The reconstruction of the glacial evolution in the LBBb is based on geomorphological surveying and mapping, and 10Be exposure ages obtained from moraine boulders and polished surfaces. Data confirms Hubschman’s proposal that the most weathered moraine (EM) was formed during a glacial advance prior the Eemian Interglacial, as well as the onset of the final retreat of the Garonne Glacier during the gLGM.
5.1. Interpretation of the CRE results
The sequence of CRE ages from the terminal moraine system of the Upper Garonne Valley is constrained by the long time since the penultimate deglaciation of the area and subsequent postglacial environmental dynamics, which together with land use changes caused by humans within the last millennia in the LBBb, have affected the preservation of the glacial records. Therefore, the reconstruction of the glacial history in the area is challenging, and our study needs to be complemented with other environmental proxies and glacial chronological data from the region.
The exposure ages of the samples (ARAN-01, ARAN-44, ARAN-45 and ARAN-46) from boulders distributed on the moraine ridges EM-2 and EM-4 (123.8 ± 8.0 ka and 124.3 ± 10.3 ka, respectively) show statistical consistence based on the Chi-2 test. They support the hypothesis that the oldest moraine system existing in the glacial terminal basin of the Upper Garonne Valley corresponds to the PGC. These dates confirm the deposit of those moraines during a glacial advance before the Eemian Interglacial period, as postulated in previous studies (Hubschman 1984; Andrieu et al. 1988).
The exposure ages from the samples taken from the highest moraines of the EM (ARAN-03 and ARAN-04) yielded 73.1 ± 4.6 and 31.1 ± 2.9 ka, respectively, showing a lack of consistency between them. These two ages suggest that these moraines might correspond to polygenic features that formed over a long time period when the glacier occupied the top of the Castillon Peak and finally shrunk by 31.1 ± 2.9 ka before the onset of the Barbazan Lake infill, located 1.5 km south (Fig. 4-B; Andrieu et al. 1988). However, it is also likely that our boulders had shifted from their original position by subsequent glacial advances or had been exhumated by postglacial erosion processes occurring on the slopes of the Castillon Peak. Therefore, more data is needed to confirm one of the scenarios and verify the link with the maximum glacial advance. Thus, given this uncertainty, for the reconstruction of the glacial history in the area, these samples were excluded.
The exposure age of 17.1 ± 1.8 ka (ARAN-07) obtained from the IM-1 moraine ridge near Burs must be excluded due to its mismatch with the Barbazan Lake records located upvalley, which indicates an older retreat than our exposure ages (Andrieu et al. 1988):
(i) A radiocarbon age of 31.160 ± 1.700 year BP (40.7–32.3 cal ka BP) at 2,263–2,274 cm (base of the core) within the basal unit (glaciolacustrine rhythmites and diamictons). Given the location of the Barbazan paleolake with respect to the IM-1, the area of the lake was covered by the ice when IM-1 was formed. Hence, this proglacial lake may only have infilled after the Garonne Glacier retreated from the IM-1. Hence, IM-1 should be older than 40.7–32.3 cal ka BP.
(ii) The sedimentary frame of the glaciolacustrine sequence shows that the Garonne ice margin was close to the paleolake (probably at IM-2) until the interruption of glacial meltwater supply. This interruption occurred at 26.600 ± 460 year BP (31.6–30.1 cal ka BP; Andrieu 1991), or just after 23.980 ± 680 year BP (29.9–27.2 cal ka BP). It is however possible that the moraine boulders from the IM-1 ridge have been affected by reworking following lLGM glacial retreat and provide thus younger ages than expected.
New CRE ages from polished surfaces on the left margin of the Marignac basin (ARAN-42: 20.7 ± 1.2 ka and ARAN-43: 24.2 ± 2.1 ka) allow us to locate the terminal position of the Garonne palaeoglacier upstream of Marignac at the time of the gLGM.
5.2. Chronology of the glacial advances in the Upper Garonne Valley
Several smooth moraine ridges were mapped within the external moraine system of the LBBb, and ascribed to the oldest glacial advance in the area (Hubschman 1984; Andrieu 1991; Stange et al. 2014; Fernandes et al. 2017). Our CRE dates reveal two moraine ridges deposited prior to the LGC in the Upper Garonne Valley at 123.8 ± 8.0 and 124 ± 9.6 ka. Thus, four CRE dates from the EM-2 and EM-4 ridges provided evidence of moraine stabilization during a time interval spanning from 128.5 ± 9.1 to 116.4 ± 9.4 ka (Table 2). Isotopic inheritance is likely to be absent for moraines far from the source (~ 20–80 km), however the youngest ages might have been exhumated following the onset of moraine stabilization (Briner et al., 2005). Therefore, the oldest age of these units most likely marks the time of moraine stabilization following the MIS 6 largest glacial advance at ~ 129 ka. As temperatures increased during the transition toward the Eemian Interglacial (Helmens 2013), the Garonne Glacier started receding and moraines stabilized. This PGC glacial advance must have covered the entire glacial terminal basin of the Upper Garonne Valley with an ice thickness ranging between 200 to 50 m (Fig. 6). According to our results, during this phase, the main Upper Garonne Glacier was ~ 78 km long and covered ~ 900 km2. Considering the position of the frontal moraine ridge EM-2 at 480 m, and the derived palaeoglacier reconstruction, the ELA was located at 1,711 ± -65/55 m (AAR), 1,719 ± -95/60 m (global AABR) and 1,704 ± -105/70 m (mid-latitude AABR), with an average of 1,711 m. This altitude suggests a reduction of 9.3 ºC with respect of current ELA at 3,139 m and assuming no change in the summer precipitation (Campos et al., 2021). As expected, these results show that glaciers had similar extents in the terminal basin during the MIS 6 and lLGM glacial advances and thus climate conditions were probably similar (Fernandes et al., 2017).
No solid chronological data are available to support a straightforward interpretation of the glacial advance during the PGC. The lack of available boulders for CRE dating in the EM-1 ridge impeded establishing the timing of the maximum extent during the PGC in the Upper Garonne Valley, whose external position on the moraine suggest that it may also belong to the PGC (Fig. 6). Indeed, there is evidence indicating that the Pyrenees were extensively glaciated during this period. In the Eastern Pyrenees, a stalagmitic flowstone stemmed from the onset of karst activity by 124.6 ± 6.9 to 121.4 ± 9.4 ka at the Niaux-Lombrives-Sabart cave occurred after the MIS 6 cold period (Sorriaux et al. 2016). In the same catchment, other age obtained from an erratic boulder located 50–100 m above the LGC moraines, reinforced the hypothesis of a previous glaciation that took place in the Ariège Valley at 133.9 ± 5.3 ka (Delmas et al. 2011). In the southern slope of the Pyrenees, an older age from the PGC has also been reported in the Aragón Valley where the outermost moraine, 80 m above the present-day riverbed, was dated at 171.0 ± 22.0 ka (García-Ruiz et al. 2013).
Geomorphological evidence from the LGC in the LBBb must thus be located in the internal part of the basin and foot slopes close to the Barbazan village. Here, the La Serre and the Burs ridges (IM-1) were generated by a piedmont glacier covering the basin during the lLGM of the LGC. The only available ages correspond to the Barbazan Lake sequence, where the onset of the proglacial lake infill, behind this moraine, started before 40.7–32.3 cal ka BP (Hubschman 1984; Andrieu 1991). In several valleys of the southern slope of the Pyrenees, glacial evidence suggests that the lLGM occurred during the MIS 4 and MIS 3, namely at 65 − 55 ka and 45 − 30 ka (Oliva et al. 2019). However, the chronology of the lLGM glacial advances in the southern slope is not as robust as in the northern slope. In the Ariège Valley, a CRE age from an erratic boulder located on the hilltop between Tarascon and Foix-Montgaillard basins yielded 79.9 ± 14.3 ka and a boulder from a lateral moraine on the confluence between the Aston and Ariège valleys yielded an exposure age of 35.3 ± 8.6 ka (Delmas et al. 2011). In addition, radiocarbon dating of the first organic remnants from the bottom of the ice-marginal deposits behind such moraines has shown the onset of the post-lLGM deglaciation between 48 and 24 ka, namely at the Estarrès Lake, Gave d’Ossau (34.2–29.7 cal ka BP; Andrieu 1987; Andrieu et al. 1988; Jalut et al. 1988), at the Biscaye peatbog, Gave de Pau (48.3–39.7 cal ka BP; Mardones and Jalut 1983), and at the Freychinède sequence, Ariège Valley (27.3–24.0 cal ka BP; Jalut et al. 1982; Reille and Andrieu 1993). Therefore, the lLGM glacial advance of the LGC in the LBBb is necessarily older than the onset of the sedimentation at the Barbazan Lake, as the glaciolacustrine rhythmites and diamictons at the base of the core sequence accumulated after the glacial retreat that followed the previous glacial advance, which likely coincided with the formation of the IM-1 moraine. Glacial advances during the MIS 3 are more evident in the Eastern Pyrenees, where moraines from Têt Valley were dated at 40.86 ± 1.9 ka (Tomkins et al. 2018) or those in the Malniu area that yielded 51.1 ± 4.8–42.6 ± 4.1 ka (Pallàs et al. 2010).
After the lLGM recession, a moraine-dammed lake formed between the IM-1 ridge and the glacier front blocked the meltwater discharges and filled the Barbazan proglacial lake (Andrieu 1991). The same author indicated that the glaciolacustrine rhythmites and diamictons at the bottom of the sequence (40.7–32.3 cal ka BP) was transported from a nearby source, probably synchronously with the formation of the moraine IM-2. However, no chronological data are yet available to discriminate whether the moraine was formed during a pulsation after the lLGM or as a result of the gLGM advance. We only can hypothesize that it must be older than 31.6–30.1 cal ka BP because sediments and pollen records showed the progressive glacial abandonment of the terminal basin, with a transition from glaciolacustrine to lacustrine sediments, a reduction of freshwater inputs from the glacier, as well as the decline of the forest coverage (e.g. Fagus sp.) and the recovery of herbaceous species (Andrieu et al. 1988; Jalut et al. 1992).
The slopes surrounding the Marignac basin were ice-free during the gLGM, as demonstrated by the exposure ages obtained from polished surfaces at 50 and 30 m above the basin floor. At 24 − 21 ka, the glacier would have abandoned the terminal basin and split into two individualized glacier tongues constrained within the Garonne and la Pique valleys as the ice shrank. At this time, the palaeoglacier of the Upper Garonne Valley flowed down 60 km along the main valley. In parallel, glacial recession was also underway during the gLGM in the Ariège Valley, where polished surfaces were dated to 18.7 ± 3.8 ka at 20 m above the gLGM Bompas moraine (490 m) (Delmas et al. 2011). In this case the gLGM was depicted with glacial advances leaving several well-preserved moraines (Garrabet, Bernière, Bompas-Arignac): a boulder from the Bernière frontal moraine, ~ 7 km from the lLGM ice limits was dated at 18.8 ± 1.3 ka (Delmas et al. 2011). Subsequently, after the gLGM, glaciers in the Upper Garonne Valley underwent a massive retreat upvalley, reaching the mouth of the highest cirques by ~ 15 − 14 ka (Oliva et al. 2021; Fernandes et al. 2021).
5.3. Mid-Late Pleistocene glacial dynamics in the Central Pyrenees in the context of European mid-latitude regions
In Eurasia, climatic models have shown that the glacial maximum of the PGC was the most extensive of the last 400 ka (Colleoni et al. 2016). This glacial maximum occurred at 140 ka (MIS 6) based on Antarctic ice core records (Winograd et al. 1992; Colleoni et al. 2016), which is also confirmed by the minimum sea-level (likely − 150 m) dated at 155 − 140 ka (Grant et al. 2014; Wekerle et al. 2016). The Greenland ice core recorded the Eemian Interglacial from 129 to 114 ka with a warming peak occurring at 126 ka (Dahl-Jensen et al. 2013). These records correspond to a sea-level stabilization at 130 − 119 ka, reaching up to 6–9 m above the current level (Hearty et al. 2007). According to the Greenland ice cores, this interglacial was up to 8 ± 4 ºC warmer than the last millennium (Dahl-Jensen et al. 2013). The beginning of the Eemian caused a massive ice discharge from Northern Hemisphere ice sheets, when glaciers flowed into the ocean leaving icebergs that drifted debris (> 150 µm) southwards as far as ~ 55ºN at 128 ka (McManus et al. 1999). In fact, disintegration of the northern ice sheets occurred during the end of the PGC, which coincided with the Heinrich Stadial 11 (~ 136 − 129; Menviel et al. 2019).
According to marine sediments from the Iberian margins, a rapid warming with an increase of annual sea surface temperatures of ~ 10 ºC followed the MIS 6 (Martrat et al. 2007). The comparison between the deep-sea cores in the Portuguese margins and European pollen records showed that in southwestern Iberia the warmest and driest period of this interglacial occurred between 126 and 117 ka (Sánchez-Goñi et al. 2005). Such warm conditions after the MIS 6 suggests non-favourable conditions to glacial development in the Iberian Peninsula. Consequently, it is likely that mid-latitude mountain glaciers, such as those existing in the Pyrenees, underwent a massive retreat leaving their terminal basins and probably the lowest peaks.
No robust chronological data on glacial records have been obtained so far to confirm the MIS 6 glaciation in the Pyrenees. Indeed, the wide temporal range (170 − 120 ka) of exposure ages and associated uncertainties do not let us to have a clear idea about the maximum of the PGC in the Pyrenees. In any case, in this work, we introduce the first 10Be CRE dataset from moraine boulders in Pyrenees that reveal the occurrence of a large glaciation during the MIS 6 in the Upper Garonne Valley. At that time, the ice covered the terminal basin and the glacial system was more extensive than during the lLGM advance. The onset of the penultimate deglaciation started at ~ 129 ka, when the Garonne Glacier abandoned the two moraine ridges (EM-2 and EM-4) located at 480 m and they stabilized.
Intense postglacial geomorphic dynamics (glacial, periglacial, slope and alluvial) have eroded glacial landforms left by previous glaciations and therefore, the glacial evidence from MIS 6 in the Pyrenees can be only found in a few valleys. The glacial deposits beyond the limits of the LGC in this mountain range have been only recently ascribed to the PGC and gradually confirmed by optically stimulated luminescence (OSL) and CRE dating techniques (Oliva et al. 2019). These deposits, which are highly degraded, presenting only few scattered boulders suitable to CRE dating, are normally located in flat areas far from slope, periglacial and alluvial processes. Several studies have been published showing the glacial and karst evidence from the PGC in the Pyrenees and interpreted from the MIS 6 (Delmas et al. 2011; García-Ruiz et al. 2013; Sorriaux et al. 2016). Within the uncertainties, these evidences are also supported by a fluvial aggradation episode at 178 and 151 ka based on the existence of fluvial and fluvioglacial sediments in fluvial terraces of the Cinca and Gállego valleys, respectively (Lewis et al. 2009).
Available dates of glacial processes during the PGC elsewhere in the Iberian Peninsula are also scarce. Similarly to what has happened with the LGC (Oliva et al. 2019), the increasing application of CRE methods is showing that the most extensive glaciers developed at the end of the PGC. CRE ages between 140 and 120 ka have been reported in several mountain regions regarding the most external moraines. This is the case of Sierra Nevada, where the lowest moraine was dated at 130–135 ka; Serra da Estrela, where the highest right lateral moraine was developed by ~ 140 ka; and even in the NW ranges, where a push moraine and a polished surface were dated at 155 and 131 ka (Table 4).
Glacial evidence of the PGC has also been detected in other European mountains confirming the maximum glacial advance of the PGC during the MIS 6. This pattern occurred in the Alps, where one erratic boulder in the Jura Mountains was dated using two cosmogenic nuclides, 21Ne and 10Be, yielding 128 and 106 ka, respectively (Ivy-Ochs et al. 2006). In the Austrian Alps, OSL was applied to glaciofluvial, fluviolacustrine and eolian sediments showing a culmination of the PGC during the late MIS 6 (149 − 135 ka) (Bickel et al. 2015). Further south, very similar ages were obtained from U-series in the coastal mountains of the Adriatic Sea revealing cold conditions during the MIS 6, with moraine development starting at 125 ka in the Bijela Gora plateau (Hughes et al. 2010) and at 131 ka for the moraines from the Mount Tymphi, at the northern slope of the Pindus Mountains (Hughes et al. 2006).
During the gLGM, the terminal LBBb was ice free with the glaciers retreating towards headwaters at 24 − 21 ka. This is supported by exposures ages from polished surfaces, at 18 km southwards the EM-1 ridge, on the lower slopes of the Marignac overdeepened basin that became ice-free by that time (Fig. 7). Therefore, our results also introduce new chronological data about glacial dynamics occurring during the gLGM in the Central Pyrenees. In other valleys of this range, glaciers showed contrasting patterns of glacial advance or retreat during the gLGM. In the Eastern Pyrenees, a glacial advance occurred synchronous with the gLGM, that was noticed in the Têt Valley where a moraine at ~ 1,690 m was dated to 25 ka (Tomkins et al. 2018); and in the La Llosa and Duran valleys, where moraines at 1,520 m and at 1,830 m were dated to 20 and 21 ka, respectively (Andrés et al. 2018). In the southern slope of the Pyrenees, glacial retreat was deduced in the upper sector of the Gállego Valley based on a paleolake located at ~ 1,500 m dammed by landslide by 20 ka (García-Ruiz et al. 2003).
In the Iberian Peninsula, a general glacial advance occurred at 22 − 19 ka followed by a massive recession (Oliva et al. 2019). In the central and northern part of the Iberia, other valleys also recorded glacial advances, such as in the Iberian Range, where the end of the gLGM was recorded based on lacustrine sediments in the Sierra de Neila was recorded prior to 21 ka (Vegas Salamanca 2007) or in NW ranges where the glacier front in the Tera Valley also remained stable until 22 cal ka BP (Rodríguez-Rodríguez et al. 2011). In the southern Iberian Peninsula, glaciers advanced during the gLGM in the Serra Nevada, where moraine stabilization in the San Juan Valley occurred at 21 − 19 ka (Palacios et al. 2016). In other European mountains, such as the Alps (Ivy-Ochs et al. 2008) or the at the Tatra Mountains (Engel et al. 2015), glaciers reached their maximum position of the LGC at 26 − 21 ka and undergone a subsequence massive retreat afterwards, at 20 − 19 ka. This suggests that the Garonne Glacier was already retreating when CO2 concentrations in the atmosphere were still low (180–200 ppm; Shakun et al. 2015).