The Principal Cordillera in the Andes of Santiago is drained by the Maipo River and its tributary, the Mapocho River, whose watersheds reach up to more than 6,000 m a.s.l. (Figure 1). The Maipo River Andean catchment includes the tributaries of the Yeso River and Volcán River, where we focused our investigation due to the identification of a well-preserved geomorphological record as well as its preferential zonal orientation, suitable for a regional paleoclimate comparison (Figure 1).

The Volcán River catchment encompasses the mountain relief located from ∼1,200 m a.s.l. up to ∼6,100 m a.s.l. and between 33°40′-33°57′ S and 70°13′-69°50′ W. It is part of the Maipo River Andean watershed, bound on the east by the Volcán San José (5,856 m a.s.l.) and Cerro Marmolejo (6,108 m a.s.l.) peaks (Figures 1, 2). La Engorda drift, as defined in this work, is located at 2,450–2,570 m a.s.l. in the namesake river valley that directly drains the San José volcano (Figure 2). Downstream, the Volcán River flows from the highest mountains toward the confluence of three fluvial systems near the town of San Gabriel, located at 1,200 m a.s.l. There, the Maipo River, coming from the south, joins the Volcán and Yeso tributary rivers (Figure 1).

Although evidence for Quaternary glaciations in central Chile has been reported since the mid-20th century, chronological constraints for glacier advancements and paleoclimatic implications are scarce. Brüggen (1950), Borde (1966), and Chiu (1991) performed geomorphological mapping in the area, partially including sediments from Quaternary glaciations in the Andean watershed of the Maipo River. More recently, Ormeño (2007), Rauld (2011), and Herrera (2016) reported geomorphological records of late Quaternary glacial advances in the localities of San Gabriel and La Engorda in the Maipo River Andean watershed (Figure 1).

From a regional perspective, evidence of glacier advances in the subtropical western slope of the Andes from ¹⁰Be exposure ages in moraines has been documented from Cordón de Doña Rosa (∼30°S), simultaneous with the global LGM and at 43 ka (Zech et al., 2006; Zech et al., 2007; Zech et al., 2008; Zech et al., 2011; Zech et al., 2017), and at 20–18 ka, ∼33–25 ka, and ∼40 ka at the El Encierro River valley (∼29°S; Aguilar et al., 2022). At Laguna del Maule (∼36°S), Singer et al. (2000) concluded that the last ice retreat in the area started between 25.6 and 23.3 ka, partly concomitant with respect to the timing of the global LGM. At the Cachapoal River valley (∼34°S), glacier stages have been proposed from ¹⁰Be exposure ages in lateral moraines at 3.8 ka, 13.5 ka, and 21.9–20.3 ka (Charrier et al., 2019).

In the Mendoza River valley located on the eastern slope of the Andes at ∼32° 45′S, Espizúa (1993); Espizúa (1999) provided ages to constrain the occurrence of glacial stages. From U-series and ²³⁰Th/²³²Th ages on travertines overlying glacial geomorphological features, and from thermoluminescence (TL) dating of non-glacial sediments, Espizúa (1993); Espizúa (1999) estimated minimum ages of 38.3 ± 5.3–31.0 ± 3.1 ka and between 24.2 ± 2.0 and 22.8 ± 3.1 ka for travertines found in between two till deposits constituting the “Penitentes Drift”; of 15.0 ± 1.5 and 15.0 ± 2.1 ka for laminated and non-glaciated sediments overlying the “Horcones Drift,” considered roughly synchronous with the global LGM; and of 14 ka BP or 11–10 ka BP for the “Almacenes Drift”. Similarly, from radiocarbon dating of non-glaciated sediments in Valle Hermoso (−35°S), Espizúa and Bigazzi (1998) and Espizúa (2004) provided minimum ages for three tills named “Hermoso I, II and III”, located at 1970, 2,170, and 2,500 m a.s.l., respectively, apparently comparable with the oldest three tills described at Mendoza River valley, with ca. 15.9 ± 0.1 ka BP and 13.6 ± 0.1 ka BP as minimum ages for the “Hermoso II” and ca. 10.6 ± 0.1 ka BP as a minimum age for the “Hermoso III” tills. In addition, Espizúa (1993); Espizúa (1999) inferred younger advances described as neoglacial (“Confluencia Drift”; 3,300 m a.s.l.) in the Mendoza River valley.

The identification and mapping of the drifts defined here were performed using aerial photographs (1:50,000), LANDSAT (15 m pixel resolution) and SPOT (5 m pixel resolution) satellite images, a Digital Elevation Model (DEM, 15 m resolution), and extensive fieldwork supported by handheld GPS location, together with sedimentological and geochronological analyses.

We described geomorphological features located in paleo-ablation zones, specifically corresponding to terminal, frontal, and lateral moraines, and outwash terraces and lacustrine deposits, following the sedimentary facies analysis associated with glacial landforms to interpret the glacial advances and related morphodynamics (Church and Ryder, 1972; Edwards, 1986; Miller, 1996). The state of conservation of these morphologies is evidence that they have not been surpassed by any further glacial development (Edwards, 1986; Miller, 1996), and therefore, terminal moraines are considered to represent glacial advances during the last glaciation. Interpretation from facies analysis is included together with the description of each corresponding drift.

We have identified four drifts named San Gabriel, in the confluence of the Yeso, Volcán, and Maipo rivers, and La Engorda, Colina River, and Las Arenas, in the Volcán River catchment within the Andean watershed of the Maipo River (Figures 1, 2), providing geochronological constraints and detailed characterization for the first three of them. We have also identified and dated well-stratified fine-grained (silt and clay) and sandy layers that we interpret as paraglacial deposits located in the watershed of the Mapocho River (Figure 1). Geochronological analyses performed in this study include optically stimulated luminescence (OSL) dating from quartz grains in sediments (Supplementary Tables SA1−A3), cosmogenic ³⁶Cl exposure dating from boulders on moraine surfaces (Supplementary Tables SB1−B4), and AMS ¹⁴C dating of organic sediment (Supplementary Table SD1). We have also collected samples for ¹⁰Be exposure dating, but the amount of quartz was not enough for the AMS analysis (Supplementary Table SC1), consistent with the dominancy of andesitic and basaltic volcanic rocks in the area.

Geochronological determination of the San Gabriel drift was based upon OSL dating performed on sandy sediments interbedded between gravel- and boulder-rich layers, constituting an outwash terrace developed downstream from a terminal moraine cut by El Yeso River at San Gabriel (Figure 3; 4). A similar geochronological analysis was performed on well-sorted sandy sediments interpreted as paraglacial facies in the Mapocho River catchment (Figure 5), reflecting low-energy sedimentary environments that enhance solar resetting (e.g., Schaetzl and Forman, 2008).

The sampling collection process for OSL dating required the insertion of tubes of 5 cm in diameter and approximately 10–15 cm in length in the sediment after the exposed outcrop had been removed. This ensured a minimum sample weight of 300 g. The sample extraction required the tube to be completely inserted into the sediment, a process that could be assisted by a rubber hammer. After extracting the tube with the sediments, the edges had to be sealed with opaque tape to ensure the protection of the sample from further light exposure. The outermost 1 cm of the sediment sampled in the tube was removed, leaving unexposed sediment in the center of the tube for optical dating. We choose the OSL method due to its advantage to obtain reliable age results from quartz- and/or feldspar-rich sediments formed in glacifluvial environments during the late Quaternary using single aliquot regeneration (SAR) and multiple-aliquot regeneration (MAR) protocols (Annex A; Supplementary Tables SA1−A3).

Geochronological determinations on morphostratigraphic units at La Engorda area were obtained by cosmogenic ³⁶Cl exposure, dating from the top of boulders located at the surface of ice-contact moraines (Supplementary Tables SB1−B4), and ¹⁴C, dating from organic material from sediments overlaying or embedded within the moraines (Supplementary Table SD1).

Sampled boulders were characterized by a low degree of weathering, with flat and isolated surfaces located on mounds corresponding to terminal moraines and in an intermorainic zone (Figure 6). The main lithology of the boulders corresponded to breccia and conglomerates in the sandy matrix (Figure 7), most possibly associated with the Jurassic Río Damas Formation (Thiele, 1980). A hammer and chisel were used to extract about 1 kg of the most superficial layer (∼5 cm) from each boulder, with the aim of maximizing the recovery of cosmogenic isotopes. A detailed explanation of the methodology is provided in the Supplementary Material (Annex B).

Radiocarbon dating was performed on bulk sedimentary organic matter obtained from the samples collected from the layers overlaying the moraines, which are interpreted to represent the average age of disseminated carbon within each sample. Samples were acid-washed to remove secondary carbonate and pretreated to eliminate labile organic matter to analyze the more resistant organic material. Conventional radiocarbon results from bulk organic matter in acidified sediments were then calibrated using Calib Radiocarbon Calibration Program (Stuiver and Reimer, 1993; Stuiver et al., 2005), and SHCal20 atmospheric database (Hogg et al., 2020). We report a 2σ calibrated age range together with the corresponding highest relative probability (Supplementary Table SD1).

Samples were extracted from stratigraphic sections observed in pits and dug from the surface of both terminal moraines M1 and M2 (Figure 6) and from the surface of the alluvial plain composed of glaciofluvial and glaciolacustrine sediments inset in these moraines at La Engorda. These radiocarbon ages are considered as minimum ages for the stabilization of the landform.

The San Gabriel drift is composed of remains of lateral and frontal moraines located at the confluence of the Maipo, Volcán, and Yeso rivers, together with an extensive outwash terrace developed downstream from this last area (Figure 8). Alluvial fans, alluvial plains, colluvium, and landslides overlay these last features. Fluvial terraces and modern rivers are inset within glacial units.

Lateral moraines are well exposed around the confluence between the Maipo and Volcán rivers, reaching altitudes of ∼1,370–1,500 m a.s.l. In the Volcán River, distal lateral moraines of east–west orientation are located on both sides of the valley close to the confluence with the Maipo River. These moraines are composed of boulders and angular gravels containing granitoids as a dominant -but not exclusive-lithology, most probably of the San Gabriel stock (Thiele, 1980), immersed within a sandy and fine-grained matrix. The moraines appear to be well preserved from erosive fluvial processes and exhibit a characteristic elongated form (Figure 8). In the confluence area of the Volcán and Maipo rivers, moraine remnants are located on the western side of the Maipo River valley, overlaying the rock substratum and reaching ∼1,400 m a.s.l. Other diamictons associated with the frontal position of the Quaternary glaciers can be recognized as remnants of frontal or bottom moraines eroded by the Volcán and Maipo rivers.

In the Yeso River valley, 2 km upstream of the confluence with the Maipo River, remains of an eroded frontal moraine area located close to the modern Yeso River at an altitude of 1,300 m a.s.l. There, a polymictic diamicton is exposed. The deposit is composed of boulders and angular–subangular gravel within a sandy and fine matrix, including boulders of breccia and conglomerate of 5–10 m in diameter that are most likely associated with the Río Damas Formation (Thiele, 1980), whose outcrops are located more than 20 km upstream in this valley (Figures 3, 8). Downstream of this moraine, an outwash terrace can be morphostratigraphically correlated with the same kind of surfaces developed at the confluence of the Volcán and Maipo rivers (Figure 8).

Outwash terraces are located more than 20 m above the thalweg of the rivers. These are composed of stratified deposits of boulders and are well rounded to subrounded gravels within the coarse and fine sandy matrix, exhibiting well-developed reverse-normal grain size grading and incipient lamination (Figure 4). These sedimentary facies can be interpreted as a product of hyperconcentrated flows produced by floods associated with ice melting in a proglacial environment. Outwash terrace deposits are normally related to episodes or pulses of glacial meltwater discharge, which generate sediment discharge that alters the fluvial bed load (Church and Ryder, 1972; Edwards, 1986; Miller, 1996).

Three OSL ages were obtained from the laminar sands located near the base and close to the top of the sediments constituting the outwash terrace in El Yeso River, close to the confluence with the Volcán and upper Maipo rivers (Supplementary Tables SA1−A3). An age of 46.0 ± 4.7 ka was obtained from the sediments in section C1, located just above the terminal moraine (Figure 9). Ages of 44.7 ± 3.7 ka and 36.4 ± 3.0 ka were obtained from the sediments in section C2, located approximately 1.5 km downstream (Figure 9), suggesting progradation of the outwash plain downstream since the frontal moraine was deposited (Figure 8). We interpret these ages as the occurrence of fluctuating melting of glaciers between ∼46 and ∼36 ka, defining an early local Last Glacial Maximum stage (ELGM) before the global Last Glacial Maximum (LGM; Clark et al., 2009; Hughes, 2021).

We interpret the San Gabriel drift as resulting from glacial advances during the Late Pleistocene. Once the glaciers reached their maximum advance within the Maipo, Volcán, and Yeso river valleys, which is registered by the lateral and frontal terminal moraines in this area (Figure 8), glacial melting began, providing meltwater that transported debris coming from the moraines. This process was aided by subglacial and supra- and intraglacial flow, as well as from the adjacent slopes, providing discharge that contributed to the formation of fluvial terraces downstream and upstream when the glaciers rapidly retreat (Edwards, 1986; Miller, 1996; Bennett and Glasser, 2009). This can explain the conspicuous terraces that developed downstream and upstream of the frontal moraine in the Maipo, Yeso, and Volcán rivers in the San Gabriel area (Figure 8).

In the Mapocho River catchment, well-stratified sandy and fine sediments are located at an altitude of approximately 1,200 m a.s.l., preserved on a terrace above the modern river thalweg (Figures 2, 5). These sedimentary facies are completely different to those of the modern deposits in the river and ravines in the area, mostly composed of gravel and coarse-to-medium sand facies. No moraines can be recognized in the area, possibly due to erosion. The position of this terrace, along with the well-stratified fine material that constitutes it and the occurrence of collapse features in the sedimentary package, suggest a paraglacial or proglacial environment (Church and Ryder, 1972). Two OSL ages from these sediments yielded 60.4 ± 5.1 and 53.9 ± 5.0 ka, suggesting glacial activity during MIS4 (Supplementary Tables SA1−A3), i.e., before the glacier advances documented in this study for the San Gabriel drift in the Maipo River Andean catchment.

La Engorda drift is composed of N-S-oriented frontal moraines named M1 and M2, which are located at 2,570 and 2,450 m a.s.l. for the inner and outer ridges, respectively, with an inboard outwash plain and alluvium overlying glaciolacustrine sediments (Figures 6, 10). Alluvial fans, glaciofluvial plains, colluvium, and landslides overlay glacial landforms in the area (Figure 10). The modern river is inset in the moraines, showing a relatively recent change in the local base level due to the rupturing of the natural dam composed of the latter deposits (Figure 11).

Frontal moraines reach approximately 180 m in height from the river thalweg, crossing the western limit of La Engorda valley perpendicularly (Figure 11). The inner ridge -M1- corresponds to a well-defined hill located to the east of the terminal moraine system, and the outer ridge -M2- corresponds to a less defined hill that is nonetheless easily observable in the field to the west of the same moraine system (Figure 6).

Moraines at La Engorda are composed of massive diamictons interpreted as till facies, with angular and rounded clasts of variable sizes ranging from some centimeters to large boulders several meters in size, embedded within a fine-to-medium sandy matrix that forms more than 50% of the deposits. Polymictic clasts are chaotically distributed within the deposit. Among these, sedimentary rocks predominate, mainly conglomerates and sandstones from the Río Damas Formation (Thiele, 1980), together with volcanic breccia associated with igneous material from the nearby San José volcano.

Exposure ³⁶Cl dating from boulders yielded ages of 14.8 ± 2.8 ka and 11.9 ± 2.0 ka for those coming from the outer -M2- ridge, 15.0 ± 3.2 ka from the one sampled in the intermoraine depression (M), and 10.0 ± 2.2 ka and 5.4 ± 1.2 ka in the case of those sampled on top of the inner–M1- ridge (Figure 11; Supplementary Tables SB1−B4). We interpret most boulders as eroded and transported rapidly during late glacial advances. The recorded boulder ages fall within an interval spanning the Antarctic Cold Reversal (ACR) and Younger Dryas (YD) chronozones and into the early Holocene. The spread in ages for most of the ∼15–10 ka set is like the spread observed when cosmogenic dating is applied to late Holocene glacier advances after taking into consideration the large uncertainty values (e.g., Fernández-Navarro et al., 2023). Radiocarbon dating of inboard paludal, fluvial, and alluvial sediments onlapping the moraine ridge to the east do not record sediment older than ∼10 ka at the base (Pits 1, 2, and 3; Figure 12; 13), being consistent with the boulders ages, except for the youngest. The younger boulder age (∼5.4 ka) can be attributed to re-exposure of the surface of the boulder, potentially covered by loose sediment, to a rockfall associated with nearby rockslides (Chiu, 1991) or to a Holocene glacial readvance; further research is needed to elucidate these possibilities.

We considered the possibility that all the boulders could be simply exhumed global LGM blocks that, at some point, were buried under a large lake. All dated boulders, however, are located on ridges at least 100 m above the floor of the valleys to the west and south. It is improbable that the boulders were exhumed after being covered by a large lake covering several km², which should have extended towards the southwest. There are no remains associated with a drainage obstacle or lacustrine sediments suggesting a lake with these dimensions existed at some point between the LGM and ∼10 ka. Only at the junction with the Colina River, about 1.5 km south of the M1 and M2 moraines on the south hillslope of the valley, an outcrop of lacustrine sediments suggests a short blockage of the Colina fluvial system, probably associated to a frontal moraine during the early LGM (Figure 14), which was later partially covered by rockslides coming from the south and the east (Chiu, 1991). This early moraine probably never reached the elevation of the ridges (more than 100 m above this point). Radiocarbon ages obtained from these lacustrine facies overlaying a diamicton deposit at this site yielded ages between 32.5 and 35.8 cal ka BP (Figure 14), which are comparable with a radiocarbon age of 24.5 ± 0.4 ka BP (28.7 ± 0.9 cal ka BP), previously obtained at the base of these lacustrine sediments by Moreno et al. (1991).

The surface of the ridges associated with the main body of each moraine is partially overlain by sandy sediments interstratified with well-sorted fine deposits exhibiting incipient lamination, which we interpret as the local reworking of material by alluvial and eolian processes at the surface of the moraines. Dark organic-rich layers and soil development are observed on top (Figure 12). All these sediments are younger than ∼8.7 cal ka BP and as recent as ∼2.8–1.1 cal ka BP, demonstrating the Holocene evolution of the surface of these moraines following the last ice retreat in this area, as suggested by the cosmogenic dates (Figures 11, 12). The lack of an obvious buried LGM surface below these Holocene sediments is consistent with the idea of only late glacial advances recorded at this elevation in the catchment, even after considering a potential lag of organic material deposition at this elevation after the LGM (Figure 11).

After deposition of the frontal moraines and initial ice retreat up the valley, a proglacial lake developed, evidenced by glaciolacustrine sediments situated below the younger glaciofluvial and alluvial material constituting the outwash plain inset, inboard the moraines (sensu Edwards, 1986; Miller, 1996; Figure 11). Glaciolacustrine deposits correspond to massive cohesive mud with interstratified incipiently laminated sandy facies. A radiocarbon date obtained at the base of the massive glaciolacustrine sediments, in this position of the landscape, yielded an age of 10.2 ± 0.2 cal ka BP (Figure 13), which is consistent with the ∼10 ka cosmogenic age determination obtained from a boulder on top of the inner moraine ridge (Figure 11). Two radiocarbon ages obtained from bulk sedimentary organic matter from fine sandy layers and organic horizons, interpreted as glacio-fluvial and alluvial sediments discordantly overlying the massive glaciolacustrine deposits, yielded ages of 6.5 ± 0.1 cal ka BP and 2.3 ± 0.5 cal ka BP, respectively, suggesting the prevalence of a proglacial environment during most of the Holocene (Figures 11, 13).

Modern fluvial and alluvial sediments inset in the moraines are composed of rounded, sub-rounded, and angular boulders, gravel, and pebbles arranged chaotically in a sandy and fine cohesive matrix, with some of these deposits exhibiting inverse-normal grain size gradation that can be interpreted as bed load and debris flow deposits, different from those constituting the outwash plain (Figure 11).

Modern glaciers in the Andes in front of Santiago are located at altitudes above 3,200 m a.s.l. in the Volcán, Yeso, and Maipo rivers watersheds (Figure 1). These glaciers are located more than 30 km to the east from the San Gabriel drift (Figures 1, 2). We interpret the San Gabriel drift moraines as a frontal moraine system, which includes lateral moraines marking the maximum altitude reached by the late Quaternary glacial tongues at that place (∼1,500 m a.s.l.), while the location of the frontal moraines determines the lowest position of the glaciers coming through the three river valleys (∼1,300 m a.s.l.). Because glacial deposits have not been observed downstream, we interpret the frontal moraines in the San Gabriel area as a terminal system, which is the largest preserved glacial advance during the late Quaternary within the Maipo Andean catchment (Figures 1, 2). This occurred between ∼46 and 36 ka according to the OSL ages obtained from outwash sediments in this drift (Figure 9), which suggest an early local Last Glacial Maximum (ELGM) in the region.

As previously shown (Figure 5), well-stratified silty and sandy sediments interpreted as paraglacial facies and dated as ∼60–54 ka are located at ∼1,200 m a.s.l. in the Mapocho River valley, at a similar altitude to the San Gabriel drift. Since no lateral or frontal moraines have been observed in this zone, additional evidence is needed to rule out the origin of these deposits to other processes, such as, for example, an ephemeral dam produced by local landslides. If those sedimentary layers have a glacial origin as we propose, it could imply the occurrence of a MIS4 glacial advance in the Mapocho River catchment before any observed in the Maipo Andean catchment.

From the study of the sedimentary record at Laguna Tagua-Tagua, located in the central valley of central Chile (∼34°S), Valero-Garcés et al. (2005) suggested wet conditions during the last glaciation before 43.5 cal ka BP and between 40 and 21.5 cal ka BP, evidenced by increased lake levels and the presence of pollen corresponding to Valdivian rainforest taxa, which appears only south of 37°S during modern times. The period between 42.1 and 40.1 cal ka BP was characterized by relatively lower temperatures and decreased moisture conditions, concomitantly with a trend toward minimum summer insolation in the southern hemisphere (Valero-Garcés et al., 2005). Regional wet conditions are consistent with paleoceanographic interpretations from ocean sediment cores obtained off central and central-northern Chile (33°S and 27°S), which suggest more humid and cold conditions being responsible for the increased runoff between ∼46 and 36 cal ka BP (Lamy et al., 1998; Lamy et al., 1999; Lamy et al., 2000; Kim et al., 2002; Stuut and Lamy, 2004; Figure 15). Thus, a period of increased precipitation and cold conditions could explain an earlier advancement of the Andean mountain glaciers in the Maipo River catchment, previous to the global LGM, which occurred between ∼46 and 36 ka BP according to the San Gabriel drift record (Figure 5). This suggests glacier advances partly contemporaneous with the earlier development of the Patagonian glaciers during the late Quaternary (Denton et al., 1999; Zech et al., 2017; García et al., 2018; Davies et al., 2020).

The age range determined from the San Gabriel drift is comparable with the ¹⁰Be cosmogenic age data of ∼50–33 ka obtained from glacial landforms in the Cordón de Doña Rosa, located at ∼30°S on the western slope of the Andes by Zech et al. (2007); Zech et al. (2008); recalculated in Zech et al., 2017; according to recent production rate data), and with the older ages of ∼35 and ∼40 ka reported in the El Encierro valley (∼29°S; recalculated from Zech et al., 2008, in Zech et al., 2017; Aguilar et al., 2022). In addition, the San Gabriel drift appears relatively contemporaneous to the Pleistocene glaciations in the Mendoza River and Río Grande catchments on the eastern slope of the Andes (32.45–35.30°S), corresponding to the older Penitentes and eventually Valle Hermoso I glacial units, respectively (Espizúa, 1993; Espizúa, 2004; Espizúa and Bigazzi, 1998; Figure 15), suggesting the zonal response of Andean mountain glaciers to regional climate changes during the Late Pleistocene at these latitudes.

It is possible that increased precipitation accompanied by cold conditions at the time of the San Gabriel drift would have been driven by a northward shift of the mid-latitude westerlies during glacial times, as previously suggested (Lamy et al., 1998; Lamy et al., 1999; Lamy et al., 2000; Stuut and Lamy, 2004; Valero- Garcés et al., 2005; Zech et al., 2007; Zech et al., 2008; Zech et al., 2011; Zech et al., 2017, 2018), which could also explain the older Penitentes drift, from which minimum ages of ∼38–31 ka BP have been reported in the Mendoza River valley (∼32° 45′S; Espizúa, 1993; Espizúa, 1999).

La Engorda drift is composed of two moraine ridges, which are located approximately 200–250 m above the Volcán River valley (Figure 10). This difference in altitude allows classifying this valley as a paleo-hanging glacier valley, since the paleoglacier of La Engorda was probably a tributary at the time of this drift and most possibly not directly connected with the main glacier filling the valley of the Volcán River.

Our geochronological data from La Engorda drift suggest two main pulses of glacial advance during late glacial times. Moraines M1 and M2 record ages between ∼15 and ∼10 ka, with a younger (single) dated boulder on the innermost, eastern edge of this drift and adjacent to the outwash-alluvial plain (Figure 11). The two younger ages in the older set of the drift fall, within 1-sigma, into the Younger Dryas (YD) chronozone, a millennial-duration cold period that punctuated the last termination between 12.9 and 11.7 ka (Broecker et al., 2010; Sweatman, 2021), and the early Holocene. In addition, our data suggest a local readvance of the glaciers in the area concomitantly with the ACR, and there are late glacial advances reported from the Patagonian Andes (Sagredo et al., 2011; Sagredo et al., 2018; Strelin et al., 2011; García et al., 2012; García et al., 2014; Reynhout et al., 2019; Davies et al., 2020). Radiocarbon ages obtained from glaciolacustrine deposits onlapping the younger moraine are close to ∼10.0 cal ka BP (minimum ages), consistent with the age inferred for this inner moraine using our cosmogenic dataset and supporting the late glacial advances at ∼15–10 ka followed by a local ice retreat afterward. The presence of fine eolian sediments locally reworked by alluvial processes on top of the moraines (Figures 11, 12) is also consistent with a local glacial advance during the late glacial advances at this locality.

Our data do not record any drift associated with the global LGM in the region, contrasting with other records such as the Cachapoal (Charrier et al., 2019) and the Encierro Valley, Cordón Doña Rosa, and Las Leñas valley (Zech et al., 2017). Between the San Gabriel and La Engorda drifts, a glacial advance that can possibly be traced to the LGM might be recorded at the confluence area of the Volcán River valley with Las Arenas valley (Figure 10). Sediments from this advance appear to be mostly covered by a large rockslide. A potential glacial advance during the global LGM in central Chile could have occurred in synchrony with generally wet regional conditions, inferred from the Laguna Tagua-Tagua sedimentary record, which is in the central valley south of the study region at 34°30′S (Valero-Garcés et al., 2005; Figure 15). The above-mentioned advance would have also occurred concomitantly with higher coastal humidity along central-northern Chile and with cold sea surface temperatures (SSTs) prevailing in the southeastern Pacific during the LGM off central Chile until ∼18 ka (Lamy et al., 1998; Lamy et al., 1999; Lamy et al., 2000; Kim et al., 2002; Stuut and Lamy, 2004; Figure 13).

Glacier advances recorded at La Engorda drift between ∼15 and 10 ka occurred concomitantly with a change in regional climate from humid to arid conditions at the late Pleistocene–Holocene transition, as inferred from the Tagua-Tagua record and the Aculeo record, located also in the central valley at 33°50′S (Valero-Garcés et al., 2005; Jenny et al., 2002; Jenny et al., 2003; Figure 15). The Tagua-Tagua record in central Chile suggests a period of increased precipitation between 13.5 and 11.5 cal ka BP, most possibly due to the northward influence of mid-latitude westerlies at that time, followed by the most arid period -inferred for the last 46 ka-during the early and mid-Holocene (Valero-Garcés et al., 2005), which is consistent with the extremely arid to arid conditions prevailing since the beginning of the Holocene until 5.7 cal ka BP, as inferred from the Aculeo record (Jenny et al., 2002). Thus, it is possible that the intensified influence of the mid-latitude westerlies would have driven increased precipitation under relatively cold conditions in central Chile, favoring positive mass balance and glacier advances at La Engorda during ∼15–10 ka, partly concomitantly with the ACR and YD periods. This happened in the context of a general warming trend of SSTs off central Chile during the latest Pleistocene, followed by relatively low SSTs during the early Holocene (Kim et al., 2002; Figure 15). Low coastal paleo-SSTs have been inferred from isotope composition of mollusk shells in northern Chile during the late Pleistocene–early Holocene transition (23°S and 25°S; Vargas et al., 2006; Flores et al., 2018). Meanwhile, similar to modern coastal paleo-SSTs have been inferred from mollusk shells during the early Holocene in central-northern Chile (31°S; Carré et al., 2012), suggesting a strong latitudinal gradient and reinforced atmospheric circulation associated with the southeastern Pacific Subtropical Anticyclone consistent with La Niña-like conditions at that time (Vargas et al., 2006), which would have favored increased aridity during the early Holocene in central Chile.

Non-glacial sediments stratigraphically located between the Horcones and Almacenes drift deposits on the eastern slope of the Andes yield U-series and ²³⁰Th/²³²Th ages of 15.0 ± 1.5 and 15.0 ± 2.1 ka (Espizúa, 1993; Espizúa, 1999). This is interpreted as a non-glacial interval during this period, which indicates that the Horcones drift deposits would correspond to a glacial advance that occurred prior to ∼15 ka. The Almacenes moraines would be associated with glacier advancements that occurred after ∼15 ka (Espizúa, 1993; Espizúa, 1999). In the Río Grande valley (∼35°S), radiocarbon ages from peats in a swamp located within the limits of the moraine system of Valle Hermoso II yield minimum values of ∼15.9–13.6 ka BP (i.e., ∼19.2–16.3 cal ka BP; Espizúa and Bigazzi, 1998; Espizúa, 2004; Supplementary Table SD1). A minimum radiocarbon age of ca. 10.6 ka BP (i.e., ∼12.4 cal ka BP; Supplementary Table SD1) was obtained from the Hermoso III tills in the Río Grande valley (Espizúa and Bigazzi, 1998; Espizúa, 2004). Although further research is needed on both sides of the Andes to increase the precision of the geochronological data, the comparison between our findings from the western slope, with previous data from the eastern slope of the Andes in central Chile and Argentina, supports the generally consistent regional trends in glacier advances during the latest Pleistocene–early Holocene transition period (Figure 15).

La Engorda drift is the first reported evidence of regional manifestations for a readvance of mountain glaciers at a time interval spanning the ACR and YD chronozones, and the early Holocene, in the Andes of Santiago and in central Chile. In the Cachapoal River catchment (∼34°20′S), located immediately south of the Maipo Andean watershed, the boulders on top of the lateral moraines exposed to ¹⁰Be exposure ages from boulders on top of lateral moraines show values of 20.3 ± 2.9 ka, 21.9 ± 5.3 ka, and one date of 13.5 ± 2.4 ka (Charrier et al., 2019). Although none of the ages in our study record the global LGM, deposits potentially associated with it might be covered with rockslide sediments just downstream of La Engorda. If this is the case, frontal moraines ridges concomitant with the global LGM at this latitude might have reached 2000 m a.s.l. in the Volcán River catchment in the Maipo Andean watershed, while in the Cachapoal River catchment, lateral moraines with similar ages are located at ∼950 m a.s.l. (Charrier et al., 2019). This sharp change in the altitudinal location of the moraines in neighboring Andean catchments could imply a strong latitudinal gradient in the precipitation pattern between ∼33.5°S and ∼34.33°S on the western slope of the Andes in central Chile at the time of the global LGM. Additional research is needed to confirm this inference. The presence of Valdivian rainforest taxa in the corresponding chronostratigraphic horizons in the Laguna Tagua-Tagua record supports the inference of increased precipitation in the central valley at ∼34.5°S concomitantly with the global LGM period (Valero-Garcés et al., 2005), while conspicuous eolian sediments dated between ∼21.6 and 17.3 ka in a paleoseismological trench in the west Andean piedmont would suggest arid conditions in addition to aeolian sediment supply from distal or abandoned alluvial systems in the Santiago valley at that time (Vargas et al., 2014). On the contrary, the similar altitudinal location of the frontal moraine dated as ∼15–10 ka at La Engorda in the Maipo Andean watershed (2,570 m a.s.l.), compared to the altitudinal location of the lateral moraine of similar age in the Cachapoal catchment (2,700 m a.s.l.), supports the idea of similar factors driving readvances of mountain glaciers at that time in the Andes of central Chile. This effect was probably driven by the intensified northward influence of the mid-latitude westerlies during the latest Pleistocene, as previously proposed for glacier advances concomitantly with the ACR period, followed by more variable conditions during the YD chronozone at Patagonia (Sagredo et al., 2018).

Glacier advances during the ELGM and the LGM, reported here in the Andean catchment of the Maipo River and in the Cachapoal River catchment (Charrier et al., 2019), respectively, occurred concomitantly with relatively higher austral summer insolation (Figure 15), with low SSTs prevailing offshore in the Southeastern Pacific (Kim et al., 2002), and during globally persistent cold conditions, as revealed by ice cores from Antarctica (Petit et al., 1999; Labeyrie et al., 2003; Villalba et al., 2009). On the contrary, the glacier advances during late glacial times in central Chile occurred concomitantly with lower austral summer insolation in the context of a general warming trend of SSTs in the Southeastern Pacific (Kim et al., 2002; Figure 15). In both scenarios, increased precipitation seems to have played a driving role for glacier advances in the Andean catchments in central Chile, as in modern times (McPhee et al., 2014). In this last sense, if a mid-to-late Holocene glacier advance is confirmed, as suggested by the youngest boulder dated from La Engorda drift and by Cachapoal catchment data (Charrier et al., 2019; Figure 15), it would be interesting to compare these mid-Holocene glacier advances with the onset of modern El Niño (Rodbell et al., 1999) and modern El Niño tropical–extratropical teleconnection patterns (Vargas et al., 2006; Figure 15), which include episodes of increased precipitation in central Chile (Rutllant and Fuenzalida, 1991).

Paleo-ELA estimates in the Andes are extremely scarce. On the eastern slope of the subtropical Andean mountains in Argentina, (Espizúa, 1993; Espizúa, 2004), estimated the altitude of the modern ELA at ∼4,500 m a.s.l. Using the accumulation–area ratio (AAR) method (Benn and Gemmell, 1997), the depression in the elevation of the ELA has been estimated at 1,000, 1,200, and 1,250 m a.s.l. at the time of Horcones, Penitentes, and Punta de Vacas drift, respectively (Espizúa, 1993; Espizúa, 1999). Altitudes of ∼3,250 ± 50 and 3,350 ± 50 m a.s.l. have been estimated for the modern ELA in the case of the Azufre and El Peñón glaciers (∼35°S).

The study of paleo-ELAs can be carried out through the application of different methods that consider hypsometric differences between glaciers. The AAR method assumes that the accumulation area of a glacier encompasses a fixed proportion of the total glacier area (Benn and Gemmell, 1997). This method requires the mapping of the extent of paleo-glaciers by means of geomorphological observations (Porter, 1981). Observations of the variability of modern ELAs on the western slope of the Andes suggest that this altitude strongly depends on the amount of snowfall during the austral winter (Masiokas et al., 2006; Garreaud et al., 2008). Precipitation during the cold austral season is therefore a key factor for the development of glacial mountain systems in central Chile.

Estimating the ELA of an ancient glacier using any morphometric method provides results for the glacier in steady state (whereby the balance state of the paleo-ELA is similar to the current ELA), which is associated with a particular climatic condition (Osmaston, 2005). According to the AAR method (Porter, 1975; Porter, 1981; Porter, 2001; Benn and Gemmell, 1997), and considering an accumulation area/total glacier area ratio of 0.6 ± 0.1 with the aim of characterizing glacier valley conditions on a global scale (Singh et al., 2011), we estimated the location of the modern ELA in the study region close to ∼4,600 m a.s.l. ELA estimates inferred from atmospheric (temperature and precipitation) variables for the Andes of central Chile (between 33°-34°S), for the period of 1958–2018, show mean values in the range of about 3,700–4,300 m a.s.l., with a significant increasing trend in the first 2 decades of the 21st century (Barria et al., 2019). Through the AAR method (with a similar ratio of 0.6 ± 0.1), we estimate that the ELA would have been situated close to 3,400 m a.s.l. and 3,600 m a.s.l. at the time of San Gabriel and La Engorda drifts, i.e., ∼1,200 m and ∼1,000 m lower than their current positions in the region, respectively.

For comparison, we also estimated the paleo-ELA using the area–altitude balance ratio (AABR) method, which is most suitable for white, snow-fed glaciers (Benn and Lehmkuhl, 2000; Osmaston, 2005). This method should not be applied to glaciers that receive massive avalanche inputs, where mass balances are non-linear or where there is extensive debris cover, as in the case of the Andean glaciers studied here. The application of this method shows ELA values estimated between ∼3,100 and ∼3,000 m a.s.l., in the case of the San Gabriel drift and between ∼3,600 and ∼3,500 m a.s.l. for La Engorda drift (Herrera, 2016).