Introduction

The Last Glacial Termination (T1; ~ 18–11.7 ka) represents the largest, most abrupt climate change of the last glacial-interglacial cycle1. T1 featured the collapse of continental-scale ice sheets, a persistent rise in both sea level and the concentration of atmospheric greenhouse gases, along with a millennial-scale sequence of climate changes that culminated with the onset of the current interglacial. Ice core data reveal a step-like pattern of increasing atmospheric temperatures featuring a synchronous but antiphased trend at millennial timescale between the polar hemispheres2,3. Recently developed precise mountain glacier chronologies in the mid-to-high latitudes of both hemispheres4,5,6,7 (see Supplementary material S1), however, are starting to reveal in-phase behavior, pointing to a globally consistent mountain glacier response to climate changes during T1. This pattern challenges the expected antiphased or out-of-phase response predicted by the bipolar seesaw paradigm. If replicated by detailed and precise glacial chronologies, this mismatch could reveal key insights about interhemispheric climate links mediated by some atmospheric component operative during glacial terminations.

Recently, Denton et al.8 proposed that changes in the Southern Westerly Winds (SWW) are the critical missing link relating insolation, atmospheric greenhouse gas concentrations, atmospheric and ocean circulation, and glacier response at millennial timescale during glacial maxima and terminations. By way of interaction with the Southern Ocean (SO), the SWW establish a coupled system that drives global ocean circulation, the atmospheric concentration of greenhouse gases, and high-latitude ocean productivity by enhancing upwelling of CO2-enriched and high-nutrient deep waters9. The efficacy of the SWW-SO coupled system is dependent upon the wind stress imparted by the SWW on the surface of the SO south of the Drake Passage (> 55° S)10. Hence, deciphering the geographical position and strength of the locus of the SWW are crucial for assessing global climate change during ice age terminations.

The SWW are the sole driver of precipitation to the Pacific and Andean divide sectors of northern Patagonia (40°–44° S)11. This region-specific response to incident atmospheric flow allows reconstructing past SWW behavior based on hydrologic balance variations preserved in stratigraphic and geomorphologic records from the Chilean Lake District and Chilotan archipelago (40–44° S). When analyzed in conjunction with similar records from southern Patagonia (50°-54°S), latitudinal shifts and intensity variations in the SWW at millennial-scale during the Last Glacial Maximum (LGM; ~ 35–18 ka) and T112,13 can be identified. However, few studies in northern Patagonia have examined in any detail glacier fluctuations following the onset of T114,15, limiting our understanding about the response of middle latitude austral glaciers and paleoclimate patterns and processes at continental, hemispheric, and global scales.

Mountain glaciers are sensitive to and provide a direct physical link to changing atmospheric conditions. For land-terminating glaciers in particular, moraines distal to present ice limits are unambiguous recorders that past climate changed. Thus, the anatomy of glacier fluctuations during T1 in the mid-latitude Andes not only offers empirical constraints on the interhemispheric synchrony of mountain glacier behavior, but also a means to examine the evolution of the SWW and associated paleoclimates. Building on prior mapping16, we present a glacial geomorphologic map and a 10Be geochronology of the inner Lago Palena/General Vintter (LPGV) basin, centered at ~ 43.9° S; ~ 71.5° W in northern Patagonia, to examine the timing and structure of glacier fluctuations during T1. This basin was covered by an eastward-flowing outlet lobe of the Patagonian Ice Sheet that was fed through several coalescing valleys during the LGM (Fig. 1)17. The foundation of our chronology comes from one of these tributary valleys located ~ 15 km from the southern shore of the lake, at the foot of Cerro Riñón (~ 1790 m asl). The Cerro Riñón valley (informal name) is carved into the North Patagonian batholith18, which was the source of the abundant glacially-transported granitoid boulders found on the local landscape.

Figure 1
figure 1

Glacial geomorphology and chronology of the (a) Lago Palena/General Vintter ice lobe with 10Be ages (n = 4) obtained from the innermost moraine ridge (PV6) and (b) the Cerro Riñón valley glacier with 10Be ages (n = 26) obtained from CR1-CR5 moraine complexes and the additional perched boulder. On top of the boxes the mean moraine age is acompanied by 1 standard deviation (σ) and the Standard Error of the Mean (SEM) including a 3% propagated production rate error20 (Table 1). Individual ages are presented along with internal uncertainty and sample ID. One age in red italics is considered an outlier. Inset maps with the location of sites mentioned in the text. Black dots correspond with glacial chronologies from RG: Río Guanaco; SUE: Seno Última Esperanza. Red dots correspond to paleovegetation reconstruction. LL: Lago Lepué. CP: Canal de la Puntilla. HM: Huelmo mire. Green dot is the location of ODP1233 sediments core. White outlines represent Northern (NPI) and Southern (SPI) Patagonian Icefields and Cordillera Darwin Icefield (CDI). This figure was created on ESRI ArcGIS v10.4 software (www.esri.com).

Results

Previous studies delineated multiple moraines alongside the eastern half of the LPGV, which have been tentatively assigned to the LGM15,16,17,19. Our geomorphological map allows identification of at least six well-preserved, closely spaced arcuate moraine ridges, the innermost of which we name PV6 (Fig. 1). We obtained four 10Be samples from boulders atop PV6, which yielded ages between 20.7 ± 0.4 and 18.9 ± 0.4 ka, with a mean of 19.7 ± 0.7 ka (Fig. 1, Table 1), calculated using the Patagonian regional production rate20 and the time-dependent Lal/Stone scaling scheme (Lm; see “Methods”)21,22. Upstream from the PV moraines, within the Cerro Riñón tributary valley, we distinguish five well-preserved moraine groups (CR1 to CR5 from the outermost to the innermost). A single 10Be sample from a perched erratic boulder resting over polished bedrock outboard of the CR1 moraines, and ~ 30 m above the modern lake surface, affords an age of 16.3 ± 0.4 ka. We interpret this date as a minimum-limiting age for local ice evacuation. The CR1 moraines lie ~ 300 m south from the lake shore and comprise a ~ 500 m long and ~ 10 m high main ridge connected to several minor ridges (Supplementary Fig. S2). We obtained seven 10Be samples that range from 16.6 ± 0.4 to 15.3 ± 0.4 ka, with a mean of 15.9 ± 0.5 ka. Directly inside CR1 are two ridges that form the CR2 moraines, clearly distinguishable from CR1 by their larger sizes (~ 700 m long and ~ 20 m high) and sharper appearance (Supplementary Fig. S2). Six 10Be samples from the largest and most continuous ridge yielded ages between 14.0 ± 0.2 to 13.1 ± 0.2 ka, with a mean of 13.5 ± 0.4 ka. Immediately inboard, separated by a meltwater channel, the CR3 moraine represents the most continuous (~ 1000 m long) and prominent (~ 30 m high) ice-marginal feature of the area. We obtained seven 10Be samples that range in age from 13.5 ± 0.3 to 12.9 ± 0.3 ka, with a mean of 13.1 ± 0.4 ka. Approximately 200 m upstream from CR3, separated by an outwash plain, a group of several discontinuous ridges covered by dense vegetation form the CR4 moraine (Supplementary Fig. S2). Two samples from the outermost moraine ridge of CR4 provide ages of 13.5 ± 0.3 and 12.9 ± 0.3, with a mean of 13.1 ± 0.5 ka (the additional sample LV17-30 (10.6 ± 0.2 ka) was excluded as an outlier23) (Fig. 1, Table 1). CR5 comprises the innermost ice-marginal features, located ~ 500 m upstream from CR4, and consists of a ~ 400 m long group of latero-frontal moraine ridge fragments elevated ~ 20 m above the floor of the most extensive outwash plain in this valley. Two 10Be samples collected from the outermost ridge yielded statistically identical ages of 12.5 ± 0.2 ka.

Table 1 10Be ages from the Lago Palena/General Vintter area calculated using the the non-time-dependent Lal/Stone scaling scheme (St;21,22) time-dependent Lal/Stone scaling (Lm;21,22 and Lifton et al. scaling (LSDn;64).

Discussion

Glacial geologic mapping and thirty new 10Be ages (1 outlier) constitute the basis of a moraine chronology for the LPGV basin that documents in detail the sequence of glacier/paleoclimatic events during T1 in northern Patagonia. Our data indicate that the retreat of the LPGV glacier lobe from the PV6 moraine at 19.7 ± 0.7 ka likely initiated the present lake, and the ice front did not subsequently re-advance out of its current basin. Considering the dispersion of the PV6 ages, we interpret this date as a maximum-limiting age for presumed large-scale glacier withdrawal during T1. The ice front then retreated more than 40% of its LGM length, prompting the detachment of the Cerro Riñón glacier shortly after 16.3 ± 0.4 ka. After this event, a moraine-building event of this tributary glacier culminated with the deposition of the CR1 moraine at 15.9 ± 0.5 ka. Subsequent advances or standstills deposited in quick succession within the error margin of the dating the CR2 to CR4 moraines between 13.5 ± 0.4 and 13.1 ± 0.4 ka. Cerro Riñón glacier then underwent net recession, only interrupted by a stillstand that constructed CR5 at 12.5 ± 0.4 ka. No ice marginal features are evident up valley, suggesting profound glacier retreat to the headwalls after ~ 12.5 ka. The chronology for the Cerro Riñón glacier informs on the timing and structure of glacier fluctuations throughout the entirety of the T1 chron within a single basin in northern Patagonia, and constrains four moraine deposition phases of similar magnitude (i.e., CR1-CR4) that culminated at ~ 15.9 ka during Heinrich Stadial 1 (HS1: ~ 17.8–14.7 ka) and between ~ 13.5–13.1 ka within the Antarctic Cold Reversal (ACR: ~ 14.7–12.7 ka), followed by an additional advance or stillstand of minor extent (i.e., CR5) at ~ 12.5 ka that is coeval with the Younger Dryas (YD: ~ 12.6–11.5 ka).

The Cerro Riñón glacier expanded and achieved its maximum extent during T1 by ~ 15.9 ka. This result differs from the majority of moraine-based records from southern South America dated so far, which document an apparent sustained and large-scale glacier recession during HS1. Notable exceptions are moraines of similar age observed along the eastern flank of the southern Patagonian Andes, such as the 10Be dated Cerro Pintado moraine at Río Guanaco (~ 50°S)24 and the 14C-constrained Lago Pinto moraine at Última Esperanza (~ 50°S)25. Subsequent to HS1, glacier activity in the Cerro Riñón valley represents the northernmost direct evidence for glacier advances during the ACR and gradual retreat during the YD in Patagonia, expanding the known geographical footprint of these glacier/paleoclimate events from 47.5o S to 43.9o S26,27,28,29. The recent interpretation of an ACR maximum based on geomorphic analysis and lake sedimentary record of former Rosselot glacier30, ~ 45 km directly west of Cerro Riñón, is consistent with our results. Considering that Patagonian ice-marginal features formed during the early phases of T1 are often closely spaced, these findings indicate that equilibrium line altitude changes (ELA) at ~ 15.9 ka and during the ACR (CR1-CR4) were similar in magnitude and were followed by a net ELA rise during the YD (CR5), accounting for modest glacier advances or standstills well within ACR limits. At similar latitude, but ~ 9000 km west of LPGV, several moraine records in New Zealand reflect synchronous glacier behavior: massive recession from the LGM limits was punctuated by deposition of the Prospect Hill moraines at ~ 15.9 ka in the Rakaia valley31, which was followed by several ACR advances, and then subsequent YD recession was interrupted by minor stillstands in multiple valleys of the Southern Alps31,32,33,34,35,36 (recalculated, see Supplementary material Table S2-S5). We propose that trans-Pacific glaciers fluctuated in unison at millennial timescales through T1, in response to zonally synchronous changes in the SWW and associated climate anomalies26,33.

We assess the representativeness of the Cerro Riñón glacier record with proxy evidence from lake sediment cores obtained in the Pacific sector of northern Patagonia, which afford valuable information for tracking the regional evolution of the SWW through the LGM and T137,38,39. Pollen records from the Canal de la Puntilla and Huelmo mire in the Chilean Lake District12 and Lago Lepué in Isla Grande de Chiloé40 indicate a treeline ~ 1000 m lower than present and presence of Magellanic Moorland communities in the lowlands during the LGM, attesting to cold and hyperhumid conditions brought by a northward shift of the SWW (Fig. 2). This was followed by rapid arboreal expansion, disappearance of Magellanic Moorland driven by deglacial warming and a southward shift of the SWW starting at ~ 17.8 ka. The interval between ~ 17.8–16.4 ka features a low lake-level stand, with peak abundance of the littoral macrophyte Isoetes (Fig. 2), signaling low SWW influence during the initial phase of T1. Discrete increases in precipitation occurred at ~ 16.4 ka and ~ 14.7 ka, as indicated by conspicuous increases in cold-tolerant hygrophilous conifers (Fitzroya/Pilgerodendron and Podocarpus nubigena, respectively). These changes suggest successive incremental increases for SWW influence in northern Patagonia. The earliest of these increases on the wind belt influence lasted until ~ 15.9 ka, as the deglacial warming trend resumed and crossed a critical threshold that favored the diversification and densification of other thermophilous rainforest trees and vines at the expense of Fitzroya/Pilgerodendron. The youngest increase in conifers (Podocarpus nubigena) during T1 took place between ~ 14.7–12.6 ka, and was followed by enhanced fire activity, a lake-level fall, and decline in conifers after ~ 12.6 ka, which suggest a decrease in precipitation related to a southward shift of the SWW. Wind-driven hydroclimate changes toward cold/wet conditions closely track glacier advances in the LPGV area. We therefore conclude that our record shows a coherent cryospheric response to significant SWW-modulated climate fluctuations during T1. In addition, we note that the reservoir-age corrected paleoclimate records from marine core ODP123341, collected offshore from northwestern Patagonia (Fig. 3), show a pattern consistent with our terrestrial-based chronology of climate change through T1.

Figure 2
figure 2

Glacier chronology of the Lago Palena/General Vintter basin and selected species from Canal de la Puntilla and Huelmo mire from the Chilean Lake District (CLD)12 and Lago Lepué pollen record from Isla Grande de Chiloé (IGC)40. Blue bars highlight pollen-based cold/wet intervals and yellow bars denote pollen-based warm/dry periods within T1.

Figure 3
figure 3

Paleoclimate proxies spanning T1. (a) Records from the North Greenland Ice Core Project (NGRIP): δ18O3; winter temperatures in purple57; summer temperatures in yellow57. (b) Glacier length of the Lago Palena/General Vintter ice lobe (this study). Yellow triangles are bracketing radiocarbon ages for the final glacial advance (100% length) of the LGM in the Chilean Lake District14. (c) Glacier-derived temperatures from Rakaia Valley, New Zealand31,32. (d) Planktonic δ18O from marine core ODP-123328. (e) Opal flux from marine core TN057-13PC9. (f) Integrated δ13C from Talos Dome (TALDICE) and EPICA Dome C (EDC)43. (g) Records from the central West Antarctica Ice Core (WAIS): CO2 in light blue and δ18O in red42. Blue bars highlight cold/wet intervals and yellow bars denote warm/dry periods inferred from palynological analyses.

Widespread glacier withdrawal in Patagonia42 was contemporaneous with distinct atmospheric CO2 (atmCO2) changes recorded at the beginning and end of T1, between ~ 18.1–16.3 ka and ~ 13–11.7 ka in the WAIS ice core43, enhanced ocean ventilation inferred from a rise in opal flux recorded in the SO9, and reduced δ13C composition of atmCO2 preserved in the EPICA Dome C (EDC) and the Talos Dome (TALDICE) ice cores from Antarctica44 (Fig. 3). Glacier advances or stillstands that stalled profound ice recession in northern Patagonia coincided with centennial-scale halts in the rising atmCO2 trend followed by atmCO2 plateaus between ~ 16.3–14.8 ka and ~ 14.8–13.0 ka43, concomitant with decreased deep SO water ventilation9 and minimum δ13C composition of atmCO2 from Antarctic ice cores44 (Fig. 3).

Our collation of mid- and high-latitude paleoclimate data from the Southern Hemisphere suggests that variations in the strength and/or position of the SWW links hydroclimate changes and glacier mass balance variations in the temperate regions of South America and New Zealand. This trans-Pacific atmospheric circulation pattern can explain the observed terrestrial changes, along with simultaneous upwelling and ventilation of deep waters in the SO. Overall, we note that negative mass balance driving glacier recession from the LGM terminus in the LPGV basin was contemporaneous with negative anomalies in SWW influence in northwestern Patagonia40, along with a sustained increase in atmCO243 that coincided with invigorated SO upwelling9. Collectively, these data indicate a poleward shift of the SWW early during T19. Subsequently, a moraine-building event of the Cerro Riñón glacier indicates a positive mass balance episode culminating at ~ 15.9 ka that was concomitant with positive anomalies in SWW at ~ 44° S40, with a pause in the rising trend of atmCO2 rapidly starting at 16.3 ka43, and subdued increase in SO upwelling9. We interpret these correspondences as a simultaneous widening of the SWW belt during the HS1. This was followed by enhanced positive glacier mass balance accounting for multiple ice readvances between ~ 13.5 and 13.1 ka coeval with positive anomalies of the SWW in northwestern Patagonia40, and a stall in the rising of atmCO2 trend43 accompanied by attenuated degassing of the SO9. We interpret these shifts as reflecting increased SWW influence at ~ 44° S with a diminished influence south of ~ 55° S, over the SO, implying a northward shift of the SWW belt. Finally, recurrent negative glacier mass balance prior to ~ 12.5 ka was coeval with negative anomalies in SWW influence in northwestern Patagonia40, resumption of the atmCO2 rising trend43, and enhanced SO upwelling9. This correspondence suggests the occurrence of diminished SWW influence at ~ 44° S and stronger SWW influence south of ~ 51° S from a southward shift of the SWW belt during the YD.

The Cerro Riñón glacier chronology also shares similarities with mountain glacier chronologies from the mid and high northern latitudes42 (see Supplementary material Fig. S3). Those studies show glacier withdrawal from their LGM limits shortly after~ 18.5 ka45,46, broadly coinciding with the onset of T1, readvances between ~ 17 and 16 ka in multiple valleys of the European Alps45,46,47 and between ~ 16.5 and 15.7 ka in western North America48, broadly contemporaneous with the deposition of the CR1 moraine during HS1. A growing body of evidence indicates that glaciers readvanced between ~ 14.5 and 13.5 ka in the Norwegian Arctic5, between ~ 14.3 and 12.8 ka in east Greenland49, and between ~ 14.5 and 12.8 ka in western North America50, coeval with the formation of the CR2–CR4 moraines during the ACR. Subsequent moraine formation events between ~ 12.8 and 12.3 ka in east Greenland51, southern Alaska6, and Scotland7 occurred during the early YD, and several sites across the European Alps during the late YD52,53,54,55,56. These advances or standstills are indistinguishable in age with the CR5 moraine when considering dating uncertainties, and occurred just before widespread and profound ice recession to the valley headwalls. We acknowledge that significant glacial activity during T1 occurred within the YD and few dates overlapping with the ACR have been reported in the majority of northern hemisphere sites42. A large number of these glacier advances, however, culminated either early or late within the interval (see Supplementary material Fig. S3), indicating that glaciers were receding through much of the YD.

The pervasive interhemispheric synchrony in mountain glaciation during T1 lies in direct contrast to the largely antiphased polar ice core records (Fig. 3). Studies have attributed the severity of cooling in Greenland ice cores throughout HS1 and YD time to a seasonality switch related to episodes of extended sea ice cover in the North Atlantic region, skewing their isotopic records toward a predominantly winter temperature signal57,58. Mountain glacier fluctuations, in contrast, respond primarily to summer temperatures which remained comparatively warm during key episodes of T1 as reflected by modelling experiments complemented with multiproxy records from Europe59.

Our results and analysis show that the timing and structure of glacier fluctuations in the LPGV basin were coeval with climate events recorded in both the northern (i.e., HS1 and YD) and southern (i.e., ACR) hemispheres. From our interhemispheric comparison we interpret synchrony of mountain glaciers driven by a global climate signal during T1, challenging the expected antiphase behavior predicted from the bipolar seesaw paradigm. This finding favors an atmospheric mechanism for generating and globally propagating millennial-scale climate variability during T1. We conclude that high-to-middle latitude mountain glaciers fluctuated in phase during T1 partly in response to summer warming due to atmCO2 concentrations brought by changes in the SWW-SO coupled system4. Our findings further support recently hypothesized climate mechanism, dubbed the Zealandia Switch, which proposed that climate variability during T1 may have been triggered by orbitally-induced Southern Hemisphere warming and globally paced by changes on the austral atmospheric and oceanic circulation at millennial timescales8.

Methods

Geomorphological mapping

Detailed geomorphological mapping of the moraine limits in Lago Palena/General Vintter basin was conducted based on aerial photographs (GEOTEC 1:70,000—www.saf.cl), satellite imagery (Sentinel 2—https://scihub.copernicus.eu/) and digital elevation models (ALOS Palsar—https://search.asf.alaska.edu/). The preliminary map was checked during two field campaigns (March, September 2017 and January 2020).

Rock samples collection

We collected 26 boulder samples for 10Be surface exposure dating from the Cerro Riñón moraines (CR complexes), in addition to one sample from a perched boulder (LV17-42) located immediately outboard of the outermost moraine limit, and 4 boulder samples from the innermost moraine ridge along the eastern shore of Lago Palena/General Vintter (i.e., PV6). We aimed for large boulders to avoid potential effects of post-depositional movements or posterior exhumation. Samples were taken from the upper ~ 4 cm of the boulder surfaces using a drill and explosive charges, avoiding areas exhibiting clear signs of erosion, such as spalling or flaking. Elevation and geographic coordinates of each rock sample were recorded with a handheld GPS unit (WGS84). We measured topographic shielding using a handheld compass and a clinometer. Moraine ages are interpreted as representing the culmination of moraine construction, and, thus, cold episodes, whereas the perched-boulder age provides a minimum age for ice retreat from the sampling site.

Quartz separation and 10Be isolation

Initial crushing and sieving of the rock samples was carried out at the Pontificia Universidad Católica de Chile and subsequent quartz and beryllium extraction at the Cosmogenic Nuclide Laboratory at Lamont-Doherty Earth Observatory following the protocol outlined in20,60. 10Be/9Be ratios were measured at Lawrence Livermore National Laboratory Center for Accelerator Mass Spectrometry. Ratios were measured relative to the standard 07KNSTD with a 10Be/9Be ratio of 2.85 × 10–12 (61; 10Be half-life = 1.36 Myr.). Analytical raw data are available in the supplementary material S1. Given relatively recent improvements described in20,60, the average analytical uncertainty is ~ 2%, with almost half the analyses between 1.7% and 2.0%.

10Be surface exposure ages

10Be ages were calculated based on methods incorporated in the online exposure age calculator (v.3—https://hess.ess.washington.edu/math/v3/v3_age_in.html)62, considering the time-dependent Lal/Stone (Lm)21,22 scaling schemes and the regional Patagonian production rate (~ 50o S)20, assuming zero erosion and a rock density of 2.65 g/cm3. We discuss ages based on Lm scaling scheme because it produces the ages that best fit with limiting radiocarbon data from the production-rate calibration site at Lago Argentino (~ 50° S)20. In addition, most of the Patagonia 10Be glacial chronologies were reported according to the same scaling scheme. We show in Table 1 that 10Be ages are statistically identical (i.e., accuracy) using other scaling methods. In the text, we report individual 10Be ages with 1σ analytical uncertainty and the standard error of the mean (SEM). Raw geographic and 10Be analytical data can be found in the Supplementary material Table S1. For comparison with other proxy records, mean moraine ages include the propagation of the analytical uncertainty and that of the local production rate (3%)20. Similarly, we recalculated available glacier 10Be chronologies from New Zealand8 by using the online exposure age calculator (v.3)62 incorporating a local production rate (~ 43.6° S)63. Recalculated 10Be ages are reported with the associated uncertainties (1σ), including a 2% propagation of the production rate error63 (See Supplementary material Tables S2, S3, S4 and S5). We note that Denton et al.8 used exposure age calculator v2.262 based on63 and reported cosmogenic ages calibrated to the year A.D. 1950. Therefore, our recalculated 10Be ages differ from those in Denton et al.8 by 1–2%, which does not alter our interpretations.