, 2004). This is, however, difficult to fully reconcile with our findings, as neither sea ice was particularly extensive at ∼39 ka, nor were Antarctic temperatures at a minimum (Fig. 3a, c, e). Furthermore, we can rule out seasonal insolation/temperature as a prominent forcing, because winter insolation is high between ∼40 and 20 ka (Fig. 3d), which would favor a less pronounced seasonal northward shift of www.clim-past.net/7/41/2011/ the westerlies and is at odds with more humid conditions documented for that very time period.
In search for other potential forcing mechanisms, we found an intriguing resemblance of the latitudinal position of the westerlies with changes in cosmic ray flux (Fig. 3b). High fluxes at ∼60 ka and between ∼40 and 25 ka, can be inferred from cosmogenic nuclides produced in the atmosphere, such as 10 Be, 14 C and 36 Cl (Christl et al., 2007;Hughen et al., 2004), and can be related to periods with a weaker geomagnetic field (e.g. the Laschamp event at ∼39 ka). Although admittedly highly speculative, we suggest that there might be causal linkages between past changes in the cosmic ray flux and the position of the southern westerlies. On the one hand, some scientists believe that cosmic-ray-induced ionization in the atmosphere can directly affect cloud formation and the atmospheric circulation (Burns et al., 2008;Usoskin and Kovaltsov, 2008). On the other hand, we point to a potential linkage via cosmic-ray-induced changes in the atmospheric chemistry, in particular ozone concentrations. Son et al. (2008), for example, have shown that the southward shift of the westerlies observed during recent decades can likely be attributed to the anthropogenic destruction of stratospheric ozone and related changes in temperature gradients. Provided that the Laschamp event also caused an ozone hole (Valet and Valladas, 2010;Vogt et al., 2009), one might analogously expect a southward shift of the westerlies at ∼39 ka. This, taken at face value, is the opposite of what one would expect from the apparent correlation between strong cosmic ray fluxes and northward shifts of the westerlies/glacial advances ( Fig. 3a and b). We fully acknowledge that at this point we are unable to resolve this discrepancy, and we are curiously anticipating the results of a more sophisticated climate model that will couple atmospheric and ocean components and include cosmic ray-induced changes in atmospheric chemistry in the absence of anthropogenic chlorofluorocarbons. As such kind of modeling studies are still in their infancy, it is possible that the sign of the forcing might be different (Calisto, M., personal communication, 2010). If not, the uncertainties of the currently available glacial chronologies would also not rule out the possibility that the Laschamp event marks the end of the early LGM in the southern Central Andes rather than the glacial advance itself, and future studies may have to disentangle a complex mixture of forcings, including cosmic rays, Antarctic temperatures and sea ice.

Conclusions
The glacial chronologies in the southern Central Andes provide robust evidence for an early local LGM at ∼39 ka, which documents increased precipitation at that time and a northward shift of the southern westerlies. Our understanding of the forcings for this shift remains incomplete, and the coincidence with the Laschamp event will fuel discussions concerning possible cosmic ray -climate linkages. These  (Christl et al., 2007), (c) sea ice duration around Antarctica (Crosta et al., 2004), (d) austral winter insolation at 60 • S, (e) deuterium from the Vostok ice core as proxy for Antarctic temperatures (Petit et al., 1999), and (f) atmospheric CO 2 concentrations (Ahn and Brook, 2008;Petit et al., 1999). linkages should be further investigated, and future global climate models might have to include cosmic-ray-induced changes in atmospheric chemistry.
Our findings also have major implications regarding the role of the westerlies in the carbon cycle. At first glance, low atmospheric CO 2 concentrations between ∼40 and 20 ka seem to corroborate the proposed link between equatorward westerlies, reduced deep ocean ventilation, and low levels of atmospheric CO 2 (Toggweiler et al., 2006). However, CO 2 levels had already dropped significantly at ∼70 ka, and no dramatic changes occurred at ∼39 ka (Fig. 3f) (Ahn and Brook, 2008). We conclude that latitudinal shifts of the westerlies did probably not exert dominant control on atmospheric CO 2 concentrations on glacial-interglacial timescales, which is in agreement with recent modeling studies (Menviel et al., 2008;Tschumi et al., 2008). Other hypotheses should therefore be pursued, for example increased Southern Ocean stratification (Sigman et al., 2010) and changes in the terrestrial carbon cycle .