Luminescence dating of glacial advances at Lago Buenos Aires ( ~ 46 S ) , Patagonia

Understanding the timing of past glacial advances in Patagonia is of global climatic importance because of the insight this can provide into the influence on glacier behaviour of changes in temperature and precipitation related to the Southern Westerlies. In this paper we present new luminescence ages determined using single grains of K-feldspar from proglacial outwash sediments that were deposited by the Patagonian Ice Sheet around Lago Buenos Aires (~46 S), east of the contemporary Northern Patagonian Icefield. The new luminescence ages indicate that major outwash accumulations formed around ~110 ± 20 ka to 140 ± 20 ka and that these correspond to the Moreno I and II moraine ridges, which were previously dated using cosmogenic isotope dating to 150 ± 30 ka. Luminescence dating at Lago Buenos Aires has also identified outwash sediments that were deposited during glacial advances ~30.8 ± 5.7 ka and ~34.0 ± 6.1 ka (MIS 3) that are not recorded in the moraine record. Younger outwash accumulations were then deposited between ~14.7 ± 2.1 and 26.2 ± 1.6 ka which correspond to the Fenix I e V moraine ridges. The combined chronology suggests that glacial advances occurred ~110 ± 20 ka to 150 ± 30 ka (MIS 6), ~30.8 ± 5.7 ka to ~34.0 ± 6.1 ka (MIS 3), and ~14.7 ± 2.1 to 26.2 ± 1.6 ka (MIS 2) at Lago Buenos Aires. Overall luminescence dating using single grains of K-feldspar has excellent potential to contribute towards the ever-increasing geochronological dataset constraining the timings of glacial advances in


Introduction
Providing robust age constraints for the timing of past glacial advances can offer an important contribution to the global climate change debate. Empirical datasets constraining climate change of the past can be used to test numerical models. Terrestrial climates in the mid-latitudes of the Southern Hemisphere are of global climatic importance because of the influence of the precipitationbearing Southern Westerlies between latitudes of~30 and 55 S, with the position of the present-day core of the Southern Westerlies at~50 to 55 S (Lamy et al., 2010). A large Patagonian Ice Sheet is known to have extended along the Andean mountain range from 38 S to 56 S during repeated glacial cycles of the Quaternary period (e.g. Glasser et al., 2008). This ice sheet may have fluctuated in response to potential changes in precipitation supply related to the migration of the Southern Westerlies (e.g. Denton et al., 1999;Hulton et al., 1994). Providing accurate and precise age constraints on past glacial advances of the Patagonian Ice Sheet can be used to test the accuracy of ice-sheet models, and consequently improve our understanding of how the Earth responds to long-term climatic change. However, providing age constraints for glacial advances of the past can be challenging as some geochronological techniques are restricted by material availability and age ranges, but can also be restricted by processes that have occurred since deposition. The most effective approach to produce robust chronologies in glaciated settings is to combine datasets which use multiple dating techniques to provide ages (e.g. radiocarbon, cosmogenic isotope and luminescence dating). Moreover, researchers are now beginning to incorporate these age sequences into Bayesian frameworks to model patterns of glacial advance and retreat (e.g. Chiverrell et al., 2013).
Luminescence dating is a technique that can determine the last time a grain of sand was exposed to sunlight prior to burial and has great potential for directly dating the sediment deposited during a glacial advance. The luminescence dating technique depends on the ability of mineral grains (typically either quartz or K-feldspars) to store energy within the crystalline structure of the grain and release it upon stimulation (e.g. exposure to sunlight). Upon exposure to sunlight the signal used for luminescence dating is reset (or bleached) before the grains are then buried, where they are exposed to natural radiation from the surrounding environment (e.g. Rhodes, 2011). Luminescence dating is most suitable for providing ages in well-bleached settings (e.g. aeolian) where the opportunity for sunlight exposure is greater than in glaciofluvial or fluvial settings. In comparison to aeolian settings, the luminescence signal of mineral grains from glaciofluvial settings is typically partially bleached as there is greater potential for shorter sediment transport pathways and attenuation of light through the turbulent, sediment-laden water columns transporting the grains (Berger and Luternauer, 1987). Glaciofluvial sediments are typically targeted in glaciated settings to maximise the opportunity for exposure to sunlight, which is greater than expected for sediments deposited as a moraine (e.g. Fuchs and Owen, 2008;Thrasher et al., 2009a). Luminescence dating of single grains can then be used to assess whether heterogeneous bleaching of the luminescence signal from individual grains has occurred and determine accurate ages by applying relevant statistical age models (Duller, 2008;Galbraith and Roberts, 2012).
A major challenge for single-grain dating of quartz from glaciated environments is that commonly only 5% or fewer of the grains emit a detectable optically stimulated luminescence (OSL) signal, and in glaciofluvial sediments from Chile as few as 0.5% of quartz grains could be detected (Duller, 2006). In contrast, a larger proportion of K-feldspar grains emit a detectable infra-red stimulated luminescence (IRSL) signal, which is typically brighter than the OSL signal emitted from quartz grains (e.g. Duller et al., 2003). Singlegrain dating using K-feldspar therefore has the potential to make luminescence dating more efficient and potentially more precise in environments that are characterised by short sediment transport pathways, and also in regions where the sensitivity of the quartz is poor e.g. glaciofluvial sediments (Duller, 2006).
Currently, OSL dating of quartz is more commonly applied for routine dating of single grains as the IRSL signal of K-feldspars is prone to the effects of anomalous fading (Wintle, 1973). Anomalous fading is the athermal depletion of the luminescence signal over time and if fading is not accurately measured and corrected for, it will manifest as an underestimation of the burial age. There are two signals commonly used for IRSL dating of K-feldspars; (1) the IRSL signal typically measured at 50 C, termed the IR 50 signal (e.g. Wallinga et al., 2000), and (2) the post-IR IRSL signal typically measured at 225 C or 290 C, termed the pIRIR 225 and pIRIR 290 signals (Thomsen et al., 2008(Thomsen et al., , 2011. Anomalous fading of the IR 50 signal is suggested by some workers to be ubiquitous to all feldspars (e.g. Huntley and Lamothe, 2001), whereas the pIRIR signal is suggested to access a more stable signal through the recombination of more distal donor-acceptor electron pairs (Jain and Ankjaergaard, 2011). Accessing a more stable IRSL signal within the K-feldspar grains for dating is advantageous as it reduces the influence of anomalous fading on the age calculation, which is especially important on a single grain level, where accurate and precise measurement and correction for fading can be difficult. Few studies have determined ages for sedimentary samples from the natural environment using the pIRIR signal of single grains of K-feldspar (e.g. Gaar et al., 2014;Trauerstein et al., 2014). However, studies have demonstrated that the pIRIR signal can provide ages in direct agreement with independent numerical age control for wellbleached samples (e.g. Reimann et al., 2012;Smedley, 2014).
The aim of this paper is to test the new luminescence dating technique using single grains of K-feldspar from proglacial sediments and complement the pre-existing chronology provided by cosmogenic isotope (Kaplan et al., 2005 and 40 Ar/ 39 Ar  dating at Lago Buenos Aires. The combined chronology will strengthen the age constraints provided for the glacial advances as it can provide checks on the internal consistency between techniques and also provide ages for different landforms preserved after deglaciation.

Study site and sample descriptions
In the past a large outlet glacier expanded eastwards at Lago Buenos Aires from the Andean mountain range on to the semi-arid Argentine plateau (Caldenius, 1932). Multiple sets of moraine ridges are preserved in the valley, extending from the Last Glacial Maximum (LGM) to~1 million years ago (Ma). The moraine complexes are locally named the Menucos (I), Fenix (I e V), Moreno (I e III), Deseado (I e III) and Telken (I e VII) moraines, ordered from youngest to oldest. Age constraints for the glacial advances are currently provided by cosmogenic isotope dating of moraine boulders (Kaplan et al., 2004(Kaplan et al., , 2005Douglass et al., 2006) and 40 Ar/ 39 Ar dating of the basaltic lava flows . The timing of the LGM at Lago Buenos Aires is recorded by the deposition of the Fenix moraine ridges and is suggested to have been broadly synchronous with sites across Patagonia from~16 to 30 ka (Kaplan et al., 2004Douglass et al., 2006), and with the responses of global ice sheets (e.g. Mix et al., 2001;Glasser et al., 2011). The Moreno and Deseado moraines are suggested to predate the LGM at Lago Buenos Aires (Kaplan et al., 2005).
Well preserved proglacial outwash plains (or sandur) at Lago Buenos Aires were targeted for luminescence dating. The lowangled slope of each outwash plain can be traced eastwards from the former ice front marked by the prominent moraine ridge. These proximal outwash plains can only have been active when the ice front was positioned at the elevation marked by the corresponding moraine ridge. Assessing the geomorphological and sedimentological context of each luminescence sample enables a direct comparison to be made between the luminescence and cosmogenic isotope ages. The geomorphology at Lago Buenos Aires was mapped in this study using Landsat imagery in combination with Google Earth imagery and ground truthing in Patagonia (Fig. 1). Differential Global Positioning System (dGPS) measurements of altitude, latitude and longitude were also used to develop an elevation profile at Lago Buenos Aires and a conceptual model describing the possible development of the sedimentary sequence for the Menucos and Fenix moraines (Fig. 2). The relationship between the moraine ridges and outwash plains that were sampled is shown by the elevation profiling and conceptual model (Fig. 2). The re-activation of pre-existing glaciofluvial channels and the presence of glaciofluvial channels that were orientated NW-SE, draining laterally across the valley and in front of the moraine ridges ( Fig. 1) complicated these relationships for some samples; this will be discussed later in Section 2.1. Cosmogenic isotope dating of the moraine ridges suggest that each moraine and related outwash plain was deposited during different glacial advances or still-stands (Kaplan et al., 2004. Luminescence dating was performed on 13 sedimentary samples taken at Lago Buenos Aires. The locations of the luminescence samples associated with the Fenix (n ¼ 9 samples), Moreno (n ¼ 3 samples) and Deseado (n ¼ 1 sample) moraine complexes are shown with the location of the existing chronology (Fig. 1). The elevation profile (Fig. 2) shows the difference in elevation (~50 m) of the Fenix and Moreno moraine complexes, and the Moreno and Deseado moraine complexes (also~50 m). Table 1 presents the details of the samples taken for luminescence dating and each sample site is shown in Fig. 3, where the moraine fragments and outwash plains related to each of the luminescence samples are highlighted in relation to Fig. 2.

Fenix moraines
The geomorphological and sedimentological contexts of four of the nine samples related to the Fenix moraines were typical of a proglacial outwash plain deposited in front of a moraine ridge  Fig. 3e). The sediments were either poorly sorted (samples LBA12OW4 or LBA12OW2), which is typical of higher energy glaciofluvial depositional conditions, or well sorted (LBA12OW3 and LBA12OW1), which is typical of stagnant water  conditions. The outwash sediments of all four of these samples were part of proximal outwash plains with low-angled slopes that can be traced westwards to the moraine ridge where the ice front was positioned when the outwash plain was active. The outwash plain of sample LBA12OW3 (410 m elevation) was eroded by the lake shorelines and lacustrine sediments that underlie the Menucos moraine ridge (Fig. 3a). The outwash plain was also constrained to the east by the Fenix I moraine ridge which has been eroded by a glaciofluvial channel at an elevation of 410 m. The age of the outwash plain of sample LBA12OW3 must therefore be older than the Menucos moraine and younger than the Fenix I moraine ( Table 1).
The outwash plain of sample LBA12OW4 (440 m) was elevated 30 m higher than the outwash plain of sample LBA12OW3 (410 m), and was likely active during an earlier phase of glaciation. The sediments dated for sample LBA12OW4 were collected from~2 km in front of the Fenix II moraine, where the ice front was positioned at an elevation of~470 m when the sediment was deposited (Fig. 3b). The outwash plain of sample LBA12OW4 was therefore active when the ice front was positioned~2 km to the west at the Fenix II moraine (Table 1). Sample LBA12OW1 was taken from a small section of an outwash plain~3 km in length and~0.2 km wide that was constrained to the east by the Fenix IV moraine ridge and truncated to the west by a younger, lower (by~10 m) glaciofluvial channel draining from the north (Fig. 3d). The low-angled slope of the outwash plain can be traced at an elevation of 450 m for~8 km from the Fenix II moraine ridge to the location of sample LBA12OW1. Similar to the outwash plain of sample LBA12OW4 elevated at 440 m, the outwash plain of sample LBA12OW1 was also likely to have been active when the ice front was positioned at the Fenix II moraine ridge (470 m elevation).
To the east of the Fenix II moraine are the poorly preserved fragments of the Fenix III moraine and proximal outwash plains (Fig. 3e). The landform targeted for sample LBA12OW2 was an outwash plain that extended for <1 km eastwards from the Fenix III moraine as a low-angled slope at an elevation of 500 m (Fig. 3e). Sample LBA12OW2 was taken from~0.5 m in front of the moraine ridge within this outwash plain. The close proximity and elevation of the outwash plain of sample LBA12OW2 suggests that this was active when the ice front was positioned at the Fenix III moraine ( Table 1).
The geomorphological contexts of the other sample sites related to the Fenix moraines were more complex. The sedimentary units that samples LBA12OW5, LBA12RF1, LBA12F4-1 and LBA12F4-2 were taken from were all poorly sorted, but it was difficult to directly link these samples to specific moraine ridges using the geomorphology. Based on the elevation profile of the outwash plains, samples LBA12OW5 and LBA12RF1 are interpreted to have been deposited at a similar time or before the Fenix II moraine ridge. However, the conceptual model ( Fig. 2) suggests that it is possible for sediments from younger glacial advances to have been deposited on top of sediments from older glacial advances. Glaciofluvial sediments preserved at different elevations at a single site may therefore have recorded multiple glacial advances within the same stratigraphy, whereas the more prominent moraine ridges relating to older glaciations were likely destroyed by overriding ice. Therefore, it is possible that the proglacial sediments sampled for luminescence dating at lower elevations within the sedimentary sequence may be older than the moraine ridges themselves.
Samples LBA12F4-2 and LBA12F4-3 were extracted from the same stratigraphic section (Fig. 4a),~100 m west of sample LBA12F4-1 (Fig. 4b). The context of these three samples is complicated due to their proximity to the central glaciofluvial channel in the valley, which was re-activated during younger glaciations. Also, post-or syn-depositional processes have deformed some of the sediments in these sections and complicated the stratigraphy (Fig. 4a). The stratigraphy of the site is outwash gravel units deformed by a silty-sand unit that contained rounded pebbles, which is interpreted to be lacustrine sediments deposited in a proglacial lake setting. After deposition, the preserved lake sediments were thrust up into the overlying glaciofluvial sediment to an angle of~40 (Fig. 4a). The sedimentary units were then covered by a laterally-extensive sedimentary sequence of horizontally-layered glaciofluvial sediments (up to~2 m thick), diamicton (up to~2 m thick), and~0.5 m thick unit of glaciofluvial sediments. The glaciofluvial sediments sampled for LBA12F4-1 were not deformed by post-or syn-depositional processes and were sampled at an elevation of~434e438 m (Fig. 4b), which was similar to the upper sedimentary units (~436e440 m elevation) of the section sampled for LBA12F4-2 (Fig. 4a). The stratigraphy at these two sites suggests that multiple glacial advances have been preserved, which is consistent with the hypothetical conceptual model for the deposition of the Fenix moraines (Fig. 2). The glaciofluvial sediments deposited in the lower part of the section that have been deformed post-or syn-depositionally (sampled by LBA12F4-2) potentially relate to an older glacial advance, and the glaciofluvial sediments in the upper part of the sequence (sampled by LBA12F4-1) relate to a younger glacial advance, which is potentially related to the Fenix IV moraine ridge in this area (Fig. 3f). The glaciolacustrine sediments that sample LBA12F4-3 was taken from are stratigraphically older than both samples LBA12F4-1 and LBA12F4-2.

Moreno moraines
The three samples taken from glaciofluvial and glaciolacustrine sediments related to the Moreno moraines were from elevations >500 m ( Fig. 3g; LBA12M1-1, LBA12M1-2 and LBA12M3). Samples LBA12M1-1 and LBA12M1-2 were taken from the same stratigraphic section, which was associated with the Moreno I moraine (Fig. 4c). Sample LBA12M1-1 was taken from glaciolacustrine sediments that were post-depositionally deformed and then covered by the glaciofluvial sediments that sample LBA12M1-2 was taken from, and so sample LBA12M1-1 therefore must have been deposited before sample LBA12M1-2. Sample LBA12M3 was taken from a well-sorted, coarse-grained sand unit deposited within a matrix-supported diamicton of the Moreno III moraine ridge (Fig. 4d). The sediments are interpreted to have been deposited in a ponded, slack water feature that formed as part of the Moreno III moraine ridge (Fig. 3g). The elevation profile (Fig. 2) suggests that the Moreno III moraine ridge and associated outwash plains sampled by LBA12M3 (530 m) were covered by the Cerro Volc an lava flow and should therefore be older than the 40 Ar/ 39 Ar age of 109 ± 3 ka.

Re-calculation of existing cosmogenic isotope ages
Cosmogenic isotope ages for the Moreno moraines calculated using earlier estimates of production rates suggest that these glacial advances occurred during Marine Isotope Stage (MIS) 6 (Kaplan et al., 2005). The 10 Be ages published in Kaplan et al. (2005) have been re-calculated in this study using the recently obtained production rates for mid-latitudes of the Southern Hemisphere, updates to scaling factors for location and altitude and other systemic improvements in the technique (see Balco et al., 2008;Kaplan et al., 2011). The production rate determined from south of Lago Argentino  is statistically identical to the production rate determined from the Southern Alps in New Zealand (Putnam et al., 2010). The re-calculated cosmogenic isotope ages suggest that the Moreno I and II moraine ridges are from MIS 6; the distribution of ages ranges from 108 ± 6 ka to 209 ± 11 ka, with Fig. 4. Stratigraphical sections showing the outwash sediments sampled for LBA12F4-2 and LBA12F4-3 (a), LBA12F4-1 (b), LBA12M1-1 and LBA12M1-2 (c) and LBA12M3 (d). The stratigraphy shown in (a) includes a unit of lacustrine sediments (sampled by LBA12F4-3) thrust up into overlying outwash sediments (sampled by LBA12F4-2), which have been post-or syn-depositionally deformed. The section shown in (b) was~100 m east of the section shown in (a), but the outwash sediments not deformed by post-or syn-depositional processes. Elevation measurements (HAE e Height Above Ellipsoid) for each section are from the dGPS measurements.
weighted mean and standard deviations of 148 ± 33 ka and 154 ± 35 ka, respectively (see Table S.1 for details).

Measurement of dose-rates for luminescence dating
Dose-rates are determined in luminescence dating to quantify the amount of natural radiation the mineral grains were exposed to per year throughout burial. The external dose-rates were determined in this study using thick source alpha and beta counting on Daybreak and Risø GM-25-5 measurement systems, respectively. The conversion factors provided by Guerin et al. (2011) were used, in addition to an alpha efficiency value (a-value) of 0.11 ± 0.03 after the measurements performed by Balescu and Lamothe (1993) ( Table 2). The water contents for all samples (5 ± 2 %) are expressed as a percentage of the mass of dry sediment and informed by field measurements. Cosmic dose-rates were calculated in accordance with Prescott and Hutton (1994).
Feldspars form a solid-solution series, ranging from anorthite (CaAl 2 Si 2 O 8 ), to albite (NaAlSi 3 O 8 ), to orthoclase (KAlSi 3 O 8 ). Density separation is routinely used for luminescence dating to isolate the K-feldspar dominated fraction but this process can be difficult for some samples. Grain-to-grain variation in internal K-content is therefore an important factor to consider for single-grain dating of K-feldspars as the internal beta dose-rate originating from the K within the different grains may vary. Bulk K-contents of the density-separated K-feldspar fractions were measured for all samples in this study using 0.1 g sub-samples on a Risø GM-25-5 beta counter; the mean and standard deviation K-content measured for all the samples was 5.0 ± 1.0% K, and suggested that the geochemistry of grains in these samples was variable. LA-ICP-MS measurements performed on single grains of K-feldspar from samples LBA12OW1 and LBA12OW4 were then used to confirm that the internal K-contents of individual grains emitting detectable luminescence signals in these samples varied from 0 to 14% ( Fig. S.1). The bedrock geology underlying the Northern Patagonian icefield includes both plutonic and basaltic rocks (Espinoza et al., 2010). The variability in internal K-contents measured for individual grains from this suite of samples is consistent with the mixture of perthitic and end member (i.e. K-or Na-rich) feldspar grains likely to have been sourced from the bedrock at Lago Buenos Aires.
Previous studies have shown a relationship between the singlegrain D e values and luminescence signal intensity measured in response to a fixed test dose for some well-bleached sedimentary samples, which contained feldspar grains of variable K-content (Reimann et al., 2012;Smedley, 2014). In these studies single-grain ages were calculated using only the D e values for the grains emitting the brightest pIRIR 180 (Reimann et al., 2012) and pIRIR 225 signals (Smedley, 2014), and assumed internal K-contents of 12.5 ± 0.5% and 10 ± 2%, respectively. The ages determined were in good agreement with independent numerical age control and suggested that the range in internal K-contents of the K-feldspar grains was reduced by selecting the brightest grains. The LA-ICP-MS measurements performed on single grains of the density-separated K-feldspar fraction from samples in this study suggest that a similar approach is necessary to determine accurate ages (Fig. S.1). Therefore, only the D e values determined for the brightest 20% of the grains for each sample were used to calculate ages, and an internal K-content of 10 ± 2% was assumed for dosimetry calculations (for more information see the supplementary materials provided in S.2).

Equipment and measurement details
Samples were collected in either opaque tubes (2 mm thick plastic rim) that were hammered into the sedimentary section, or under an opaque cover sheet to prevent contact with direct sunlight during sampling. All samples were prepared under subdued red lighting conditions in the laboratory. The sedimentary samples were first treated with a 10% v.v. dilution of 37% HCl and with 20 v.v. of H 2 O 2 to remove carbonates and organics, respectively. Dry sieving isolated the 180e210 mm or 180e250 mm diameter grains, and density separation using sodium polytungstate provided the <2.58 g cm À3 (K-feldspar-dominated) fractions. The K-feldspar grains were not etched in hydrofluoric acid because of concerns about anisotropic removal of the surface (Duller, 1992). Grains were mounted into 10 Â 10 grids of 300 mm diameter holes in a 9.8 mm diameter aluminium single-grain disc for analysis.
All luminescence measurements were performed using a Risø TL/OSL DA-15 automated single-grain system equipped with a 90 Sr/ 90 Y beta source delivering~0.04 Gy/s. The variability of the beta dose between grains for the source used in this study was found to be negligible with a relative standard error of~2%. Each single-grain disc was located using the laser at room temperature to prevent annealing of the pIRIR signal during the time it takes to find the disc (Smedley and Duller, 2013). Stimulation was performed using an infra-red laser (830 nm) fitted with an RG-780 filter to remove any shorter wavelengths , and a blue detection filter pack (BG-39, GG-400 and Table 2 Dosimetry calculations for luminescence samples from Lago Buenos Aires. An internal K-content of 10 ± 2% was applied to determine the internal beta dose-rate for these samples (Smedley et al., 2012).

Sample
Depth ( Corning 7-59) was placed in front of the photomultiplier tube. The signal was recorded for a total of 1.7 s, and the IRSL signal was summed over the first 0.3 s (10 channels) of stimulation and the background calculated from the final 0.6 s (20 channels). The instrument reproducibility of the single-grain measurement system was measured using the pIRIR signal (2.5%; Smedley and Duller, 2013) and incorporated into the calculation of the D e values. The single aliquot regenerative dose (SAR) protocol using the pIRIR 225 signal in this study is shown in Table 3. Five screening criteria were applied to the resulting data throughout the analyses unless otherwise specified. Grains were only accepted if (1) the response to the test dose was greater than three sigma above the background, (2) the test dose uncertainty was less than 10%, (3) the recycling ratio (including the associated uncertainties) was within the range of ratios 0.9 to 1.1, (4) recuperation was less than 5% of the response from the largest regenerative dose (96 Gy) and (5) the single-grain D e values were not part of a population of very low doses that were identified by the finite mixture model (FMM) to be inconsistent with the geological context of the sample (e.g. Smedley, 2014). A luminescence age was then determined by dividing the D e value by the environmental dose-rate that the grains were exposed to throughout burial.

5.2.
Assessing the pIRIR 225 signal for single-grain dating

Optical bleaching
For luminescence dating in an environment where the opportunity for bleaching is limited (e.g. glacial settings) it is important to consider how rapidly the luminescence signal used for dating bleaches in response to optical stimulation (e.g. sunlight). Although the pIRIR signal may be more stable over time than the IR 50 signal, several multiple-grain studies of coarse-grained K-feldspar have demonstrated that the pIRIR signal bleaches more slowly in response to optical stimulation than the IR 50 signal Murray et al., 2012), which in turn bleaches more slowly than the OSL signal from quartz (Godfrey-Smith et al., 1988). If the signal used for luminescence dating bleaches more slowly in response to sunlight in nature, it is more likely that larger residual doses are incorporated into the D e values determined in the laboratory. This has led some authors to prefer the IR 50 signal for single-grain dating of K-feldspar in environments where there is limited exposure to sunlight instead of the pIRIR signal Blombin et al., 2012). However, Smedley et al. (2015) have demonstrated that the pIRIR 225 signal bleaches more rapidly than the pIRIR 290 signal for individual grains of K-feldspar. The authors also performed short laboratory bleaching experiments on two well-bleached samples and one partially-bleached sample with independent age control to demonstrate that the bleaching rates of the pIRIR 225 signal for individual grains of Kfeldspar did not have a dominant control on the single-grain D e distributions. Therefore, the pIRIR 225 signal was preferred for single-grain dating of K-feldspar in this study.

Recovering a laboratory dose
Dose-recovery and residual-dose experiments were performed on sample LBA12F4-2 to assess the suitability of the measurement protocol using the pIRIR 225 signal. The residual D e value measured after a given dose of 50 Gy and an 8 h bleach in the solar simulator (SOL2) was 2.7 ± 0.4 Gy (n ¼ 25 grains). The residual-subtracted dose-recovery ratio determined was 1.12 ± 0.07 and demonstrated that the measurement conditions were appropriate for dating these samples. The overdispersion value calculated for doserecovery experiments of sample LBA12F4-2 was 21 ± 1% and quantifies the minimum amount of scatter beyond measurement uncertainties that can be incorporated into a single-grain D e distribution from measurements performed in the laboratory.

Anomalous fading
Fading measurements are typically performed for luminescence dating of K-feldspar grains to determine whether fading correction of the luminescence signal is necessary to provide accurate ages. Measuring accurate fading rates (termed g-values) from single grains of K-feldspar is difficult due to the low sensitivity of the IRSL signals from some of the grains. Therefore, fading measurements were performed in this study on a large suite of medium aliquots to determine the best estimate of the fading rate for this suite of samples. Five multiple-grain aliquots from 12 of the 13 samples (n ¼ 60 aliquots) were used to determine g-values. Fading rates were determined for both the pIRIR 225 signal and the less stable IR 60 signal built into the pIRIR 225 protocol (termed the IR 60/225 signal), and were normalised to two days (Auclair et al., 2003). The g-values measured using the IR 60/225 and pIRIR 225 signals formed a normal distribution around a central value (Fig. 5a) so the weighted means and standard errors were calculated for each of the two datasets to provide the best estimate of fading. The g-values determined were 4.1 ± 0.1%/decade (IR 60/225 signal) and 1.0 ± 0.1%/ decade (pIRIR 225 signal). The fading rate determined using the pIRIR 225 signal for this suite of samples (1.0 ± 0.1%/decade) was consistent with fading rates typically measured using the pIRIR 225 signal of other samples (e.g. Roberts, 2012;Trauerstein et al., 2014), and also similar to g-values measured for the OSL signal of quartz (e.g. Buylaert et al., 2012;Thiel et al., 2011). Given that luminescence dating of quartz is not routinely corrected for fading, the measurement of g-values for K-feldspar that are consistent with quartz fading rates implies that this magnitude of effect is negligible, and may just be an artefact of the measurement procedure (e.g. Buylaert et al., 2012;Thiel et al., 2011). Therefore, fading correction of the pIRIR 225 signal was not performed in this study.

Selection of statistical age models
Overdispersion values calculated for single-grain D e distributions provide an estimate of the amount of scatter in a dataset beyond that of measurement uncertainties. This information can often be used in conjunction with the knowledge of the geological context of a sample to determine which statistical model is appropriate for providing an age (Galbraith and Roberts, 2012). The range of overdispersion values from 26 ± 1% (sample LBA12OW1) to 80 ± 1% (sample LBA12F4-1) for the single-grain D e values of samples in this study suggests that the D e distributions for some of the samples were more scattered than others (Table 1). Given that the samples had similar geochemical and depositional contexts it is likely that the extent of bleaching upon deposition in this glaciated  (Galbraith and Roberts, 2012). However, it is important to consider that other factors, such as luminescence characteristics, microdosimetry and post-depositional processes can cause scatter in D e distributions. Arnold and Roberts (2009) summarise the overdispersion values for single-grain D e distributions determined for quartz from a range of samples thought to have been well bleached upon deposition. The overdispersion values ranged from 0 to 40%, with a mean and standard deviation of 20 ± 9%. For the samples in this study it is hypothesised that a threshold of 50% can be used to differentiate between well-bleached and partially-bleached samples from Lago Buenos Aires. To test this hypothesis the luminescence ages determined for four samples that could be directly linked to specific moraine ridges by the geomorphology were compared to the cosmogenic isotope ages. Two of the samples had overdispersion values 50% (samples LBA12OW3 and LBA12OW1), and two had values >50% (samples LBA12OW4 and LBA12OW2). Examples of the sedimentology and single-grain D e values for one of the samples with an overdispersion value of 50% (LBA12OW1; Fig. 6a, c) and >50% (LBA12OW2; Fig. 6b, d) are shown in Fig. 6.
The CAM ages calculated for the four samples are compared to the cosmogenic isotope ages of the moraines in Fig. 6e to assess the validity of the suggested overdispersion threshold used for age model selection. The two samples with overdispersion values of 50% had CAM ages in agreement with the cosmogenic isotope ages, and the CAM ages for those samples with overdispersion values of >50% overestimated the cosmogenic isotope ages. The agreement between the CAM ages and cosmogenic isotope ages for the samples with overdispersion values 50% confirms that these samples were likely to have been well bleached upon deposition. To calculate accurate ages with the MAM, the amount of scatter characteristic of a well-bleached sample from the natural environment under investigation (s b ) needs to be quantified. Therefore, the mean of the overdispersion values determined from the two samples determined to have been well bleached upon deposition (~30%) was used to define the s b value for the MAM in this study; this overdispersion value was similar to the 20 ± 9% determined for single grains of quartz (Arnold and Roberts, 2009). The MAM was then used to calculate ages for the samples with overdispersion values >50% and the ages calculated agreed with the cosmogenic isotope ages (Fig. 6f). Based on the nature of the single-grain D e distributions and the depositional contexts of the samples, an overdispersion threshold of 50% was used for age model selection for all the samples in this study.
The sedimentology of sample LBA12OW1 (overdispersion 50%; Fig. 6a) was well sorted and characteristic of less energetic depositional environments with greater opportunity for bleaching upon deposition than a more energetic depositional environment. In contrast, the sediment of sample LBA12OW2 (overdispersion >50%; Fig. 6b) was poorly-sorted and characteristic of higher-energy depositional environments, typical of more turbulent water columns, where there is less opportunity for bleaching upon deposition. These observations are consistent with research assessing the relationship between sedimentary lithofacies and the potential for bleaching in a glacial environment (Fuchs and Owen, 2008;Thrasher et al., 2009a,b).  5. Histograms of the fading rates measured for the samples using the IR 60/225 (a) and pIRIR 225 (b) signals. A Gaussian fit has been applied to these data (solid line). Repeated L x /T x measurements were performed using the IR 60/225 and pIRIR 225 signals after progressively longer storage times (inserted between steps 2 and 3 of Table 3), from a prompt measurement after~0.2 h to the maximum delay time of~680 h i.e.~1 month. Note that fading measurements were not performed on samples LBA12M1-1 and LBA12M1-2. The distribution of the signal-intensity emitted from single-grain Kfeldspars for the suite of glacial samples extracted from Lago Buenos Aires are also shown (c). Data are plotted as the proportion of the total light sum that originates from the specified percentage of the brightest grains.

Results
The total number of grains analysed to determine D e values for each sample ranged from 600 to 1900 (see Table S.2 for details). Fig. 5c presents cumulative light sum plots (after Duller, 2000) for all the samples in this study. The number of grains passing the screening criteria and providing D e values ranged from 6.4 to 30.0% of the total grains analysed, and the brightest 20% of the grains for each sample accounted for between only 1.3% and 6.0% of the total grains analysed (Table 1; Fig. S.4). The single-grain ages determined for all of the samples in this study are shown in Fig. 7 in relation to the cosmogenic isotope ages provided for the moraine ridges  and the 40 Ar/ 39 Ar age provided for the Cerro Volc an lava flow . The luminescence age of each sample is also shown in Figs. 3 and 4. The Fenix and Moreno samples can be clearly differentiated by the luminescence ages into different glacial periods, which is consistent with the pre-existing chronology.

Luminescence ages for the Fenix moraines
The luminescence ages for the glaciofluvial sediments associated with the Menucos and Fenix moraines are in chronological order in accordance with the sedimentological and geomorphological context of each sample, e.g. sample LBA12OW3 (14.7 ± 2.1 ka) is the youngest and sample LBA12F4-2 (30.8 ± 5.7 ka) is the oldest. The luminescence ages are also in good agreement with the pre-existing chronology provided for the Fenix moraines. For those samples where the geomorphological and sedimentological context was difficult to interpret (samples LBA12OW5, LBA12RF1, LBA12F4-1 and LBA12F4-2), the use of the geomorphology, elevation profile and conceptual model suggest that the luminescence ages were feasible in the context of the surrounding chronological dataset.
The luminescence age of LBA12OW5 (20.2 ± 4.5 ka) suggests that this sample was deposited at a similar time to the Fenix I and Fenix II moraine ridges. This is feasible because sample LBA12OW5 was taken from a relic glaciofluvial terrace (elevated~415 m) of the central meltwater channel draining through the valley (Fig. 3b), which was truncated by younger meltwater drainage events since formation. Also, the luminescence age provided for sample LBA12F4-1 suggests that these glaciofluvial sediments were deposited at a similar time to sample LBA12OW2 and the Fenix III moraine ridge, but within uncertainties of the cosmogenic isotope ages for the Fenix IV and V moraines. This is feasible as the sample was taken from the upper part of the sedimentary sequence at this site. The luminescence age for sample LBA12F4-2 is older than sample LBA12F4-1 and is therefore internally consistent with the sedimentology of these sections. The age of sample LBA12F4-2 (30.8 ± 5.7 ka) is, however, at the upper limit of agreement with the cosmogenic ages for the Fenix IV and V moraines. The age of sample LBA12F4-2, in addition to the age of sample LBA12RF1 (34.0 ± 6.1 ka), suggests that glaciofluvial sediments have been preserved at Lago Buenos Aires from an earlier glacial advance (or advances) that has later been overrun by the glacial advances of the Fenix IV and V moraine ridges. The conceptual model (Fig. 2) demonstrates that preservation of outwash sediments from an earlier glaciation is feasible as these sediments were taken from units that were either deposited at lower elevations in the sedimentary sequence or have been thrust up from lower elevations by post-or syn-depositional processes.

Luminescence ages for the Moreno and Deseado moraines
The fundamental principles of luminescence dating require that there is an upper age limit to the dating technique; this limit is dependent upon factors inherent to the sample, mineral type and signal used for analysis. The accumulation of energy from ionizing radiation within the crystalline structure of a mineral grain can reach saturation at larger doses and therefore can complicate luminescence dating of older samples. In luminescence dating saturation can be determined from the dose-response curve, and a prudent limit above which D e estimation may be unreliable has been suggested to be two times D 0 (the characteristic dose in an exponential fit) by Wintle and Murray (2006). Duller (2012) has demonstrated that single grains of quartz have different saturation limits; this variability may impact upon the single-grain D e distributions for older samples, and so it is important that the saturation limits of the K-feldspar grains used for dating older samples in this study are investigated.
Single-grain D 0 values were calculated for the proglacial samples associated with the Moreno and Deseado moraines to investigate the grain-to-grain variability in D 0 values of K-feldspar grains. Means and standard deviations were calculated for the 2D 0 values determined for the grains used to determine D e values (i.e. the brightest 20% of the grains); the results ranged from 489 ± 161 Gy to 651 ± 252 Gy. There was a large amount of variability in the 2D 0 values measured for individual grains. For example, the 2D 0 values determined for single grains from sample LBA12D2-1 ranged from 182 to 1423 Gy. Based on the underlying 40 Ar/ 39 Ar age of the Cerro Volc an lava flow, the Deseado II moraine and the associated glaciofluvial sediment (sample LBA12D2-1) must be >109 ± 3 ka. Therefore, the expected D e value for sample LBA12D2-1 is >390 Gy, according to the dose-rate of 3.56 ± 0.42 Gy/ ka determined for this sample (Table 2). This implies that some of the grains analysed were approaching saturation for sample LBA12D2-1 as the grains were characterised by 2D 0 values as low as 182 Gy, but a large proportion of grains had 2D 0 values in excess of the expected D e value. In practise the luminescence signals of grains that were approaching saturation are unlikely to influence the D e distribution as the L n /T n ratios cannot be interpolated on to a dose-response curve (Duller, 2012). Moreover, large and asymmetric uncertainties are determined for grains at the limit of saturation, which means the statistical age models will weight the age towards those D e values more precisely known i.e. those not approaching saturation. Thus, luminescence dating of samples associated with the Moreno and Deseado moraines is not restricted by saturation of the pIRIR 225 signal.
The overdispersion values for the suite of samples related to the Fenix moraines ranged from 26 to 80%, whereas the overdispersion values for the Moreno and Deseado samples ranged from only 38 to 52%. The smaller range in overdispersion values potentially suggests that the proportion of scatter in these D e distributions caused by partial bleaching upon deposition (which will be similar to that experienced by outwash associated with the Fenix moraines) accounts for a smaller proportion of the overall doses. Residual signals caused by the effects of partial bleaching have previously been reported to account for a larger proportion of the burial dose for young fluvial sediments in comparison to older samples (Jain et al., 2004). The reduced influence of partial bleaching on the samples associated with the Moreno and Deseado moraines suggests that the CAM can be applied to determine accurate ages for these sediments; this is supported by the excellent agreement between luminescence ages determined for the Moreno and Deseado moraines, in addition to the cosmogenic isotope ages.
The two luminescence ages determined for the Moreno I moraine ridge (LBA12M1-1 e 132 ± 19 ka; LBA12M1-2 e 134 ± 22 ka) were within ±1 s of one another and the cosmogenic Fig. 7. Comparison of the luminescence ages provided in this study with the cosmogenic isotope ages (Kaplan et al., 2004;Douglass et al., 2006) that were re-calculated by Kaplan et al. (2011). The ages are shown as weighted means and standard deviations, which do not include those ages identified as outliers in the original publication. Also shown is an 40 Ar/ 39 Ar age  for the Cerro Volc an lava flow constraining the deposition of the moraine ridges at Lago Buenos Aires. isotope age determined for the Moreno I moraine ridge (~150 ± 35 ka). The luminescence age determined for the ponded feature incorporated into the Moreno III moraine ridge (111 ± 17 ka) was within uncertainties of the luminescence and cosmogenic isotope ages determined for the Moreno I and II moraine ridges. However, the age was at a lower limit of the age interval provided for the Moreno I and II moraines when considering the uncertainties, and may suggest that the grains within the ponded feature were deposited after the glacial advance. Currently, there are no cosmogenic isotope ages constraining the deposition of the Deseado moraine ridges. The new luminescence age provided for sample LBA12D2-1 in this study suggests that the outwash plain associated with the Deseado II moraine ridge was deposited 123 ± 18 ka, and therefore the outwash associated with the Deseado II moraine was deposited during MIS 6 at a similar time to the Moreno moraines.

Discussion
The new luminescence ages provided in this study (Fig. 7) contribute towards the continued advancement and refinement of the luminescence dating technique in glaciated environments, but the ages also provide insights into the glacial history preserved at Lago Buenos Aires since MIS 6.

Moreno and Deseado moraines at Lago Buenos Aires
Ages provided for the Moreno and Deseado moraines in this study using luminescence and cosmogenic isotope dating suggest that sediments have been preserved at Lago Buenos Aires from glacial advances from~110 ± 20 ka to 150 ± 30 ka. Palaeoenvironmental records linked with Patagonia suggest that it is possible for glacial advances to have occurred during MIS 6. Dust in the EPICA Dome C ice core (75 S, 123 E) which has been geochemically fingerprinted to Patagonia recorded high concentrations during MIS 6, indicating that glacial conditions likely prevailed at this time (Lambert et al., 2008). Also, the d 18 O record indicates that global ice volumes were increased during MIS 6 (Imbrie et al., 1984).
It is interesting to observe that glacial advances during MIS 6 have been preserved at Lago Buenos Aires, but there are currently no sediments recorded from MIS 4, which is a time with similar palaeoenvironmental records to MIS 6 (i.e. increased global ice volumes). In order to further define the glacial record and provide comparisons to the palaeoenvironmental record, more age constraints for the timing of glacial advances prior to MIS 2 and 3 are required from across Patagonia and the Southern Hemisphere.

Fenix moraines at Lago Buenos Aires
The luminescence ages for the deposition of outwash sediments associated with the Fenix moraines at Lago Buenos Aires (Fig. 8e) appear to fall into two groups (grey shading in Fig. 8): earlier advances during MIS 3 and later advances during MIS 2 (referred to as phases MIS 2a and MIS 2b in Fig. 8).

Glacial advances: MIS 3
The oldest Fenix moraine ridges preserved at Lago Buenos Aires are the Fenix IV (26.2 ± 1.6 ka) and Fenix V (25.0 ± 2.3 ka) moraines , but luminescence dating for glaciofluvial sediments suggests that outwash plains were deposited at Lago Buenos Aires as early as 30.8 ± 5.7 ka (LBA12F4-2) and 34.0 ± 6.1 ka (LBA12RF1) during MIS 3 (Fig. 8e). The older outwash plains were deposited at lower elevations in the sedimentary sequence, which may have allowed them to be preserved when the more prominent moraine ridges were destroyed during subsequent glacial advances. dD) data, with age model correlated to the GISP2 ice core time scale using methane and d 18 O from 9 to 57 Ka BP (Brook et al., 2005). (b) Global d 18 O records developed by the SPECMAP project (Imbrie et al., 1984), (c) Dust concentrations from the EPICA Dome C ice core (75 S, 123 E) (Lambert et al., 2008). (d) Sea surface temperature records from ODP core 1233 (41 S, 74 W) (Kaiser et al., 2005). (e) Cosmogenic isotope ages (Douglass et al., 2006;Kaplan et al., 2004Kaplan et al., , 2005 re-calculated by Kaplan et al. (2011) and luminescence ages (this study) for Lago Buenos Aires. Also shown are: (1) radiocarbon ages from the Chilean Lake District (Denton et al., 1999; re-calibrated by Putnam et al., 2013); (2) cosmogenic isotope ages from Lago Pueyrr edon (Hein et al., 2010), the Strait of Magellan and Bahia Inútil (Kaplan et al., 2008;McCulloch et al., 2005) that were all re-calculated by Kaplan et al. (2011); (3) cosmogenic isotope ages from the Última Esperanza region, Chile (Sagredo et al., 2011). Evidence for glacial advances during MIS 3 has also been recorded by radiocarbon dating of plant macrofossils that were overrun by ice in the Chilean Lake District (Fig. 8e; Denton et al., 1999). Cosmogenic isotope dating of moraine boulders (37.1 ± 1.3 ka, 38.7 ± 1.7 ka and 38.0 ± 3.9 ka) in the Última Esperanza region has also identified moraine ridges preserved from glacial advances during MIS 3 (Sagredo et al., 2011). Finally, Darvill et al. (2015) used cosmogenic nuclide dating of outwash sediments to determine glacial limits from MIS 3 for the Bahia Inútil lobe in southern Patagonia, which were more extensive than the MIS 2 limits. No other published evidence of glacial advances at this time exists from Patagonia, but a trend of increasing dust levels in the EPICA Dome C ice core (Fig. 8c) and low sea surface temperatures (SST) off the west coast of Chile (Fig. 8c) are seen between~31 ka and~37 ka; this may be related to an increase in ice extent at Lago Buenos Aires.
The timings of earlier phases of glaciation that occurred during MIS 3 in the Chilean Lake District, at Lago Buenos Aires and at Bahia Inútil suggest that these phases occurred across~12 of latitude in Patagonia. However, summer insolation in the Southern Hemisphere was high~31e37 ka (Berger and Loutre, 1991) and temperatures in West Antarctica (81 S) were warmer than during full glacial conditions~18e25 ka ( Fig. 8a; Brook et al., 2005). The conditions recorded in the palaeoenvironmental archive therefore question why glacial advances~31e37 ka (MIS 3) in Patagonia reached similar lateral extents to the advances during the LGM of MIS 2 (~18e25 ka).

Glacial advances: MIS 2
Cosmogenic isotope ages of the Fenix IV (26.2 ± 1.6 ka) and Fenix V (25.0 ± 2.3 ka) moraine ridges records the major glacial advances at Lago Buenos Aires during MIS 2 . However, the timing of these major advances occurred later at Lago Buenos Aires than is suggested by the evidence preserved elsewhere in Patagonia (Fig. 8e). The major glacial advances at Lago Pueyrr edon (28e30 ka; Hein et al., 2010;Kaplan et al., 2011), in the Chilean Lake District (27 cal. ka; Denton et al., 1999) and in the Strait of Magellan and Bahia Inútil (~27e29 ka; McCulloch et al., 2005;Kaplan et al., 2008Kaplan et al., , 2011 were all consistent and occurred~2 ka prior to the major glacial advances at Lago Buenos Aires (~25 ka). Moraine ridges that were deposited at a similar time to Fenix IV and V at Lago Buenos Aires (~25 ka) are preserved within the limits of larger advances occurring~27 ka at Lago Pueyrr edon (Hein et al., 2010;Kaplan et al., 2011), Bahia Inútil (Kaplan et al., 2008;McCulloch et al., 2005) and in the Chilean Lake District (Denton et al., 1999). Moraine deposition in Patagonia~25 ka ( Fig. 8 grey shading; MIS 2a) occurs at the same time as cooler SST off the west coast of Chile (Fig. 8d) and high concentration of dust in the EPICA Dome C ice core (Fig. 8c).
Multiple glacial advances have then been recorded at Lago Buenos Aires from~18 to 23 ka by the Fenix I to III moraine ridges. Fig. 8e shows that the glacial advances at Lago Buenos Aires from 18 to 23 ka are in phase with records provided from across Patagonia (41e53 S). The Rio Blanco moraine limit 3 in the Lago Pueyrr edon valley records an advance from cosmogenic isotope dating between 20 and 22 ka, respectively (Hein et al., 2010;Kaplan et al., 2011). Glacial advances are also recorded in the Strait of Magellan (~19e20 ka) and Bahia Inútil (~19e25 ka) (Kaplan et al., 2008;McCulloch et al., 2005), which were deposited at similar times to the moraine ridges in the Chilean Lake District at~19, 21 and 25 cal. ka BP (Denton et al., 1999). The palaeoenvironmental record reports that environmental conditions in the Southern Hemisphere fluctuated from~18 to 23 ka (Fig. 8 grey shading; MIS 2b) according to the dust concentrations in the EPICA Dome C ice core (Fig. 8c) and minor fluctuations in SST recorded in the ODP core off the coast of Chile (Fig. 8d). Also, temperatures recorded in the Siple Dome ice core in western Antarctica (Fig. 8a) cooled from 23 ka into a distinctly cooler period occurring~22e23 ka, which was followed by a rapid increase in temperature~21 ka, and the global ice volume increased (Fig. 8b). The similar timings of events in the palaeoenvironmental record and glacial advances across Patagonia suggest that fluctuations in the extent of the Patagonian Ice Sheet were sensitive to changes in climate throughout the LGM.
The glacial record from Patagonia suggests that the final glacial advance occurred~18 ka, which corresponds to the deposition of the Fenix I moraine ridge. Cosmogenic isotope ages for the final glacial advance preserved at Lago Pueyrr edon suggest that it occurred~19e20 ka (Hein et al., 2010;Kaplan et al., 2011) where lake formation then persisted from 15 to 17 ka (Mercer, 1976;Mercer and Sutter, 1982;Turner et al., 2005). This is consistent with cosmogenic isotope dating of former ice limits around Cerro Tamango, Cerro Oportus and Sierra Colorado at Lago Pueyrr edon, where the ice lobe had retreated to within 10e15 km of the present-day extent from 15.6 to 19 ka, with a still-stand or advance occurring~16.9 ka (Boex et al., 2013). Similarly, the final extent of glaciation is recorded at 18 cal. ka in the Chilean Lake District (Denton et al., 1999) and~17 cal. ka in the Strait of Magellan and Bahia Inútil Kaplan et al., 2008Kaplan et al., , 2011. The lacustrine sediments preserved in the sedimentary record at Lago Buenos Aires indicate that a lake formed from~18 ka. A later glacial advance then deposited the Menucos moraine at 16.9 ± 1.4 ka  similar to the still-stand or advance reported at Lago Pueyrr edon (Boex et al., 2013). Glacial conditions must then have persisted for the glaciofluvial sediments sampled by LBA12OW3 (14.7 ± 2.1 ka) to have been deposited. The sedimentary record at Lago Buenos Aires therefore suggests that glacial advances potentially occurred here later than in the rest of Patagonia (i.e. at Lago Pueyrr edon, in the Strait of Magellan and at Bahia Inútil).
Palaeoenvironmental records suggest that glacial conditions could have prevailed across Patagonia up to~15 ka (Fig. 8).
Although the temperature records in West Antarctica (Fig. 8a) and the SST in the ODP core off the coast of Chile had begun to increase from~18 ka, the conditions were still commensurate with large ice volumes present in Patagonia up to~15 ka: (1) the d 18 O records indicate that the global ice volume remained extensive until~15 ka; (2) the dust concentrations in the EPICA Dome C ice core were still decreasing from the MIS 2 maximum; and (3) the SST were increasing, but did not reach significantly warmer temperatures than the MIS 2 maximum until after~15 ka. Therefore, it is plausible that ice persisted at Lago Buenos Aires until~15 ± 2 ka but evidence for this has not been preserved (or discovered) elsewhere. After 15 ± 2 ka the ice front at Lago Buenos Aires retreated into the mountains and re-stabilised during the Younger Dryas,~100 km from the LGM position, and~50 km from the present ice front (Glasser et al., 2012).

Factors controlling glacial advances across Patagonia
The differences in timings of glacial advances during the LGM at sites across Patagonia that span a north-south transect of the Patagonian Ice Sheet may be related to a number of controlling factors, which operate at regional or hemispheric scales. The differences in the topography between the valleys have been suggested to control the style and timings of glacial advances (e.g. Glasser and Jansson, 2005). Overdeepening of the valleys east of the Patagonian Ice Sheet over time caused by the topographicallycontrolled, fast-flowing glaciers (e.g. Lago Buenos Aires) could potentially change the distribution of the ice volume so that the lateral extension of the glacial advance is reduced during larger but later glaciations (Kaplan et al., 2009). Evidence of glacial advances may also not be preserved in all the valleys, especially where overridden by younger glacial advances (e.g. as suspected for MIS 4 glacial advances at Lago Buenos Aires). The glaciofluvial sediments in this study dated to 30.8 ± 5.7 ka (LBA12F4-2) and 34.0 ± 6.1 ka (sample LBA12RF1) provide a good example of these issues as similar aged deposits have rarely been reported from other sites in Patagonia, likely because of the lack of preservation and/or discovery.
For glacial advances at similar latitudes to have been synchronous across Patagonia but vary at different latitudes, large-scale shifts in climate may be controlling glaciation more than regional factors. Boex et al. (2013) have suggested that inconsistencies between glacial advances at different latitudes in Patagonia are related to the precipitation supply controlled by the migration of the Southern Westerlies during the LGM. During the LGM the Southern Westerlies expanded northwards and so the core is reported to have been located~45e50 S (Hulton et al., 1994;Denton et al., 1999). Positioning of the Southern Westerlies at latitudes 45e50 S would have increased the input of precipitation to the piedmont glacier at Lago Buenos Aires (~46 S) and may have facilitated ice expansion at Lago Buenos Aires at this time but not in the rest of Patagonia. However, more evidence is required to substantiate this hypothesis.

Conclusions
Luminescence dating of proglacial samples from Lago Buenos Aires (~46 S), east of the Northern Patagonian Icefield, provides rare evidence in Patagonia for glacial advance (s) during MIS 3 (~31e37 ka). Evidence of the oldest moraines preserved at Lago Buenos Aires from the LGM (MIS 2) is recorded from~25 ka, in phase with LGM advances in the Strait of Magellan and Bahia Inútil, but out-of-phase with the LGM at Lago Pueyrr edon and the Chilean Lake District. Multiple fluctuations in glacial extent are then recorded at Lago Buenos Aires from~17 to 23 ka by three advances, which are in phase with glacial advances that occurred at different latitudes across Patagonia (e.g. Lago Pueyrr edon, Strait of Magellan and Bahia Inútil). In contrast, the timing of the final glacial advance of the LGM preserved at Lago Buenos Aires (~15 ± 2 ka) was out-ofphase with the rest of Patagonia. The palaeoenvironmental records suggest that glacial conditions could have persisted in Patagonia up to~15 ka and so the discrepancy in timings of glacial advances across Patagonia is either related to the lack of preservation of sediments at some sites, or regional discrepancies in climate during deglaciation.
Overall, luminescence dating using single grains of K-feldspar has excellent potential to contribute towards the ever increasing geochronological dataset constraining the timings of glacial advances in the Patagonian region. Improving and increasing these datasets will continue to advance our understanding of the synchronicity of glacial advances across Patagonia, and then begin to unravel the potential signature of past climatic changes on these dynamic systems.