Late Pleistocene and Holocene glaciation in the Pyrenees: a critical review and new evidence from 10Be exposure ages, south-central Pyrenees

A compilation and a critical assessment of the 14C data set available show that the chronology of glacial events in the Pyrenees is not well constrained. After reviewing the literature on glacial reconstruction, we suggest a simplified subdivision of the Pyrenean last glacial cycle record into Last Pleniglacial, Deglaciation, and Neoglacial. To improve the numerical glacial chronology, we provide 10Be surface exposure ages for 5 glacial erosion surfaces, 9 moraines and 2 erratics in the Upper Noguera Ribagorcana Valley (south-central Pyrenees). Published corrected 14C data and new 10Be exposure ages indicate that the major phase of moraine building recorded in this valley during the Last Pleniglacial probably occurred after 25 ka BP. This age calls in question the generally accepted hypothesis of a very early deglaciation of the Pyrenees ca 70–40 ka BP, and strongly suggests that the Pyrenees could have been in pleniglacial conditions during the Last Glacial Maximum (LGM). However, we do not exclude the possibility that the maximum glacier extent during the last glacial cycle had taken place much earlier than the LGM, as indicated by some published U-Th, AMS 14C and OSL data. We suggest that pleniglacial conditions could have taken place during a longer (⪢30–20 ka) period than generally assumed, and that the Last Pleniglacial could include several glacier fluctuations recorded irregularly in different valleys, with a last major glacier readvance taking place around the LGM. In addition, the Deglaciation is represented by a series of moraines deposited between ca 13.7±0.9 and 10.1±0.6 ka. This moraine series indicates a highly variable climatic pattern that is partly correlated with Greenland Stadial 1 (the Younger Dryas), and suggests that the Deglaciation could have continued into the early Holocene.


Introduction
The complex oceanographic, atmospheric and glacial system in the North Atlantic region is considered to have played a significant role in the rapid and global climate fluctuations experienced during the late Quaternary (e.g. Alley and Clark, 1999). Recent studies clearly demonstrate that the Iberian Peninsula and surrounding offshore areas are highly sensitive to the variability of the North Atlantic system (e.g. Pérez Obiol and Julià, 1994;Cacho et al., 1999Cacho et al., , 2000Cacho et al., , 2001Sanchez Goñi et al., 2000Combourieu et al., 2002;Moreno et al., 2002).
The Pyrenees are located in the north of the Iberian Peninsula, between the Atlantic and the Mediterranean. In this context, the Pyrenean glacial record is a potential source of relevant paleoclimatic data. Despite its relatively low resolution, when correlated and combined with other marine or terrestrial paleoclimate proxies, the glacial record of mountain areas may provide additional constraints on regional precipitation patterns, atmospheric temperature, and on predominant wind regimes.
Current knowledge of the Pyrenean glacial record is based on numerous local geomorphological and sedimentological studies that provide a relative chronology of glacial events valid for individual valleys. Similarities between valleys have given rise to tentative correlations and several valuable regional syntheses (Taillefer, 1969;Bru et al., 1985;Andrieu et al., 1988;Bordonau, 1992;Bordonau et al. 1992;Serrat et al., 1994;García Ruiz and Martí Bono, 1994;Serrano and Martínez de Pisón, 1994;Gómez Ortiz and Salvador, 1994;Peña Monné et al., 1997;Gómez Ortiz et al., 2000;Calvet, 2004). However, numerical chronological data are scarce, and the resulting chronological framework remains largely speculative. The lack of a robust numerical chronology hinders the correlation with other paleoclimate proxies, and constitutes the main obstacle to extracting paleoclimatic information from the glacial record of the Pyrenees.
With the advent of new geochronological techniques such as cosmic ray exposure dating, it is now possible to improve the chronological dataset of the last Pyrenean deglaciation. The Upper Noguera Ribagorçana Valley (UNRV) in the south-central Pyrenees is specially suited to 10 addition, Hetu and Gangloff (1989) had reported a till deposit located 3 km beyond the limits of the outermost frontal moraines in the Ossau Valley (a tributary of the Oloron Valley), which probably corresponds to an even older glaciation. Additional indirect evidence based on karst stratigraphy and speleothem U/Th chronologies support the interpretation of several phases of extensive glacier development since 350,000 yr, which may be correlated with the last glaciation, MIS 6, 8 and, with less certainty, with MIS 10 (Sorriaux, 1981;Bakalowicz et al., 1984;Quinif and Maire, 1998). More recently, Peña Monné et al. (2004) reported two OSL ages c. 155 ka for a glaciofluvial terrace laterally related to the most external till outcrop in the Gállego Valley, providing numerical evidence to attribute these sediments to the penultimate glaciation (i.e. correlative with MIS 6).
While there is strong evidence that the Pyrenean glacial record includes deposits from several glaciations, the record of the Last Glaciation is the best preserved, showing the freshest forms and the largest volumes of glacial deposits. Despite variations in the morphostratigraphy of different areas, the record of the last glacial cycle in the Pyrenees is consistent with the following evolutive phases.

The Last Pleniglacial
The term Pleniglacial is used in the literature to refer to the period of maximum and nearmaximum glacial extent, irrespective of the age of glaciation (e.g. Taillefer, 1969;Serrano and Martínez de Pisón, 1994). In the present paper we use the term Last Pleniglacial to refer to the same concept, but restricted to the Last Glaciation. Both in the northern and southern Pyrenees this period is recorded by the most prominent morainic suite and by the thickest juxtaglacial successions. According to Bordonau (1992Bordonau ( , 1993 and Bordonau et al. (1992), these large depositional forms necessitate stability of ice volumes for long periods. Instead of using the term Pleniglacial, these authors defined the following phases: "pre-maximum stabilisation", "maximum", and "post-maximum stabilisation".
There is general consensus on the view that the Pyrenean last glacial maximum took place several thousands of years earlier than the global Last Glacial Maximum, LGM, (Mardones and Jalut, 1983;Vilaplana, 1983a,b;Andrieu et al., 1988;Jalut et al, 1988Jalut et al, , 1992Bordonau, 1992;Gillespie and Molnar, 1995;Florineth and Schluchter, 2000;García Ruiz et al., 2003;Calvet, 2004). This idea is suggested by 14 C dates from sediments in the basal units of glaciolacustrine successions which, according to their geomorphological location, must have been deposited during or after the Last Pleniglacial period. The most ancient dates come from the work in the Pau Valley (northern Pyrennees) by Mardones and Jalut (1983). They reported an age of 38,400 14 C yr BP for the nearly basal units of the lacustrine succession of Biscaye, deposited inwards from the Last Pleniglacial piedmont moraines. Extrapolating from the oldest dates in this succession, Mardones and Jalut (1983) calculated an age of c. 45 ka BP for the start of the lacustrine deposition and suggested an age between 70 and 50 ka BP for the piedmont moraines.
Similarly, the glaciolacustrine succession of Llestui (Noguera Ribagorçana Valley, southern Pyrenees), yielded 14 C ages between ≥ 34,000 and ≥ 31,500 14 C yr BP (Vilaplana, 1983b). These figures were the main cornerstones on which the widely cited chronological frameworks of Vilaplana (1983a,b), Bordonau et al., (1992) and Bordonau (1992) were built. The glacial maximum was placed between 70-50 ka BP by Vilaplana (1983a,b), around 60 ka BP by Bordonau et al. (1992), and was later shifted to 50-45 ka BP by Bordonau (1992) to include the arguments raised by Montserrat (1992). According to Montserrat (1992), independent theoretical criteria suggest that the optimum conditions for the maximum advance of Pyrenean glaciers during the Last Glaciation took place between 50-45 ka BP. This author pointed out that during this period, ablation could have been reduced in the Pyrenees because of a minimum in summer radiation (Berger, 1978), and an increased vertical temperature gradient determined by low CO2 concentration values in the atmosphere (Barnola et al., 1987). The contrast between the low latitude radiation minimum and the high latitude radiation maximum would have produced a reinforced latent heat flow to the Pyrenean latitudes and an increased precipitation (Montserrat, 1992). A number of additional glaciolacustrine successions across the Pyrenees have yielded radiocarbon dates older than 20 ky BP, as summarized in Tables 1 and 2. These dates are generally accepted as providing minimum ages for the start of the deglaciation and are often quoted as proof of an early Pyrenean glacial maximum (e.g. Andrieu et al. 1988;García Ruiz et al., 2003;Calvet, 2004).
The validity of the oldest radiocarbon dates from glaciolacustrine successions was first questioned by Turner and Hannon (1988), who were reluctant to accept a marked climatic asynchroneity between the Pyrenees and northern Europe. They considered that the oldest part of the palynological succession presented by Mardones and Jalut (1983) was not consistent with the regional evidence and rejected the dates ranging between 29,500 and 38,400 yr BP.
According to Turner and Hannon (1988) the basal sediments at Biscaye and probably other sites (see Table 1) were likely to be aged by a hard water effect and/or by presence of old reworked organic matter. Jalut et al. (1992) accepted that a hard water effect was likely, but argued that it could only be responsible for an ageing of less than 1.5 ka. However, these authors did not fully discuss the possible ageing due to the mineral carbon effect or the possible input of organic matter reworked from older sediments, and they considered the original dates of Mardones and Jalut (1983) to be roughly correct.
The first attempt to quantify the possible ageing of radiocarbon dates from the Pyrenean deposits was done in the glaciolacustrine succession of Llestui (Noguera Ribagorçana Valley, southern Pyrenees), which had initially yielded 14 C ages between ≥ 34,000 and ≥ 31,500 14 C yr BP (Vilaplana, 1983a,b). Bordonau et al. (1993) resampled the same succession to attempt a separation of the mineral and organic carbon components present in bulk sediment samples.
They measured the activity of the CO2 released by igniting the samples in two steps, at c. 400ºC and at c. 900ºC. The step at higher temperature should correspond to oxidation of the mineral component and yielded ages about 40,000 yr BP. In contrast, the step at 400ºC, which must roughly correspond to the oxidation of the organic component, provided two new dates of 18.2±0.6 14 C ka BP and 21.6±0.9 14 C ka BP (Table 1). Despite the fact that these corrected dates of Bordonau et al. (1993) show age inversion and that they display slightly high  13 C values, they must be closer to the true ages of the Llestui succession than the former >30 14 C ka BP measured by Vilaplana (1983a,b).
Palynological and dating studies also indicate that 14 C dates obtained from glaciolacustrine successions in the Pyrenees must be taken with caution. Reille and Lowe (1993), and Reille and Andrieu (1995) show that radiocarbon dates from lacustrine sediments are often inconsistent with the biostratigraphical correlations between neighbouring sites found at similar altitudes.
Hence, in agreement with Turner and Hannon (1988), Reille and Andrieu (1995, page 18) conclude that the ages between 38,400 and 31,900 14 C ka BP for the basal succession of Biscaye are too old, and suggest an alternative age of less than 25 ka BP. Redating of the same succession by Reille and Andrieu (1995) provided an age of 12,480 14 C yr BP for sediments found at greater depth than those dated by Mardones and Jalut (1983) as 29,500 14 C yr BP (see Table 1). In addition, the same authors suggest that most of the dates given by Andrieu (1987Andrieu ( , 1991 and Jalut et al. (1988) from the successions of Barbazan, Estarrès and Castet may be considered too old (Table 1).
More recently, additional dates from glaciolacustrine sediments and bog deposits in the southern Pyrenees have been made available. Some of these are AMS 14 C dates from pollen concentrates which, in principle, should not be affected by the mineral carbon ageing. The oldest date comes from the glaciolacustrine juxtaglacial sediments of Linás de Broto that yielded an age of 30,380 14 C yr BP . This succession was emplaced when large lateral moraines were being built and when the Ara Valley must have been extensively glaciated (Last Pleniglacial), with an ice tongue about 30 km long (Martí Bono and García Ruiz, 1993;Serrano and Martínez de Pisón, 1994;Peña Monné et al., 1997). Another interesting date from the neighbouring valley to the west (Gállego Valley, Fig. 1) is taken from the basal sediments of the El Portalet peat bog, which yielded an age of 28,300 14 C yr BP (García Ruiz et al., 2003;González-Sampériz et al., 2004, in press 2006. This site is located at the bottom of a glacial cirque and the basal sediments are interpreted as the result of deposition in a proglacial lake. Since the site is at high altitude and close to the glacial divide, the date obtained was interpreted by García Ruiz et al. (2003) as evidence of an early age for deglaciation. However, according to González-Sampériz et al., (2004;in press 2006), the sedimentological and palynological record in the El Portalet succession suggests a later glacier readvance overriding this site around the LGM. Bakalowicz et al. (1984) interpreted that clastic deposition in the Niaux-Lombrives caves (Ariege Valley) took place when the entrances of the cave system were glaciated. Based on U-Th dating of speleothems, these authors established that the last period of clastic deposition, and hence local glaciation took place between 90 and 19 ka. This means that the Ariege Valley must have been extensively glaciated with a >40 km-long glacier during the whole period. This glacier extent corresponds to >75% of the maximum Ariege glacier length suggested for the last glacial cycle (Taillefer, 1985).
Preliminary optically stimulated luminescence (OSL) dates by Sancho et al. (2002) suggest a correlation between the lowermost till outcrop in the Cinca Valley (2 dates yielding 49.5±3.9 and 62.7±6.2 ka) and a fluvial terrace (6 dates from different localities ranging from 53.8±3.8 to 65.1±4.5 ka). According to these authors, the OSL dates suggest a glacial outwash origin for the dated terrace and an age greater than 50 ka for the maximum glacial extent. Similarly, Peña Monné et al. (2004) sampled proximal and distal glaciofluvial deposits related to the lowermost till outcrop in the Gállego Valley, yielding OSL dates of 85±5 and 67±8 ka, respectively. In addition, for waterlain sediments included in a well preserved frontal moraine located c. 2 km upvalley, they provide two OSL ages of 35±3 and 36±2 ka. Although all published OSL dates from the southern Pyrenees may correspond to depositional ages, no detailed discussion is yet available about the possibility of partial bleaching prior to deposition, and an ageing effect cannot be completely ruled out.

The Deglaciation
Separated by more than 10 km from the Last Pleniglacial features, minor recessional moraines or push moraines are commonly found at increasing heights along the valley axes.
An early "valley glacier phase" was distinguished by Bordonau et al. (1992) and Bordonau (1992) including two deposits (the Seminari de Vilaller frontal moraine in the Noguera Ribagorçana Valley, and La Masana glaciolacustrine complex in the Valira Valley), which had previously been attributed to a "post maximum stabilisation" by Vilaplana (1983bVilaplana ( , 1985. Bordonau (1992) tentatively assigned an age >26,000 14 C yr BP to his "valley glacier phase" by correlating it with the Barbazan-Coumanié-Labroquère glacial episode (Garonne Valley, northern Pyrenees). The age is based on the interpretations and near-basal date of the proglacial glacio-lacustrine sequence at Barbazan obtained by Andrieu et al. (1988), and Andrieu (1989Andrieu ( , 1991. A much more distinctive and widespread feature of the deglaciation is a group of tills and moraines restricted to high mountain areas (Taillefer, 1969;Martí Bono and García Ruiz, 1994;Peña Monné et al., 1997;García Ruiz, et al., 2003). They are clearly distinguishable from the Last Pleniglacial moraines because they are found tens of km upvalley, and are morphologically fresher. These moraines formed after disconnection of the main glacial tributary system, and result from stabilisation or readvance of single ice tongues measuring several km in length. The period of deposition of these moraines has been referred to as the "disjunction stage" by Taillefer (1969), and as the "mountain glacier phase" by Mardones and Jalut (1983). The latter authors associated this phase with a cold period, according to their interpretation of the palynological and sedimentary record of the glaciolacustrine sequence in Biscaye (northern Pyrenees). They assigned an age between c. 24,000 and 13,250 14 C yr BP to this glacial phase based on interpolation of their single bulk sediment radiocarbon dates (Table 1). Partly based on the same data and following a similar terminology, Bordonau (1992) distinguished an "altitude glacier phase" which comprises from older to younger, an "altitude valley glacier episode" and a "cirque glacier episode". Following the interpretations and using the dates from the north-Pyrenean glaciolacustrine and bog successions of Freichinede, Balcère and Laurenti studied by Jalut et al. (1982) and Andrieu et al. (1988), Bordonau (1992) considered that the glacial fronts had already receded to high mountain areas by 16 14 C ka BP. Hence, he suggested an age of about 16 to 13 14 C ka BP for his "altitude glacier phase". Similarly, Reille and Lowe (1993, p. 74), conclude that the upper valleys in the Pyrenees were mostly deglaciated prior to 15 14 C ka BP. A minimum age for deglaciation of some cirque areas c. 13.5 14 C ka BP is deduced from lake Redó d'Aigüestortes in the Tor Valley (Copons and Bordonau, 1996), and c. 16.5 14 C ka BP from la Grave peat bog in the Têt Valley (Delmas, 2005) (Table 2).
Cirque-floors are commonly occupied by conspicuous rockglaciers Serrano and Agudo, 1998). Rockglaciers are especially abundant towards the eastern Pyrenees but are present all along the mountain range. Serrat (1979) argued that the rockglacier dynamics would have been favoured by a period of cold and dry climatic conditions and suggested that there would have been a period of preferred development of rockglaciers towards the end or after the main deglaciation. He acknowledged the coexistence of cirque glaciers during the development of rockglaciers but suggested that this major period of rockglacier development could be correlated with the cold and dry conditions of the Lateglacial period, as deduced from palynological records (e.g. Jalut, 1974), and from sedimentological arguments (Vilaplana et al., 1983(Vilaplana et al., , 1989. Based on these considerations, Bordonau (1992) formally defined a "rockglacier phase" taking place after the deglaciation. He suggested that the correlation made by Serrat (1979) could be restricted to the Younger Dryas, and assigned an age of 11,000 to 10,000 14 C yr BP to the "rockglacier phase". Such restrictive chronology for the major phase of rockglacier activity was rejected by García Ruíz and Martí , and Serrano and Martínez de Pisón (1994). They consider that during a general shift from glacial to periglacial dynamics, rockglaciers can activate asynchronously depending on local factors.
Hence, they suggest that rockglaciers in different areas may have been active at different times since the deglaciation to the Holocene, including reactivations during cold historical fluctuations. These ideas are in agreement with the observation of present-day active rockglaciers Serrano and Agudo, 1998;Chueca et al. 1994;Chueca and Julián, 2005;Serrano et al., 1991Serrano et al., , 1999Serrano et al., , 2006.

The Neoglacial
Apart from rockglaciers, most cirques display single or several glacial moraines. Taillefer (1969) distinguished a "neoglacial stage" during which cirque moraines would have been formed after a period of complete or almost complete deglaciation. In contrast, Bordonau (1992) interpreted most of the cirque moraines as part of the main deglaciation, and included them in his late "altitude glacier phase" (the so called "cirque glacier episode"). However, he implicitly recognised that part of them could have been generated in later (Holocene) periods. Although geomorphic data do not allow a clear separation of Lateglacial and Holocene glacial features, it is generally assumed that some of the cirque moraines were deposited during the Holocene (Taillefer, 1969;Bordonau et al., 1992;García Ruiz and Martí Bono, 1994;Serrano and Martínez de Pisón, 1994). Existence of early Holocene glacial advances is supported by Gellatly et al. (1992), who inferred glacial phases by conventional 14 C dating of a bog succession located between two morainic arcs in the Cirque de Troumouse (northern Pyrenees, Table 2). They distinguished a probable Holocene phase which occurred earlier than 5,190 14 C yr BP, and another phase of moraine construction, which they correlated with a shift towards inorganic sedimentation in the bog between 5,190 and 4,654 14 C yr BP. According to Gellatly et al. (1992), there have been at least two further Holocene advances involving moraine construction. Julián and Chueca (2002) also indicate some morainic remnants found at high altitude in cirque areas which they attribute to pre-Little Ice Age times (perhaps corresponding to the Lateglacial) based on lichenometric dating.
In addition, historical glacial phases have been documented. The few present-day glaciers in the Central Pyrenees are bordered by a series of fresh morainic ridges. According to descriptions and engravings made during the 18 th and 19 th centuries (Martínez de Pisón, 1989;Martínez de Pisón and Arenillas, 1988), and according to lichenometric dating (Chueca and Julián, 1996), these moraines mark the location of the glacial margins during historical times.
Together with other equally fresh moraines found in completely deglaciated valleys, these were attributed to the Little Ice Age (Martí Bono et al., 1978;Martínez de Pisón, 1989;Bordonau et al., 1992;Martí Bono and García-Ruiz, 1994). Copons and Bordonau (1994) described two moraine generations in the Maladeta Massif (Central Pyrenees), which indicate two episodes of glacier advance in the Little Ice Age. Based on correlation with the fluctuations established in the Alps, on dendroclimatogical data from the central Pyrenees, and on a discussion on the possible causes of the Little Ice Age, the first episode was considered to have occurred during the 18 th century (probably coinciding with the Maunder Minimum), while the second episode was assigned to the 1820s, with a possible causal link with the Tambora volcanic eruption (Copons and Bordonau, 1994). Similarly, based on lichenometric dating, Chueca and Julián (1996) identified two phases of glacier readvance attributable to the Little Ice Age (1600-1620 AD and 1820-1830 AD).

The glacial record in the Upper Noguera Ribagorçana Valley
The study area is located in the UNRV between 3,240 and 900 m altitude (Fig. 2). The geological, geomorphological and sedimentological record of the UNRV was studied in detail by Vilaplana (1983a,b), Vilaplana et al. (1989) and Vilaplana and Bordonau (1989). In this section we summarise the most relevant data on glacial reconstruction, with especial emphasis on the geomorphological features suitable for surface cosmic ray exposure dating. The information included in this section is taken from Vilaplana (1983a) and from additional aerial photointerpretation and fieldwork carried out in 2003.
Paleozoic granodiorites of the Maladeta batholith crop out in the northern UNRV (Fig. 2).
South of the Maladeta batholith, bedrock is largely composed of quartz-poor sedimentary paleozoic units including limestones, slates and sandy slates (Mey, 1965). However, granodioritic blocks are pervasive in tills and moraines cropping out in the study area, both in areas of granodioritic and sedimentary bedrock.
About 9 km south of Vilaller the valley is relatively narrow and shows no signs of glacial erosion. Associated with a widening of the Noguera Ribagorçana Valley at the convergence with the Noguera de Tor Valley there are a few terrace remnants at c. 15-20 m above the present-day talweg (Vilaplana, 1983a). Some of these deposits have been interpreted as glaciofluvial in origin, but since they have no contact with glacial sediments, they can hardly be situated in the relative chronological framework (Vilaplana, 1983a,b).
The lowest till outcrops in the valley are found about two km south of Vilaller (c. 940 m a.s.l., Fig. 2). They correspond to the Sant Antoni subglacial till, whose surface displays abundant rounded to subangular granodiorite boulders of metric size (Vilaplana, 1983a,b). The present day Noguera Ribagorçana river incised this deposit by c. 20 m. These subglacial tills were deposited when the Noguera Ribagorçana glacier was more than 30 km-long, close to its assumed maximum extent (Vilaplana, 1983a,b).
The volumetrically most important glacial deposits in the UNRV correspond to a left lateral moraine whose remnants and associated juxtaglacial obturation terraces can be traced discontinuously for >10 km along the eastern valley flank ( (Vilaplana, 1983a,b;Vilaplana and Bordonau, 1989). The glaciolacustrine rhythmites of Llestui were 14 C dated by Vilaplana (1983a,b), and the dates were later corrected by Bordonau et al. (1993) to c. 18-20 ka BP (see the present paper section 3.1., for details). Bordonau (1992) attributed the lateral moraines of Artigalonga and Llestui together with their associated juxtaglacial deposits and terraces (termed the Artigalonga-Llestui moraine system in the present paper) to the "post-maximum stabilisation phase" (Fig. 4).
About 2-3 km north of Vilaller, and at slightly higher (up to c. 20 m) elevations than the Artigalonga-Llestui moraine system, the Noguera Ribagorçana Valley-flanks display sparse granodioritic erratic boulders. These erratics mark an earlier period of slightly more extensive glaciation than the period of deposition of the Artigalonga-Llesui moraine system.
Despite the large deposits of the Artigalonga-Llestui moraine system, there is no frontal moraine which can clearly be considered a correlative. The only possible candidate is the Seminari de Vilaller moraine (Vilaplana, 1983a,b) located 1 km north of Vilaller and at c. 990 m a.s.l. (Fig. 2). This is a till deposit which clearly forms a frontal moraine rising about 8 m over the Noguera Ribagorçana talweg. Vilaplana (1983a,b) and Bordonau (1992) considered that the altitudinal contrast between this frontal moraine and the lowest remnants of the Artigalonga-Llestui moraine system is too strong to allow correlation. This was the reason for Bordonau (1992) to attribute this morainic frontal arc to his "valley glacier phase", considered to correspond to a slight readvance during the deglaciation after the "post-maximum stabilisation phase" (Fig. 4).
North of Seminari de Vilaller moraine, no glacial deposits but few glacial erosion surfaces are observed for 15 km along the valley axis. The best preserved roches moutonnées on granodiorite are found at Refugi de l'Hospitalet, about 1,460 m a.s.l. (Fig. 2).
About 1,530 m a.s.l. and close to the UNRV talweg, there is a 0.5 km-wide group of morainic remnants which are termed in the present paper the Santet moraines (referred to as the Barranc de Bessiberri deposits by Vilaplana, 1983b) (Fig. 2). Roadcuts show that these moraines are made up of sandy till, with abundant granodioritic boulders that protrude conspicuously at the surface. The gently undulated terrain of the Outer Santet moraines allows the distinction of up to three discontinuous subdued ridges, which can be interpreted as deposited by a receding ice front. In contrast, the more prominent Inner Santet moraine (Fig. 5) rises 10-15 m over the valley floor and can be interpreted as the result of a longer stabilisation or a glacial readvance, when the ice tongue of Mulleres (and maybe Conangles) was c. 6.5 kmlong. Upwards (c. 1,5 km) along the valley axis, at about 1,600 m a.s.l. there are two minor tightly-spaced left lateral moraines which are known as the Hospital moraines. The prolongation of these moraines towards the valley bottom indicates that they must be younger than the Santet moraines. Accordingly, they indicate two further small ice stabilisations or readvances. The Santet and Hospital moraines are included in the "altitude valley glacier episode" (early "altitude glacier phase", Fig. 4) distinguished by Bordonau (1992).
At about 1,720 m a.s.l. in the Mulleres Valley axis, there is a protruding granodioritic roche moutonnée (Fig. 6a) which must have been deglaciated before the deposition of a bouldery moraine located a few meters upvalley (Fig. 2). This deposit, termed the Mulleres moraine, can be ascribed to a minor period of ice stabilisation or readvance when the Mulleres ice-tongue was c. 4 km-long.
At about 2,200 and 2,300 m a.s.l. on the Pleta Naua slope, there are three arcuated moraines referred to as Pleta Naua moraines (Fig. 2, Fig. 6b). Situated on an open slope oriented to the SE, they may have been deposited by small (< 2 km 2 ) niche glaciers disconnected from the main Mulleres ice-tongue. Similar moraines are perched on the opposite southern flanks of the Mulleres Valley. The outer arcs were assigned to the "altitude valley glaciers episode" (early "altitude glacier phase"), whereas the inner moraines were attributed to the late "altitude glacier phase" (the "cirque glacier episode", Fig. 4) by Bordonau (1992).
In addition, the opposite (southern) flank of the Mulleres Valley also shows a single morainic arc oriented north at c. 2,460 m a.s.l., termed the Tallada moraine (Figs. 2, 6b). In contrast with the Outer and Inner Pleta Naua moraines, it shows a sharp moraine crest and its boulders display a low degree of alteration and lichen growth. The moraine surrounds a closed depression which, entrenched between two peaks, holds a permanent snow field. Despite the contrasting freshness of this morainic arc Bordonau (1992) also assigned it to his early "altitude glacier phase" (the "altitude valley glaciers episode", Fig. 4). Equivalent features associated with permanent snow fields are also present at the head of Salenques Valley, interpreted by Chueca et al. (2002) to be formed during the Little Ice Age. Approximately midway between these two erosion surfaces and c. 2,000 m a.s.l., the upper Bessiberri Valley displays a right fronto-lateral moraine remnant which must have been deposited during a glacier stillstand or readvance when the Bessiberri glacier was about 2 kmlong. At about 2,700 m a.s.l. there is a set of moraines referred to as Llastres de Bessiberri moraines (Figs. 2, 6c). Despite being topographically higher than the Pleta Naua moraines, they show a similar alteration pattern and morphology. South of Llastres de Bessiberri, there is an active rock glacier Chueca et al., 1994;Serrano and Agudo, 1998;Chueca and Julián, 2005), which is interepreted to derive from burial of a Little Ice Age glacier . This interpretation is consistent with lichenometric dating of limiting moraines (Chueca et al., 1994).

Methods
Granodioritic glacial erosion surfaces and moraine boulders were selected for 10 Be surface exposure dating to cover the widest possible range of glacial phases (Table 2). To minimise the risks of shielding by till, soil cover, or snow, we chose the highest and most prominent roches moutonnées located far from areas covered by till. Similarly, in an attempt to avoid underexposure caused by block rotation or till cover, only boulders well anchored in till were selected. From these, boulders protruding more than 1 m over surrounding till were favoured.
We selected up to three samples per moraine whenever suitable blocks were available. Only the two upper centimetres from selected rock surfaces were sampled.
Samples were crushed and sieved to extract the 250-1000 m granulometric quartz fraction.
Minerals other than quartz were dissolved by HCl and H2SiF6, and remaining quartz grains were cleaned using sequential HF dissolutions to remove any potential atmospheric 10 Be (Brown et al., 1991;Kohl and Nishiizumi, 1992;Cerling and Craig, 1994). Between 10 to 30 g per sample of clean quartz cores were then completely dissolved in HF and spiked with 300 g of 9 Be carrier (Bourlès, 1988;Brown et al., 1992). Beryllium was separated by successive solvent extractions and alkaline precipitations (Bourlès, 1988;Brown et al., 1992). Measurements of (Raisbeck et al., 1994). AMS measurements were performed with reference to NIST Standard Reference Material 4325 using the certified 10 Be to 9 Be ratio of (26.81.4)  10 -12 . Because 10 Be production rates are all referred to ICN standards, we normalized the measured concentrations based on the NIST standard to ICN standards by increasing them by a factor of 1.143, according to Middleton et al. (1993). Analytical uncertainties (reported as 1) were calculated based on counting statistics, a conservative estimate of 3% instrumental variability and a 50% uncertainty in the chemical blank correction. See Table 3 for additional details.
Among the slightly different 10 Be production rates available in the literature, the one deduced from the Köfels calibration site (Kubik et al., 1998;Kubik and Ivy-Ochs, 2004) is undeniably the most suitable when considering similarities in altitude, geographic location, and exposure age-range with respect to the study area. The re-evaluated sea level high latitude (SLHL) production rate of 5.44 10 Be at g -1 y -1 quartz reported by Kubik and Ivy-Ochs (2004) was based on the nucleonic component scaling of Lal (1991), with no latitude scaling for the muonic component. Based on 1) the original 10 Be concentration data and the topographic and environmental correction factors reported in Kubik and Ivy-Ocks (2004 , Tables 1 and 2, samples KOE4 to KOE101) for the calibration site of Köfels, 2) the attenuation lengths and muonic contributions of Braucher et al. (2003), and 3) the widely used scaling method of Stone (2000), we recalculated a SLHL production of 5.410.17 at g -1 y -1 quartz. Using this "corrected" Köfels-derived SLHL 10 Be production value, we applied the scaling method of Stone (2000) to deduce the exposure ages reported in Figs 2, 7 and 8. The limitations of these and other recently developed scaling criteria (Dunai, 2000;Stone 2000) were reviewed by Desilets and Zreda (2003), who provide a more reliable scaling procedure based on mass shielding depth (instead of altitude) and effective vertical cutoff rigidity (instead of geomagnetic or geographic latitude).
Building on the approach of Desilets and Zreda (2003), Pigati and Lifton (2004) provide an altitude-latitude scaling method that takes into account geomagnetic intensity variations and polar wander, and their time-integrated effects on production. Table 3 includes the exposure ages calculated for the Noguera Ribagorçana samples based on both Stone (2000) and Pigati and Lifton (2004), showing that results from these two scaling methods only differ by a maximum of +7.5/-6.4 %, depending on altitude.
Even though uncertainty in the scaling models is difficult to assess, realistic errors for the nucleonic component attenuation length may be > 10% of the widely used scaling polynomials of Lal (1991), which are also implicit in Stone (2000) polynomials, and in the range of +5/-2% in the case of Desilets and Zreda (2003) scaling, implicit in Pigati and Lifton (2004). As no corrections for erosion, sample thickness or irradiation geometry are applied, the calculated exposure ages for the UNRV must be seen as minimum estimates. However, the maximum deviation between calculated and true exposure ages in the most unfavourable cases is estimated as < 9% (c. 2.7% corresponding to a mean erosion of 1.5 mm ka -1 on granodiorite, c. 1.7% corresponding to 2 cm sample thickness, and c. 4.5% corresponding to the difference between irradiation on a flat geometry vs. the geometries of sampled surfaces based on Masarik and Wieler, 2003). Based on present day seasonal distribution of snow in the sampled sites, the effect of snow cover is assumed to be negligible, at least for the Holocene.

Results
Dating of the relevant moraine groups and erosion surfaces were performed to provide a chronology of the UNRV deglaciation. Twenty-five samples corresponding to 18 blocks from 9 different moraines, 2 erratic blocks, and 5 erosion surfaces were processed. Despite the fact that the sampling was designed to minimise the likelihood of selecting previously rotated or till covered blocks, some 10 Be dates are incoherent when compared with dates from the same moraine or dates from neighbouring sites (Table 3, Fig. 7). Such aberrant cases were avoided by withdrawing 1) one date (ART01) differing by more than two standard deviations with respect to the older dates from the same moraine, and 2) dates inconsistent with the geomorphologically-based relative chronology (see Fig. 7 and the following paragraphs for details). The minimum exposure age of each moraine was then deduced by calculating the error weighted mean of the remaining dates. The resulting 10 Be exposure ages obtained from the UNRV are summarized in Figs. 2 and 8.
According to their location, the Sant Antoni till and the Seminari de Vilaller moraine must be among the oldest glacial deposits in the UNRV. However, when compared with exposure ages from features located upvalley, the exposure ages of 12.31.8 ka (STA01) and 6.11.3 ka (SMV01) are too young and can be attributed to erosion of till cover or to block rotation. Hence, these two key localities did not yield dates that can be interpreted as ages of deposition or deglaciation. 10 Be data from boulders on the Artigalonga moraine provide exposure ages of 18.62.2 ka (ART02), 16.63.8 ka (ART03) and 4.41.0 ka (ART01). The last one is an obvious outlier, whose underexposure may be due to block rotation. The two older dates indicate an exposure age of 18.11.9 ka (weighted mean) for the Artigalonga-Llestui moraine system, corresponding to a minimum age for the culmination of the major phase of moraine building in the UNRV (Fig. 3, 7 and 8). In line with the geomorphological evidence, an older exposure age of 21.34.4 ka is calculated for the Tinabre erratic (TIN01), which must have been deposited before the Artigalonga-Llestui moraine system, when the UNRV glacier near Tinabre was about 20 m thicker (Fig. 2). Given that significant erosion or rotation of blocks cannot be ruled out, the Artigalonga and Tinabre exposure ages must be interpreted as minimum ages of deposition.
The Refugi de l'Hospitalet erosion surfaces are found on a moderate slope which minimises the chance of till cover and, hence, the 10 Be exposure ages obtained for RLH01 (13.23.3 ka) and RLH02 (11.71.5 ka) are considered to provide relatively accurate deglaciation ages, and minimum ages for the disconnection between the Salenques and Mulleres glaciers, with a weighted mean of 11.91.4 ka (Fig. 2). Statistically indistinguishable exposure ages for three block surfaces sampled from the Inner Santet Moraine (IST01, 14.41.4 ka; IST02, 12.31.5 ka, and IST03, 16.12.8 ka) yield a weighted mean exposure age of 13.70.9 ka. The exposure age of HOS03 (25.32.4 ka, Fig. 7), when compared with the internally consistent exposure ages of Refugi de l'Hospitalet and the Santet Moraines, must be considered too old for the age of the Hospital Moraine, indicating inheritance. The angular boulder from where HOS03 was retrieved must have been sitting on a higher ridge or rocky slope for several thousands of years before being deposited on the moraine.
The lowest erosion surfaces along the Bessiberri Valley (OBS01) provide an exposure age of 16.32.2 ka, which must correspond to a minimum age for the disconnection of the Bessiberri ice tongue from the Noguera Ribagorçana glacier. Further upvalley, another roche moutonnée (BES01) yields a consistent deglaciation age of 11.71.5 ka. Between these two erosion surfaces, there is a fronto-lateral morainic ridge at c. 2,000 m a.s.l. (Bessiberri moraine in Fig.   2) and a smaller morainic ridge found at a higher altitude. Despite the lack of suitable boulders for sampling, OBS01 and BES01 bracket the age of these moraines, providing strong support for correlation with the Santet Moraines.
These dates mark the period when the Mulleres ice tongue was about 3.5 km long, and are statistically indistinguishable from the exposure ages determined for Outer Pleta Naua moraine (weighted mean of 10.20.7 ka), and Llastres de Bessiberri moraine (weighted mean of 10.10.6 ka), suggesting that they could be formed during the same phase of moraine building.
However, given the age uncertainties these moraines could also correspond to different glacier fluctuations. The age of the Inner Pleta Naua moraine is less clear because different boulders yielded conflicting exposure ages. Block rotation of IPN01 (4.40.7 ka) or inheritance for IPN02 (11.21.6 ka) cannot be ruled out.

Sources of ageing in radiocarbon dates from glaciolacustrine successions
Next, we discuss several theoretical points that will be applied later in our discussion on the reliability of the Pyrenean 14 C dataset. As generally acknowledged, 14 C dates from limnic material may be subject to a number of errors (e.g. Sutherland, 1980;Lowe et al., 1988;Bowman, 1990;Lowe and Walker, 1997). First, a hard water error may be produced when old carbon is leached from bedrock lithologies into lake waters. Uptake of such inert carbon by organisms in the lake results in 14 C-depleted organic matter, and yields 14 C ages that are too old.
Second, mineral carbon contamination may affect sediment samples if old carbon from bedrock was introduced into the lake as fine-grained detritic sediments. The effect may be most important in areas of carbonate-rich bedrock, but it may also occur in recently deglaciated terrain where inert carbon may be released from other lithologies by glacial erosion (Sutherland, 1980). Third, organic carbon contamination may result from inclusion, in the lake sediments, of allochtonous organic matter reworked from older formations.
The fractionation laws of the different carbon isotopes provide an indirect way of evaluating to what extent dates are affected by the hard water effect or by mineral carbon contamination.  13 C values higher than the normal values found in limnic organisms (i.e. > -25 3 ‰, Bowman, 1990;Lowe and Walker, 1997) indicate a hard water error in the case of AMS samples, and/or a possible mineral carbon contamination in the case of bulk sediment samples. In contrast, detecting contamination by organic carbon is a more complex matter: its effect can only be estimated by biostratigraphical comparison, detailed taphonomical study of the microfossil assemblage, or by independent AMS dating of the different organic fractions (Nambudiri et al., 1980;Lowe et al. 1988;Nelson et al., 1988).
The magnitude of the hard water ageing may reach up to several hundred years (e.g. Andree et al., 1986;Moore et al., 1998). In contrast, the ageing produced by the mineral and/or organic carbon contamination may be much higher depending on the relative amount of allochtonous vs. autochtonous carbon component and the original concentration in 14 C of the former. When autochtonous organic production rates are small, they can be overwhelmed by even the smallest amounts of old mineral or organic allochtonous carbon, resulting in a large ageing effect (Olsson, 1979;Sutherland, 1980;Björck and Håkansson, 1982). Sparse vegetation cover and high rates of mechanical weathering may be responsible for increased erosion and enhanced allochtonous input of mineral and organic carbon into glaciolacustrine successions. These conditions are likely to be met at the start of glaciolacustrine sedimentation during the first stages of the deglaciation, when near-full glacial conditions still prevail (Sutherland, 1980). While the ageing effect may be maximum in minerogenic sediments deposited during full glacial conditions, it can be significantly reduced or negligible during later deglaciation phases, when the relative detritic input is decreased by 1) the protective effect of the vegetation cover, 2) a greater distance from the glacier front, and by 3) the increased local lacustrine organic production. The correction needed for 14 C dates from contaminated postglacial lacustrine sediments is highly variable; with reported values ranging up to 25,000 yr (Nambudiri et al, 1980).

Reliability of minimum limiting radiocarbon dates from Pyrenean glaciolacustrine successions
As shown in section 3, the deglaciation chronology in the Pyrenees has been based on a 14 C dataset obtained from glaciolacustrine, lacustrine and bog successions in a number of localities.
Next, we discuss the validity of the most relevant Pyrenean 14 C dates, based on the compilation presented in Tables 1 and 2. The oldest dates from the Pyrenean dataset (those > 20,000 14 C yr BP) correspond to glaciolacustrine sediments deposited in juxtaglacial lakes during the Last Pleniglacial, or to proglacial sediments deposited during the early stages of glacial retreat. According to the arguments given in section 7.1., dates from these successions are relatively prone to ageing due to contamination.
As shown in section 3.1., suspicion of ageing due to mineral carbon contamination was confirmed in the case of the juxtaglacial sediments of Llestui in the UNRV, where the initial dates of 34.0-31.5 14 C ka BP were corrected to c. 21-18 14 C ka BP .

Based on biostratigraphical correlation and on a new radiocarbon dataset Reille and Andrieu,
(1995) (see Table 1 and section 3.1. for details) confirmed that the basal sediments of Biscaye were affected by an important ageing effect. Despite the 38,400 14 C yr BP of Mardones and Jalut (1983), these sediments were assigned a maximum age of 25,000 14 C yr BP by Reille and Andrieu (1995). Accordingly, the dates significantly older than 25 14 C ka BP and 21 14 C ka BP for the Llestui and Biscaye successions respectively can be considered to be unreliable (see the assessment column in Table 1).
In the northern Pyrenees other successions were emplaced during deglaciation, which yielded ages older than 20 14 C ka BP (Tables 1 and 2). Despite the fact that contamination has not been proven for any of these sediments, biostratigraphic criteria suggest that the 14 C dates from the glaciolacustrine sites in the northern Pyrenees such as Barbazan (c. 31-26 14 C ka BP), Estarrès (c. 27 14 C ka BP) and Castet (c. 25 14 C ka BP) were probably affected by significant ageing (Reille and Andrieu, 1995). In addition, the oldest dates from the Lake of Lourdes (33.7 ka BP) and Le Monge (34.5 ka BP) are unreliable given that they were obtained from tills, a kind of deposit in which the dated carbon is probably not coeval with deposition. All these dates might easily need corrections of several thousands of years and, in consequence, they can be considered as suspect (see Table 1). Similarly, in the southern Pyrenees there are several dates for which significant ageing is not proven but can be suspected. The oldest dates come from Els Bassots in the UNRV, a succession which yielded not only conflicting ages between c. 31.4 and c. 20.2 14 C ka BP (Table 1), but also age inversion (Bordonau, 1992). As suggested by the youngest date, the oldest dates from this site are probably affected by allochtonous mineral carbon input from the same bedrock units as Llestui, which is located only a few km upvalley.
Despite using minute samples, and provided that a careful selection of the organic matter is made, AMS dating is the best choice to avoid contamination from mineral carbon (Cwynar and Watts, 1989). Accordingly, the pollen concentrate AMS 14 C dates in the southern Pyrenees (Portalet, Linás de Broto, Corral de las Mulas and Tramacastilla paleolake, see Table 1 and references therein) can probably be regarded as the most reliable dates available for the early stages of the Pyrenean deglaciation. However, given the absence of detailed biostratigraphical and taphonomical studies devoted to assessing the accuracy of these dates, organic contamination cannot be ruled out. Reille and Lowe (1993) have suggested that a portion of Pinus, Artemisia and Ephedra pollen grains found in the lateglacial successions of the Pyrenees might come from far field areas such as North Africa. Even minute amounts of this allochtonous pollen could be recycled from older formations, inducing a significant ageing in the AMS 14 C dates, especially for the virtually sterile successions corresponding to the early deglaciation period. It is also worth noting that the Linás de Broto date (30,380 14 C ka BP, Table 1) corresponds to deposits coeval with extensive (Last Pleniglacial) glaciation, whereas the Portalet date (28,300 14 C ka BP, Table 1) is interpreted as a minimum age for deglaciation of cirque areas in the Gállego Valley. Considering that a difference of 2,000 years is insufficient to pass from full glacial conditions to nearly full deglaciation, there are two alternative possibilities; either 1) the Portalet date puts a minimum age for the end of glacial transfluence across the Portalet col, instead of a minimum age for the whole Gállego Valley deglaciation, or 2) some of these AMS 14 C dates are in error.
The aim of the above discussion is to show that most of the 14 C ages that have generally been taken as proof of an early deglaciation centred about 70-40 ka BP do not stand up to a rigorous analysis. Our review of 14 C dates shows that the widely cited chronology of Bordonau (1992) was mainly based on data that may now be considered unreliable or, at least, suspect (Table 1).
In addition, the correction of the Llestui succession from 34 ka BP (Vilaplana, 1983a,b) to around 18-20 ka BP  has often been overlooked in the literature. If the corrected age of the Llestui succession is accepted, the ages of the "post-maximum stabilisation" and the "valley glacier" phases of Bordonau (1992) need to be shifted towards significantly younger ages.

Suggestions for constructing a chronology of the Pyrenean deglaciation
Studies in different valleys have subdivided the glacial record according to their associated distribution of ice masses, giving rise to terms such as "disjunction stage" (Taillefer, 1969, Serrano, 1991, "valley glacier phase", "altitude glacier phase", "cirque glacier episode" (Bordonau, 1992;García Ruiz et al., 2003). However, it is recognised that morainic arcs corresponding to the same glacial pulse may have formed at different heights and distances from divides, depending on local topographic configuration, exposure and snow feeding. Some areas might have been largely or totally deglaciated when others were still occupied by long valley glaciers or by cirque glaciers. This may be especially relevant when comparing the central Pyrenean glaciers with the possibly relatively starved glaciers of the drier southeastern Pyrenees (Taillefer 1969;Calvet 2004). For these reasons we consider that locally-derived descriptive terms are inappropriate when making intervalley correlations across and along the mountain range, and when attempting the construction of a general chronological framework. Given the present knowledge, we suggest that a simple and broad division between Last Pleniglacial, Deglaciation, and Neoglacial is all that is required to easily accommodate all the local observations. Such a simple approach minimises errors of attribution between the local geological record and the glacial phases framework. As soon as new dates from different local records and different dating methods are made available, the distinction of subphases and intervalley correlations should emerge naturally.
The "pleniglacial" concept applied to the last Pyrenean glaciation corresponds to the period when the maximum or near-maximum glacial extent, and the major phase of moraine building took place. As summarized in section 2, at least three lateral moraines and a group of frontal moraines record this phase in several valleys in the northern and south-western Pyrenees. This moraine group spreads over several km along the valleys axes, and is interpreted as resulting from several periods of glacial readvance or stability. In this context, the distinction between "pre-maximum", "maximum" and "post-maximum" phases made by Bordonau (1992) may not provide a complete picture of the Late Pleniglacial period since it was deduced from observations in valleys where terminal moraines are mostly lacking, and where the glacial record is fragmentary. Taken together, the three phases defined by Bordonau (1992) can be accommodated as part of the Last Pleniglacial concept. Similarly, as summarized in section 2.2, the "valley glacier phase" of Bordonau (1992) included the record of only two localities; the Seminari de Vilaller frontal moraine (Noguera Ribagorçana Valley), which had previously been attributed to the "post-maximum phase" (Vilaplana, 1983a,b), and the Massana glaciolacustrine complex (Valira Valley), which had been ascribed to a glacial stabilisation after the glacial maximum (Vilaplana, 1985). According to our definitions, both localities can also be included in the Last Pleniglacial fluctuations rather than to subsequent deglaciation phases.

Implications for the early Pyrenean deglaciation hypothesis
The 18.11.9 ka cosmic ray exposure age of the Artigalonga-Llesui moraine system must be interpreted as a minimum age for deposition given that a rejuvenation by erosion of the morainic surface is likely. However, an additional constraint on the age of the Artigalonga-Llestui moraine system is provided by the corrected date of 21-18 14 C ka BP (equivalent to c. 25.5-21.0 cal ka BP, Reimer et al., 2004) derived from the glaciolacustrine deposits of Llestui (Bordonau et al., 1993, see Table 1 and section 2.1. for a detailed explanation). Considering that the glaciolacustrine sediments sampled by Bordonau et al. (1993) correspond to lower, and thus older, stratigraphic levels than the moraine top, the difference between the 10 Be exposure age and the actual age of the last phases of moraine building must not exceed a few thousand years.
In addition, despite its relatively large error bar, the exposure age of the Tinabre erratic (21.34.4 ka) suggests that the phases of maximum glacier extent and thickness (Last Pleniglacial conditions) recorded in the UNRV are probably younger than 25 ka.
As summarised in section 3.1., the Last Pleniglacial and most of the deglaciation in the Pyrenees are generally believed to have taken place tenths of thousands of years before the Last Glacial Maximum (LGM). As discussed in section 7.2. such a view is based on 14 C data from glaciolacustrine deposits whose accuracy is highly suspect. The corrected 14 C data of Bordonau et al. (1993) and the new 10 Be chronology presented here do not support the generally accepted hypothesis of an early deglaciation. Our chronology suggests that the Noguera Ribagorçana glacier could have been in a pleniglacial condition until the LGM, or until a few thousand years earlier.
However, the Noguera Ribagorçana chronology could be consistent with a relatively early start of the Last Pleniglacial. We recognise that the record of the Last Pleniglacial may be more fragmentary in the UNRV than in other Pyrenean valleys where several terminal moraines possibly corresponding to slightly more extensive glacial phases are preserved. Hence, the Artigalonga-Llestui moraine could correspond to the late stages of the Last Pleniglacial, whereas some of the relatively old 14 C dates from juxtaglacial deposits (Table 1) or the relatively old OSL dates (Sancho et al., 2002, Peña Monné et al., 2004, if confirmed, could correspond to its early stages. Such an explanation would allow us to reconcile all the observations, and would imply a relatively long Last Pleniglacial (e.g. between >> 30 and c. 20 ka).

Chronology of the Deglaciation in the Upper Noguera Ribagorçana Valley. Implications for Lateglacial and early Holocene glacial fluctuations in the Pyrenees
The lack of significant glacial deposits between the till outcrops corresponding to the Last Pleniglacial and the deposits from subsequent glacial phases in most Pyrenean valleys suggests fast deglaciation subsequent to the Last Pleniglacial. Although this may be the case in the Noguera Ribagorçana, the distribution of erosion surfaces suitable for 10 Be sampling does not provide a proper constraint on the deglaciation rate. However, according to dated erosion surfaces, disconnection between the Bessiberri and the Noguera Ribagorçana glaciers took place before c. 16.32.2 ka, and disconnection between the Salenques and Mulleres glaciers took place before c. 11.91.4 ka (Fig. 2). These dates are slightly younger than the former estimations of c. 16,000-13,000 14 C yr (equivalent to c. 19.3-15.2 ka cal BP, Reimer et al., 2004) for retraction of glaciers to high mountain areas in other valleys in the Pyrenees suggested by Bordonau (1992) and Reille and Lowe (1993).
The record of the Deglaciation in the UNRV reveals a series of glacial stillstands or readvances which, according to 10 Be exposure ages, occurred between c. 13.70.9 ka and c.
10.10.6 ka (Fig. 2). Within this relatively short period the Noguera Ribagorçana glacial fluctuations formed the composite Santet moraines (three subdued ridges plus a prominent one), the Hospital moraines (two minor ridges) and the Mulleres moraine (one minor ridge, 10.41.0 ka, weighted mean). The Outer Pleta Naua moraines (weighted mean of 10.20.7 ka) were formed by a small glacier that had disconnected from the Mulleres ice tongue. Also in this relatively short period, the Bessiberri glacier formed the Bessiberri moraine (possibly a correlative to the Inner Santet moraine), and some smaller morainic remnants found upvalley including the Llastres de Bessiberri moraine (weighted mean of 10.10.6 ka). Exposure ages clearly indicate that these glacial oscillations can be partly correlated with the Greenland Stadial 1 (spanning 12.65 -11.5 k GRIP yr BP, Björck et al., 1998), widely known as the Younger Dryas. Hence, according to the evidence from the Noguera Ribagorçana, the Lateglacial and perhaps the early Holocene periods can be envisaged as a dynamic period involving numerous glacial oscillations, which suggest a highly unstable climatic pattern. Exposure ages support the inclusion of the Mulleres, the Outer Pleta Naua and the Llastres de Bessiberri moraines in the Deglaciation rather than in the Neoglacial despite the fact that they may have been formed, especially the latter, during the early Holocene. 6.5 km long with its terminus at c. 1,560 m a.s.l., the Bessiberri glacier was c. 3.5 km long, its front descending to only c. 2,000 m a.s.l. Such asymmetric pattern, can be attributed to 1) predominant northwesterly winds, which produce large leeward snow accumulation on southeast-oriented slopes (Taillefer, 1982), and to 2) atmospheric temperatures higher in the late-afternoon than in the early-morning, which could lead to increased ablation of snow on westward-oriented valley heads. The UNRV data demonstrate that during the Lateglacial and perhaps the early Holocene some of the factors affecting glacial equilibrium displayed a pattern that resembled the one today. In addition, these data illustrate that geomorphological correlations based on moraine altitude between different valleys can be misleading and may result in significant errors.
The conflicting exposure ages for the Inner Pleta Naua (4.40.7 ka for IPN01 and 11.21.6 ka for IPN02) can be interpreted in the following opposite ways: IPN02 provides roughly true ages of deposition, while IPN01 is underexposed due to block rotation, or IPN01 provides roughly true ages of deposition, while IPN02 is affected by inheritance. The first option would yield an age of deposition akin to that of the Outer Pleta Naua moraine, which seems to be consistent with the apparently similar weathering aspect of both moraines. However, the occurrence of several glacial pulses at c. 10 ka is not reflected in the probably correlative Llastres de Bessiberri, lending some support to the second interpretation. If the exposure age of IPN01 is interpreted as the age of deposition, the Inner Pleta Naua moraine could be correlative with some of the moraines estimated to be Holocene (between 5,190 and 4,654 14 C yr BP) by Gellatly et al. (1992) in Lac des Aires (Cirque de Troumouse, northern Pyrenees, Fig. 1).
However, given the marked contrast in orientation and altitude between both areas, this correlation is difficult to accept, and the paucity of exposure age data for the Inner Pleta Naua moraine prevents us from drawing a definitive conclusion on the age of this moraine. As for the Neoglacial, the most recent evidence of glacial activity in the UNRV is provided by the Tallada, and the Salenques moraines. These are extremely fresh morphologically, are located close to the present-day permanent snow fields, and can be attributed to the Little Ice Age.

Conclusions
There are some 14 C, OSL and U-Th data suggesting that the last Pyrenean glacial maximum took place earlier than the Last Glacial Maximum (LGM). However, the 14 C data supporting this hypothesis are fewer and less conclusive than generally acknowledged. Most of the 14 C dates used to support a last Pyrenean glacial maximum taking place between 70-40 ka BP can indeed be rejected, and others are suspect. In addition, some corrected 14 C dates suggest extensive glaciation at c. 20-18 ka BP (c. 25.5-21.0 cal ka BP), contradicting the generally accepted hypothesis of largely deglaciated valleys during the LGM. We thus consider that the chronological framework of the Pyrenean deglaciation should be built with a critical and restrictive approach, involving several dating methodologies, such as 14 C, U-Th, luminescence, and cosmic ray exposure dating. It is of paramount importance however that any new dating should be provided with all the necessary information to enable an evaluation of its validity and significance in terms of deglaciation pattern.
In some cases, the deglaciation phases previously defined in the Pyrenees have been used for correlation purposes as if they were chronological units. This practice should be avoided, especially when comparing the glacial records of distant areas. We suggest the use of a simplified and flexible relative chronological framework based on the concepts of Last Pleniglacial, Deglaciation and Neoglacial phases. The "pre-maximum stabilisation", the "maximum", the "post-maximum stabilisation", and the "valley glacier" phases of Bordonau (1992) can all be included into the Last Pleniglacial concept (Fig. 4). As indicated by the glacial record of several valleys in the northern and southern Pyrenees, the Last Pleniglacial includes a minimum of three glacier fluctuations that took place when the Pyrenean glaciers were close to their maximum extent. This period was responsible for the largest phase of moraine building recorded in the Pyrenees.
The UNRV glacial record shows three distinctive periods of glacial deposition: First, the Last Pleniglacial is represented by perched erratics (e.g. the Tinabre erratic) and the largest phase of moraine building and juxtaglacial deposition in the UNRV (the Artigalonga-Llestui moraine system). The combination of new 10 Be exposure ages and corrected 14 C dates available in the literature indicates that these features were probably formed after c. 25 ka. Second, the Deglaciation period in the UNRV is represented by a series of moraines appearing in different tributary valleys (the Outer and Inner Santet, Hospital, Mulleres, Outer Pleta Naua, Bessiberri, and Llastres de Bessiberri moraines). According to 10 Be exposure data, most of these Deglaciation moraines are restricted to the short period ranging between c. 13.70.9 and c. 10.10.6 ka. Third, the Neoglacial is represented by a few fresher undated moraines (Salenques and Tallada moraines), which were probably built during the Little Ice Age.
According to published corrected 14 C dates and new 10 Be exposure ages, the major phase of moraine building in the UNRV (responsible for deposition of the Artigalonga-Llestui moraine system during the Last Pleniglacial) could have been synchronous with the global LGM, or could have slightly predated it. Thus, the data available does not lend support to the generally accepted hypothesis of a very early (70-30 ka BP) deglaciation of the Pyrenees with respect to the global LGM. However, we do not exclude the possibility that the maximum glacier extent during the last glacial cycle had taken place much earlier than the LGM. A simple way of reconciling the relatively young ages of the UNRV glacial features with the older 14 C data (e.g. the 30 14 C ka BP date from Linás de Broto glaciolacustrine succession, Table 1) and the recently published OSL ages (c. 85 to 35 OSL ka from glacial and glacio-fluvial sediments in the Cinca and Gállego valleys), if confirmed, would be to consider that pleniglacial conditions could have taken place during a longer period than generally assumed. Hence, the data available suggest a long Last Pleniglacial (e.g. >>30-20 ka), which would include several glacier fluctuations recorded irregularly in different valleys, with a last major glacier readvance taking place around the LGM. There is little information at present on the amplitude and duration of retreating phases separating different pleniglacial advances. Some evidence has been provided by U-Th dating of cave sediments, which indicates that between 90 and 19 ka the length of the Ariege glacier would have been maintained at more than 75% of its assumed maximum length for the last glacial cycle. Further geochronological work in other Pyrenean valleys is needed to test the validity of such interpretations.
The earlier stillstands or readvances that built the morainic ridges during the Deglaciation between c. 13.70.9 ka and c. 10.10.6 ka correlate well with the Greenland Stadial 1, generally referred to as the Younger Dryas. Exposure ages suggest that the last fluctuations of the Deglaciation in the UNRV might have extended into the early Holocene. The numerous glacial fluctuations during the Deglaciation suggest a highly unstable climatic pattern for the Last Termination and, perhaps, the early Holocene in the Pyrenees. 10 Be surface exposure dating allows correlation of moraines between tributary valleys in the UNRV. This reveals an asymmetric distribution of ice masses during the Lateglacial and perhaps the early Holocene, demonstrating that altitudinal correlation of moraines between different valleys is unreliable. support to the RISKNAT group (SGR2001/ 00081), and subsidized the facilities of the new Cosmogenic Isotope Laboratory held at the Universitat de Barcelona (project 2003PIRA 00256). We are indebted to the staff of the AMS-Tandetron facility at Gif-s-Yvette for technical assistance during 10 Be measurements. We thank Dr. Marc Calvet and an anonymous reviewer whose constructive criticism enabled us to significantly improve an earlier version of our manuscript.           Table 3. Error bars are 1 standard deviation. Dotted dates are consistent with the geomorphological relative chronology and are considered to yield actual exposure ages. Crossed dates provide too old exposure ages, are inconsistent with the relative chronology, and are interpreted as affected by inheritance.
Squared dates yield too young ages, are inconsistent with the geomorphological relative chronology and are interpreted as affected by block rotation or removal of till cover. IPN01 and IPN02 are consistent with the relative chronology but yield conflicting exposure ages relative to each other. See detailed data analysis in text.  Pallàs et al. Fig. 3 Bessiberri m.
Hospital m.