Centro de Investigaciones Geoquímicas y Procesos de la Superficie, Facultad de Ciencias Exactas, Físicas y Naturales, Centro de Investigaciones en Ciencias de la Tierra, Consejo Nacional de Investigaciones Científicas y Técnicas, Unversidad Nacional de Córdoba, Córdoba, Argentina
 The origin of dust deposited in East Antarctica is not fully understood yet. This study demonstrates that, contrary to Nd, Sr isotopes are strongly fractionated in the finest-sized material representing potential dust sources in South America. Analysis of Nd isotopes suggests that Argentine loess, Southern Ocean sediments and Antarctic dust, may all have a related origin or can represent comparable mechanisms of eolian sediment transport and deposition. This analysis also concludes that the Patagonian signature can explain a high proportion of the isotopic composition of dust trapped in Antarctic ice during glacial periods. The Puna-Altiplano plateau emerges as the second possible important sediment contributor explaining the crustal-like signature found in East Antarctic dust and should be considered as a potential source area when Quaternary paleoclimate is reconstructed from Argentine loess, Antarctic dust and South Atlantic sediment records.
 Mineral dust has an important role in paleoclimate studies because it is a proxy indicator of aridity in the continents and of changes in the global wind systems. This is one of the reasons why dust records from polar ice cores have been extensively used to deduce past changes in atmospheric circulation and climatic conditions in the source region [e.g., Petit et al., 1999]. In addition, it has served to validate atmospheric global circulation models (AGCM) during the last glacial periods.
 By contrasting the isotopic (Sr and Nd) composition of dust in ice cores with those from sediments representing southern hemisphere (SH) potential source areas (PSAs), it has been established that Patagonia was the main sediment supplier to East Antarctica (EA) during glacial times [Basile et al., 1997]. More recently, Delmonte et al.  reviewed this issue and, by enlarging the PSAs and ice core sampling, concluded that the dust deposited in EA during glacial times derived almost exclusively from southern South America (SSA). These new data were useful to discriminate the main dust source region in the SH but did not differentiate any specific areas of SSA, as the new available information promoted a wider range and overlapping isotopic compositions for dust characterizing SSA [Gaiero et al., 2007]. Nevertheless, the idea of a unique Patagonian origin still persists in some of the most recent research works on Antarctic ice cores [e.g., EPICA Community Members, 2006]. Clearly, the South American dust source issue still needs to be deciphered.
 New systematic isotopic data of dust exported from Patagonia indicate that despite of latitudinal position the region export sediments with a uniform chemical and isotopic signature and could only explain part of the origin of EA dust deposited during glacial periods [Gaiero et al., 2007]. This rises the question whether other PSA in the SH have contributed to the long-range dust transported to Antarctica. A contribution from Australia to EA is suggested for interglacial times [Revel-Rolland et al., 2006] but this is considered negligible for glacial periods, and moreover unsupported by model simulations [e.g., Lunt and Valdes, 2002]. Thus, in this work the focus is placed in SSA where extensive loess records are a clear testimony of much larger dust activity during the last glacial period. Implicit in this contribution is to stress the importance of grain-size on Sr isotopic composition of dust as formerly observed [e.g., Delmonte et al., 2004] and demonstrate, by means of an isotopic mixing model, that it is possible to disentangle other PSAs in SSA that could explain the non-Patagonian signature found in EA dust.
 Most of data set for this contribution was obtained from the bibliography [see Gaiero et al., 2007, and references therein] except new isotopic data of top soils samples from Central Argentina (see location in Figure 1 and data in Table S1 in the auxiliary material). Isotopic determination procedures are fully described elsewhere [e.g., Gaiero et al., 2007].
3. Results and Discussion
3.1. Changes on the Isotopic Composition for Long-Range Transported Dust
 The very fine grain-size materials from the EA ice cores (average ∼2 μm) [e.g., Delmonte et al., 2004], have an isotopic composition range with a clear offset on the 87Sr/86Sr axis compared to coarser sediment samples from different southern environments (Figure 2a). The grain-size of dust becomes finer when it is transported over long distances. This sorting effect modifies the mineralogical composition of the material, hence its isotopic composition especially in the Rb-Sr system, indicting that clay-sized fractions contain higher Rb/Sr ratios and, consequently, higher 87Sr/86Sr ratio than coarser materials. This Sr isotopic fractionation is observed on fine sediments from the Argentine loess (ArL) formation, marine sediment cores from the South Atlantic Ocean and the Scotia Sea (Figure 2a) and in Patagonian sediments (Figure 2b). On the contrary, the Sm-Nd system is not affected by this sorting effect as demonstrated by sediment provenance studies [e.g., Smith et al., 2003, and references therein] and illustrated on Figure 2c for Patagonian sediments. The conservative behavior of ɛNd(0) on different grain-size fractions of Patagonian sediments can be used to select EA samples with a pure Patagonian signature (dotted circle in Figure 2a). Applying this interpretation and according to Figure 2a, ice core dust samples with more radiogenic Sr and unradiogenic Nd values (non-Patagonian isotopic signature) suggests three possible origins: direct transport from the ArL formation and/or from the Patagonian shelf or a mix source of eolian sediments from different SSA dust sources. In the following sections these possibilities are explored.
3.2. Origin of the Non-Patagonian Signature Found in East Antarctic Dust
3.2.1. Role of the Patagonian Shelf
 The role of the Argentine/Patagonian shelf as dust source during cold periods is controversial [Wolff et al., 2006, and references therein]. Some arguments stated below prevent the use of published shelf sediments isotopic data as representative of this area during glacial times. Similar to the main Patagonian rivers bed load and ArL samples, the isotopic composition of shelf sediments have lower ɛNd(0) values compared to Patagonian dust (Figure 2a). About 90% of the present sediment mass delivered from Patagonia to the shelf are eolian [Gaiero et al., 2003] and it is very likely that such scenario was more pronounced during glacial periods due to a general reduction of the hydrological cycle [e.g., Yung et al., 1996]. Then, it is expected that sediments deposited at the exposed continental shelf during glacial times contained a dominant Patagonian dust signature. Moreover, the shelf is composed primarily of sandy sediments [e.g., Guilderson et al., 2000], suggesting that during high stand periods most of the fine particles supplied by the atmospheric pathway were removed from the shelf and deposited on the Argentine Basin [Lonardi and Ewing, 1971]. Therefore, the bulk topmost layer of piston cores previously employed to define a shelf source could actually represent a mix of fluvial sediments deposited during the Holocene and hence, they do not represent the true shelf composition for low stand sea level during cold periods. In summary, during low stands sea level, dust entrained from the Patagonian shelf should contain a similar isotopic composition compared to dust derived from the adjacent continent. In addition, paleoenvironmental studies indicate a small rapid sea level rise in the Argentine shelf around 14 kyr BP [Guilderson et al., 2000], which coincided with a period of almost no changes in dust fluxes to EA thus suggesting a minor role for the transport from this area [Wolff et al., 2006, and references therein].
3.2.2. Role of the Argentine Loess (ArL) Formation
 Isotopic data of ArL samples (open/crossed circles in Figure 1) were employed to evaluate the relevance of this area as dust contributor to EA during glacial periods [Basile et al., 1997; Delmonte et al., 2004]. This region is not considered a present-day dust source in SSA [Prospero et al., 2002] and contrasting with Patagonia, it is not clear whether it was a significant dust source during the last cold periods. This is because the ArL region has geomorphic, climatic and biogeographical characteristics that differ from the Patagonian region and, most likely, provided a different environmental response during glacial times. This is based on the extensive eolian deposits of last glacial age found in the Pampean plains as opposed to no records found in Patagonia except for very localized areas [Zárate, 2003].
 Nevertheless, loess records in many continents are a clear testimony of much larger dust activity during the last glacial periods and the identification of the source areas that contributed to the ArL formation could place strong constrains to the understanding of the environmental conditions of the source areas and the long-range transport of dust into EA. Nd isotopes of the ArL deposits have different composition in a clear N-S gradient (26S–39S) [Smith et al., 2003; this work] (Figure 1b). Sedimentological [e.g. Zárate, 2003] and isotopic [Smith et al., 2003; this work] evidences indicate a main northern Patagonian origin through SSW winds for the southern ArL (Figure 1). Further north, the loess origin is less understood [Zárate, 2003]. Accordingly, Smith et al.  constrained the loess provenance discussion and interpreted the latitudinal shift as a direct W to E transport of sediment from the Central-West Argentina (CWA) (Figure 1) [see Prospero et al., 2002] containing the isotopic gradient of the Andean volcanic rocks. Contrarily to the assertion of these authors, a volcanic gap exist between 27°–34°S (Figure 1) and no available geochemical data for rocks and sediments from this area can be used to check this conclusion. Furthermore, Smith et al.  explain this transport by a northward movement in the Polar Front, which could promote an increase in intensity and frequency of the westerly belt further to the N (e.g., up to 30S). This mechanism is frequently used to explain the increased moisture recorded on the western side of the Andes during the Last Glacial Maximum (LGM) [e.g., Lamy et al., 1999]. However, north of 37S the height of the Andes increases notably and represents a wall for the westerly winds from the Pacific [Seluchi et al., 1998]. Presently, the prevailing wind direction along the CWA is from the S [Smith et al., 2003, and references therein] and there is no geological evidence indicating westerly winds direction of Quaternary age. According to Smith et al. , the Puna-Altiplano plateau (PAP) is a significant source area only for loess north of 30S. However, the importance of the PAP as dust contributor to the loess deposits was suggested [e.g., Bloom, 1990], but was little explored by researchers working in the Argentine loess problem.
3.2.3. Role of the Puna-Altiplano Plateau (PAP)
 The PAP is a high elevated basin (3750–4000 m) that extends over 1000 km and is ∼200 km wide. On all timescales, the climatic condition of this region is closely associated to the upper air circulation determining easterly zonal flow aloft supporting increase moisture conditions (<5% of the time), and mostly westerly winds prevailing in middle and upper troposphere causing dry conditions from May to October [Garreaud et al., 2003].
3.3. Provenance Mixing Model to Explain the SSA Signature in Antarctic Dust
 A mixing isotopic model between dust exported from Patagonia and PAP-like sediments reproduce well the ArL composition (Figure 2a). The northernmost loess sequences from El Lambedero, show a slightly different Sr isotopic composition probably as a consequence of local lithological sources in this inter-montane area. In agreement with a SSA origin, this model also explains the isotopic composition of marine sediment deposited during glacial periods [Walter et al., 2000].
 Data from Figures 2b and 2c allow recalculating a new mixing curve for the finest-grain sediments (see legend). The new isotopic mixing curve supports atmospheric transport models, indicating that most of EA dust deposited during the LGM derives from SSA, where an 80–90% have a Patagonian signature [Lunt and Valdes, 2002]. Notably, the finest-grain marine sediments and loess samples are also explained by this mixing curve. The dominant Patagonian signature can be primarily related to a shorter pathway transport compared to the PAP. Nevertheless, the variable presence of the PAP signature indicates that there could be changes in the transport efficiency from SSA to EA and/or changes in the climatic conditions of the sources. Model simulations and dust particle measurements suggest that, in general, glacial-interglacial changes in the aerosol transport from SSA to EA were small when compared to changes in the source region, the later explaining the high dust input to Antarctica during the LGM [Wolff et al., 2006]. Source changes in Patagonia appear mostly governed by the persistence and strength of westerly winds. During glacial periods, intensification of westerlies resulted in increased precipitation, which in turn led to an expansion of the Patagonian ice field, thus increasing fluvial erosion in the Andes [e.g., Petit et al., 1999]. This seemed to promote a strong rain shadow on the eastern plateau [McCulloch and Davies, 2001], therefore enhancing dust mobilization, also favored by stronger katabatic winds [Clapperton, 1993]. Source changes in the PAP seem to be linked to significant changes in precipitation on all time scale [Garreaud et al., 2003]. Paleoenvironmental records for the last 325 kyr indicate changes between hyperarid, semiarid and wet conditions [e.g., Fritz et al., 2004]. Therefore, the pure Patagonian signature found in EA dust during some stadials of Marine Isotopic Stadial (MIS) 2, 4 and 6 (Figure 2a) could indicate the dominance of wetter (e.g., MIS 2 and 4) and hyperarid conditions (most of MIS 6) [Fritz et al., 2004], that temporarily reduced the PAP source. Unfortunately, the lack of longer SSA continental record and the limited number of EA dust samples representing the oldest glacial periods (MIS 8 and 12) preclude any plausible explanation for the increased PAP signature suggested in Figure 2a.
4. Concluding Remarks
 Isotopic data indicate that Patagonia and the PAP could represent the most important SH provenance areas for glacial dust deposited in EA. Along isotopic data, common environmental characteristics observed in both source areas provide a better understanding of provenance and increased dust load in EA during cold periods. Explosive volcanism activity was widespread in Patagonia and in the PAP and could represent an extra significant supplier of fine particles for subsequent dust emission. Both areas are located in semi-arid environments with numerous enclosed basins [Clapperton, 1993]. Repeated oscillations between dry and humid conditions periodically replenish the lowland areas and permit a constant potential availability of material that is subsequently deflated by the strong and persistent winds presently observed in both areas [see Gassó and Stein, 2007] (see section 3.2.3 and auxiliary materials). These regions are crossed by the zonal climate belts between subtropical and cold temperate region of the SH. Patagonia is strongly influenced by the SH westerlies, while climatic characteristic of the PAP is mostly governed by the subtropical westerly jet stream. In the PAP dust could be directly injected into the tropospheric winds and hence transported long distances. This could be inferred from the observation of modern SH jet streams (Animation 1) (http://squall.sfsu.edu/scripts/shemjet_archloop.html). During the Austral winter the jet stream crosses the PAP and continues its zonal pathway up to 130–145E, where frequently is deflected to EA, joining the circumpolar vortex (see case study in the auxiliary materials). Although the Patagonian plateau is characterized by a lower topography compared to the PAP, usually the transport momentum from jet streams produced intensification of surface wind speed promoting strong westerly events [Labraga, 1994]. Recent evidence indicates that dust up-lift during strong westerlies reached the free troposphere after 48 hours [Gassó and Stein, 2007].
 The possibility of the existence of subtropical dust sources during the LGM was thought to depend largely on an equatorward shift in the Hadley circulation [Chylek et al., 2001]. However, the presence of a high-elevated desert area at these latitudes could also explain dust transport to Polar Regions. Compared to the PAP, the two most important subtropical southern hemisphere dust sources (South Africa and Australia) are dominated by low relief, having lower potential energy and hence much lower rates of sediment production than high relief areas [Pye, 1995]. This, along with data from this contribution, is in agreement with back trajectory analysis of continental transport to EA, indicating more efficiency from SSA than from Australia during the LGM [Lunt and Valdes, 2001].
 This work was supported by Antorchas, CONICET, FONCYT, IAI, and SECyT/UNC. I thank P. Depetris, S. Gassó and E. Piovano for helpful discussion and ms revision. Thanks to J.-L. Probst, F. Brunet and P. Brunet at the LGTM/Toulouse for isotopic measurements. Reviews by B. Delmonte and A. Bory led to improvements in the manuscript.