The stratigraphy of lake Endletvatn on northern Andøya, northern Norway, has been revisited to improve the understanding of the palaeoenvironment in the region during the Last Glacial Maximum (LGM). Four high-quality cores were analysed with respect to various lithological parameters and macrofossil content, supplemented by 47 AMS radiocarbon dates. The sediments indicate a low-energy environment with a mean sedimentation rate of 0.5 mm a−1. We infer perennially frozen ground in the surroundings during the LGM. Climate proxies indicate a high arctic climate (i.e. July mean temperatures between 0 and 3°C) throughout most of the LGM. The warmest periods are marked by a rise in seed, moss and animal fossils, and often also by higher organic production in the lake. These periods took place from 21.4 to 20.1, from 18.8 to 18.1, around 17 and from 16.4 cal. ka BP onwards. The shifts between the different climatic regimes occurred rapidly – probably during one or two decades. The present data do not support recently published conclusions stating that Picea, Pinus and Betula pubescens grew on Andøya during parts of the LGM. The highest relative sea level after the final deglaciation on northern Andøya is bracketed between 36 and 38 m a.s.l. It occurred between 21.0 and 20.3 cal. ka BP, peaking around 20.7 cal. ka BP. The final deglaciation of the northern tip of Andøya occurred 22.2 cal. ka BP. Then the western margin of the Andfjorden ice stream receded to the Kjølhaugen Moraine and shortly thereafter to the Endleten Moraine. Our research confirms that northern Andøya is a key location for understanding the natural environment in northwestern Europe during the LGM.
The oldest onshore postglacial sediments in Norway are found on the northern tip of the island Andøya (Fig. 1). Radiocarbon dates from basal sediments from three lakes (Fig. 1C) have yielded ages between 26 and 22.2 cal. ka BP (Vorren 1978; Vorren et al. 1988; Alm 1993). Thus, analysis of this region could potentially unravel the palaeoenvironment during most of the Last Glacial Maximum (LGM). A review of the earliest work related to the deglaciation history and palaeoclimate in this region can be found in Vorren & Elvsborg (1979). Subsequently, several onshore and offshore investigations have been conducted in the Andøya–Andfjorden area aiming at elucidating the palaeoenvironment of the region, in particular the glaciation history (e.g. Vorren et al. 1983; Møller et al. 1992; Hald & Aspeli 1997; Lambeck et al. 2002, 2010; Plassen & Vorren 2002; Vorren & Plassen 2002; Nesje et al. 2007), and the history of the flora and fauna (Fjellberg 1978; Foged 1978; Vorren 1978; Vorren et al. 1988, 2009; Alm & Birks 1991; Alm 1993; Alm & Willassen 1993; Solem & Alm 1994; Vorren & Alm 1999; Kullman 2006, 2008; Elverland et al. 2007; Aarnes et al. 2012; Parducci et al. 2012).
Parducci et al. (2012) recently concluded that Picea and Pinus grew on Andøya c. 22 000, 19 200 and 17 700 cal. ka BP, based on analysis of sediment DNA, while Kullman (2006, 2006, 2008,2012), based on radiocarbon dating of a Betula tree root, concluded that Betula trees grew in the area 20 cal. ka BP.
Nesje et al. (2007) used cosmogenic surface exposure dating of perched boulders/bedrock together with the mapping of block fields and their associated clay mineralogy in order to constrain the surface geometry of the LGM ice sheet along a profile from Andøya towards the mainland. The surface exposure dating of erratics and bedrock on northern Andøya based on 10Be provided age estimates between 56 and 20 10Be ka (Nesje et al. 2007), indicating that the LGM ice sheet did not reach the mountain plateau of northern Andøya. However, these authors could not exclude past cover by non-erosive cold-based local glaciers. Exposure dates from the lowest-altitude locality in the study area, Store Æråsen (105 m a.s.l., Fig. 1C), gave ages of 36–45 10Be ka.
Lambeck et al. (2002) modelled the glacial rebound of the Scandinavian Ice Sheet. If the observed evidence of Vorren et al. (1988) from Andøya shall match the predicted values, Lambeck et al. (2002) indicated that the ice sheet had to stand at the shelf edge, and be 1000–1500 m thick in the Andøya region. Lambeck et al. (2010) indicate that the overall maximum thickness occurred somewhat earlier than c. 23 cal. ka BP.
Detailed stratigraphies and chronologies spanning the LGM have emerged from the adjoining continental slope (e.g. Dahlgren & Vorren 2003; Laberg & Vorren 2004; Rørvik et al. 2010). Studies of the seabed morphology have shown that large ice streams occupied Andfjorden (Vorren & Plassen 2002; Ottesen et al. 2005) and Vestfjorden-Trænadjupet (Ottesen et al. 2005; Knies et al. 2007; Laberg et al. 2007) during the last glaciation (Fig. 2). An important result of these marine-geological and geophysical studies is that the margin of the ice sheet in this region has fluctuated more often and more rapidly than was previously realized.
Some of the published results and interpretations are contradictory. Thus the aim of this paper is to reappraise the chronology and the palaeoenvironmental LGM history on Andøya.
Andøya is characterized by mountains rising sharply to 300 to 600 m a.s.l., flanked by extensive areas of mire-covered strandflat. The up to 505-m-deep fjord Andfjorden, which is situated to the east and north of Andøya, was an important drainage outlet for the Fennoscandian Ice Sheet during the LGM, and the ice stream draining through Andfjorden led to the formation of a dense pattern of glacial lineations (Vorren & Plassen 2002; Ottesen et al. 2005).
The continental margin off the Lofoten–Vesterålen islands comprises a narrow and thin crustal segment overlain by a clastic sedimentary wedge of Permian through to Palaeogene age. The bedrock on Andøya is composed largely of Precambrian gneissic rocks. A restricted near-coastal expanse of Mesozoic coal-bearing sedimentary rocks occurs ∼10 km south of the study area, as well as beneath the Quaternary cover in Andfjorden (Bergh et al. 2008).
Material and methods
Endletvatn is presently a NW–SE-oriented 1.2-km-long lake near the northern tip of Andøya Island (Fig. 1). A sub-basin in the southwestern extension is now filled with sediments and overgrown as a mire. In 2002 and 2003, four cores (from site 6, Fig. 1C) were sampled from this part of the original lake. Here, the sub-basin was mapped using ground-penetrating radar (GPR) from the Geological Survey of Norway (Fig. 1C). The GPR was a digital pulse EKKO 100 (Sensor & Software Inc., Canada). The recordings comprised six N–S profiles and five E–W crossing profiles, with a total length of 2223 m. A source of 1000 V and an antenna with a centre frequency of 100 MHz were used for all profiles. A marked reflection was recorded in all profiles (Fig. 3). This reflection can be followed to a depth of 9–11 m. The limited range of depth recording is probably due to a strong reduction of the signal in the fine-grained basin sediments. Evidently, the marked reflection represents the boundary between coarse-grained sediments below (bedrock or/and diamictons and sandy gravel) and the fine-grained basin infill sediments above. The four C-cores at site 6 all reached this boundary between 12 and 13 m, that is, below the range of the GPR recordings, but in line with the interpolation of the GPR recordings. The C-cores were situated at or close to the deepest part of the sub-basin.
The four cores C1–C4 were retrieved with a 100-mm Geonor clay sampler, mainly in PVC tubes. During coring the tube lengths were about 2 and 1 m long. The 2-m cores were cut to lengths of between 1 and 1.5 m after retrieval. The basal core sections in C1 and C2 were collected in aluminium tubes (C1, core section 12–12.6 m, and C2, core section 11–12 m). The material in aluminium tubes was extracted using a hydraulic piston.
Physical properties, including p-wave velocity, wet-bulk density and magnetic susceptibility (MS) were measured using a Multi-Sensor-Core-Logger (MSCL). The measurements were carried out on unsplit cores, except for the magnetic susceptibility measurements of the cores in aluminium tubes. Here, measurements were taken after removal from the tube.
On the split cores, colour determinations were made using a Munsell Soil Colour Chart. They were photographed at 0.5-m overlapping intervals. For each vertical centimetre, 3-cm3 samples were retrieved and analysed for loss on ignition (LOI) and water content (LOW).
Seven grain-size analyses were carried out. Prior to the measurements, carbonates and organic matter were removed with acetic acid and hydrogen peroxide, respectively. After allowing the chemicals to react overnight, the samples were washed with de-ionized water (twice after each treatment). Subsequently, sodium polyphosphate was added to each sample for dispersion and the samples were left on a shaking table overnight. The grain-size measurements were carried out with a Cilas 1180 L laser-diffraction particle size analyser (range 0.04–2500 μm). Data processing and statistical analyses were performed using self-programmed routines and the software gradistat (Blott & Pye 2001). The results are presented in volume per cent.
Qualitative element-geochemical measurements of core C3 were performed using an Avaatech XRF Core Scanner equipped with a rhodium X-ray source. The measurements were carried out with a 2-mm down-core slit size and a 12-mm cross-core slit size using the following settings: 10 kV, 1000 μA, 10-s measuring time, no filter. During the measurements, the sediment surface was covered with a 4-μm ultralene foil. Selected results are presented as element ratios to minimize the influence of water and matrix effects (Tjallingii et al. 2007; Weltje & Tjallingii 2008). Prior to the measurements, a colour image of the core was acquired using a Jai L-107CC 3 CCD RGB Line Scan Camera installed on the XRF Core Scanner.
Botanical macrofossils were washed out using sieves with mesh sizes of 1, 0.2 and 0.063 mm. All fragments >0.063 mm were collected. The sieves were cleansed by means of compressed air between samples to be washed out. The macrofossils were identified and counted under a stereomicroscope. Identifications of vascular plant remains were according to Beijerinck (1947) and Berggren (1981), and the seed/fruit collection at the Department of Arctic and Marine Biology, University of Tromsø. The abundance has been adjusted to a fixed volume of 35 mL per sample. The samples were retrieved from the core in 2-cm-thick slices.
Bryophyte macrofossils were studied in core C3. The bryophyte abundance was determined by simple counting, where all free parts of a species were given equal weight: leaf fragments, entire leaves and shoots of different size were given the weight 1. Nomenclature of mosses follows Hill et al. (2006).
Two samples for cosmogenic surface exposure dating were obtained from the Kjølhaug and Endleten moraines respectively (Fig. 1). They were retrieved with a hammer and chisel from the uppermost 2 cm of horizontal crests of boulders. The samples were processed for 10Be from quartz following procedures based on methods modified from Kohl & Nishiizumi (1992) and Child et al. (2000). AMS measurements were carried out at the PRIME Laboratory, Purdue University, USA, and measured 10Be/9Be was corrected by full chemistry procedural blanks. The 10Be concentrations were converted to exposure ages using a 10Be half-life of 1.5 Ma. To calculate apparent exposure ages, the CRONUS-Earth ver. 2 10Be/26Al exposure age calculator (Balco et al. 2008) was used. The calculator uses a sea-level high-latitude (>60°) nuclide production rate of 4.96±0.43 atoms g−1 a−1 (10Be) scaled to altitude and latitude using algorithms derived by a number of authors. Variations in calculated ages are <4% between the different scaling models, and here we quote the ages obtained by applying the Lm model (Balco et al. 2008). A correction was applied for sample thickness using an attenuation coefficient of 160 g cm−2 and a rock density of 2.65 g cm−3. We conservatively corrected for snow shielding assuming 0.3 m of snow during 4 months per year. Shielding factors were calculated with a snow density of 0.3 g cm−3.
Forty-seven samples from units A–K were radiocarbon-dated by AMS (Table 1), most at the Laboratory for Radiological Dating in Trondheim, Norway (samples named Tra in Table 1). Some samples had their age determined at the Svedberg Laboratory in Uppsala, Sweden (samples named TUa in Table 1). Three samples were analysed at the 14Chrono Centre for Climate, the Environment, and Chronology at Queen's University in Belfast, Ireland (samples named UBA in Table 1). Bulk as well as macrofossil samples were analysed. In some cases, a bulk sample and a macrofossil sample from the same level were dated. All TUa/Tra-sample types were dried at 105°C for 24 h before weighing. The bulk samples comprised 0.5 and 1.0 vertical centimetres, whereas the macrofossil samples could comprise as much as 11 vertical centimetres. Dates were calibrated according to IntCal09 (Reimer et al. 2009).
Table 1. Radiocarbon dates. The sample types are macrofossil samples (M), bulk samples (B) and algae (A). Macrofossil samples comprise mainly bryophytes and seeds. Dates were calibrated according to IntCal09 (Reimer et al. 2009), using 1 sigma. In cases where the relative area of the probability distribution was divided, the mean of the larger area was chosen to represent the sample's point in the age–depth curve.
Depth (cm below surface)
14C age (cal. a BP; 1σ)
Mean 14C age (cal. a BP)
13 800–13 981
15 012–15 279
14 963–15 255
12 808–13 149
14 874–15 258
16 860–17 075
18 621–18 833
17 116–17 543
18 505–18 702
18 549–18 706
18 083–18 306
15 109–15 885
18 975–19 318
19 890–20 294
20 929–21 446
13 840–14 153
14 207–14 770
16 789–16 970
19 231–19 427
18 100–18 292
17 070–17 459
18 260–18 490
20 068–20 537
20 186–20 568
20 248–20 557
15 436–16 083
19 843–20 430
17 183–17 557
17 456–17 849
18 615–18 793
18 735–18 988
20 067–20 407
19 832–20 146
22 096–22 438
21 866–22 410
22 071–22 377
21 754–22 220
22 956–23 575
16 679–16 896
19 222–19 549
16 984–17 257
18 260–18 490
18 265–18 489
18 557–18 885
21 072–22 085
The results of the radiocarbon dates are listed in Table 1. The four cores were analysed with respect to lithological related parameters and divided into 11 lithostratigraphic units and nine correlating horizons (Fig. 4). Using the correlating horizons and lithostratigraphic units the age determinations (except for two) were transferred to the C3 core (Fig. 5). The two deleted samples were retrieved near core breaks (C1/1105–1109 cm and 1203–1206 cm, Table 1), and are obviously contaminated by younger plant remains.
For the rest, a best-fit curve is drawn for the calibrated ages (Fig. 5), given (i) that there is no hiatus below Unit K in core C3 (the facts that all four cores have the same stratigraphy and do not show any obvious signs of erosion, and that the individual units have almost the same thicknesses indicate that hiatuses in units B to J are absent, except possibly in the lowermost part, namely Unit A); (ii) that the sedimentation rate was almost constant within each individual lithostratigraphic unit (the lithology within each unit is relatively uniform, indicating relatively constant sedimentation rates); and (iii) that the boundary between units J and K is at 15.1 cal. ka BP. The boundary between units J and K was previously dated by Vorren (1978) to 15.2, 15.1 and 13.7 cal. ka BP, as well as by Vorren & Alm (1999) to 15.2 cal. ka BP. Furthermore, the corresponding boundary in the adjacent lakes Nedre and Øvre Æråsvatn, respectively, has been dated by Vorren et al. (1988) to 15.1 and 15.1, and by Alm (1993) to 15.1 cal. ka BP. Thus it seems safe to suggest an age close to 15.1 cal. ka BP.
Several of the dates fall far outside the best-fit curve (Fig. 5). Hence, it is obvious that some of the dating results are in conflict. The main reason for these discrepancies is probably the content of reworked sediments. Most of the bulk samples from units K, J, I and the upper part of Unit H are obviously too old (area 1 in Fig. 5). We believe that this is due to a mix of contemporary and older reworked organic material. It should be noted that the two samples from the same horizon in two different cores (C2 and C4) give the same age. This probably indicates that the amount of reworked material is of the same magnitude in this horizon. Too old ages are also given by three bulk samples in units G, E and F (area 2 in Fig. 5).
Macrofossil samples from units E and G seem to provide too young ages (area 3 in Fig. 5). We have no simple explanation for this deviation. In contrast to this, three of the lowermost macrofossil samples (from units C and B in core C3) are probably too old (area 4 in Fig. 5). The two next lowermost samples were extracted from the same layer. The algae (Table 1) gave a slightly younger age than the macrofossils (21 987 versus 22 224 cal. a BP). We rely on the age determination of the algae as the algae were produced concurrently with the sediments settling in the basin. For the other samples giving older ages in this group, we suggest that some of the macrofossil leaves were reworked from older deposits.
The δ13C values range from −11.8 to −33.4‰ (Table 1). According to Mackie et al. (2005) and references therein, it is expected that marine organic matter should have δ13C values of −10 to −22‰, whereas marine bulk organic matter has δ13C values ranging from −19 to −22‰. In contrast, terrestrial plants in the Northern Hemisphere record (with C3 photosynthesis) have mean δ13C values of around −27‰, with the full range being −32 to −20‰. In the present material, δ13C values from the macro-algae layer in Unit C yielded values of −22.1, −24.9, −23.8 and −26.1‰ (from TUa-5341A, TUa-4925, TUa-4940 and TUa-5794, respectively). All these values lie outside the expected δ13C range for marine organic matter and marine bulk organic matter. Vorren et al. (1988) also investigated δ13C values in the nearby Lake Nedre Æråsvatn, where they found the same results. They explained the anomaly using the results of Deuser et al. (1968), who found that marine plankton from cold waters with an abundant supply of CO2 yielded δ13C values as light as −28‰. Vorren et al. (1988) argued that marine algae could behave similarly with regard to carbon sources and photosynthetic pathways. This could be the explanation for the light δ13C values in the 14C samples from the macro-algae layer, and thus no corrections for marine samples were performed.
Lithostratigraphy – results and interpretations
The split cores were placed side by side, and nine correlating horizons were defined (Figs 4, 6). Based on changes in colour, grain size (Figs 4, 6, 7), fluctuations in LOI, LOW and MS (Fig. 6), as well as chemical composition (Fig. 8), 11 lithostratigraphic units (A – lowermost to K – uppermost) were identified. Although the values are not identical, LOI, LOW and MS show the same general trends through all four cores (Fig. 6).
Unit A is found in cores C1 and C3, and probably rests on bedrock. In C1, it comprises massive sand and gravelly sand with irregular lenses of laminated mud and sand. Unit A in core C1 is highly deformed and compacted (low water content). We suggest that this is due to glacial tectonics occurring during the last glacial re-advance across the area. Unit A in core C3 is a sandy layer containing pebbles (Fig. 9A). These coarse sediments were probably glacifluvially deposited during or just after the deglaciation of the area.
Unit B is found in cores C1 and C3, and Unit C in cores C1, 2 and 3. It comprises laminated clayey silt with sand laminae (Fig. 7). Faults occur in core C1, and in C3 there is an angular unconformity near the top of Unit B (Fig. 9A). Unit B and the lower part of Unit C contain some fine-sand laminae (Figs 7, 9A, 7). They may derive from small turbidites. Well-sorted sand laminae occur in Unit B and at the base of Unit C. The contents of Ca and S are relatively high (Fig. 8). A dark layer occurs in the middle of Unit C (Fig. 9A) where the brown seaweed (Desmarestia aculeata) and green microalgae (cf. Vorren et al. 1988) occur in abundance in this 8-cm-thick laminated interval (about 20.5 cal. ka BP). This indicates that Lake Endletvatnet was transgressed by the sea for a short period during the LGM. The brown seaweed was also found in abundance in the marine LGM sediments in Lake Nedre Æråsvatn (Vorren et al. 1988). The species has a circumpolar distribution, extending south to Portugal in Europe. It is also found in sub-Antarctica and Antarctica (Algaebase 2012). It is one of the most common halophytes in Svalbard today, where it is typically found in the intertidal zone and down to at least 20 m (Jaasund 1965).
The upper boundary between units C and D is sharp, and represented with a change to a lighter colour. Units D, E, F and G contain laminated silt and clay (Figs 4, 9B). Unit D is characterized by low LOI, increasing LOW (Fig. 6), and relatively low Ca and S contents (Fig. 8), while Unit E is distinguished by a small peak in LOI, a sudden drop in LOW, and relative increases in Ca and S (Figs 6, 8). The upper boundary towards Unit F is gradual, detected mainly by a change in colour. Unit F is in many respects similar to Unit D, that is, with low LOI values, and relatively low Ca and S contents. However, the LOW values are higher than in Unit D. Unit G is characterized by high organic content, reflected by several LOI-peaks (Fig. 6) and relatively high Ca and S contents (Fig. 8). This unit resembles Unit E in most respects. The boundary between units G and H is gradual, marked by a slight colour change and a decrease in LOI.
Units H, I and J also contain packages of finely laminated clayey silt (Fig. 9B). Unit H is somewhat disturbed by minor faults (Fig. 9B). The faulting is probably the result of small-scale internal gliding, possibly owing to compaction, along the margin of the basin. The most characteristic feature of Unit I is the pronounced peak in MS and a temporary decrease in LOW (Fig. 6). The visual boundary between units I and J is marked by a gradual change in colour.
From the lower part of Unit C and upwards to the base of Unit K, the grain-size distribution and primary structures are fairly uniform. Finely laminated clayey silt dominates. The sedimentation rate is low and almost constant, about 0.5 mm a−1. This indicates a stable and low-energetic physical environment during the 7000 years; that is, there have been no severe erosional events transporting coarse-grained sediments from the nearby mountain sides to the basin. An arid climate with permafrost and a sparse vegetation cover could provide conditions leading to this type of sedimentary environment.
The variations in the chemical composition (Fig. 8) could be explained by changes in the provenance of the sediments transported to the basin, or by changes in the intra-basinal sediment production. There is no indication of changes in the drainage area of the basin since the area was deglaciated. Thus, the provenance of the sediment transported to the basin by small brooks has probably not changed. However, changes in sediment source related to aeolian transport may have occurred. The trimodal/bimodal grain-size distribution of samples from units D, E and G can best be explained by aeolian input (Fig. 7). The shoreline regression that occurred after Unit G time (see below) exposed new land areas and consequently new source areas for aeolian erosion. The peak in magnetic susceptibility in Unit I may possibly derive from aeolian sediments from newly exposed areas. The changes in the LOI curve mostly follow the changes in chemical composition. There is a relative increase in S in units C, E, G and K. Calcium follows the same trend, but Ca also increases in units I and J. The internal production of organisms in the lake (mainly algae) is probably the main reason for the variations in chemical composition.
Unit K is found in C1, C2 and C3. In core C3 this unit is directly underlying a gyttja sequence (Fig. 9B). Radiocarbon dates indicate that this boundary marks a hiatus of varying length. The hiatus in core C3 is ∼3500 years. In core C1, the sediment comprises deformed laminated clayey silt with some organic content. This unit is characterized by a sudden increase both in LOI and in LOW (Fig. 6).
Biostratigraphy – results and interpretations
Vascular plant remains
Three types of seeds were found regularly in all four cores, namely Poaceae, Papaver and Brassicaceae. The majority of the Poaceae material probably belongs to the genus Puccinellia and, as stated by Alm & Birks (1991), numerous arctic Puccinellia species should be considered. All Papaver seeds belong to the group Papaver sect. Scapiflora, namely the arcto-alpine Papaver-group (cf. Berggren 1981). Seeds of the species belonging to this group are indistinguishable from each other, but all are distributed in alpine or arctic habitats. In the present detailed analysis of core C3 (Fig. 10) only the Draba-type is recorded. However, another Brassicaceae seed type also occurs in the C3 sequence in a parallel sample from the 985–987 cm level.
No seeds were found in units A and B. In Unit C, Poaceae and Papaver occur regularly below and above the layer of marine algae (Fig. 10). Unit D is practically without seed occurrences, whereas E and F have sporadic occurrences. Unit G has a regular occurrence of Poaceae and Papaver seeds. In the middle part of Unit H there are sporadic occurrences of the two seed types, whereas they are absent in the lower and upper part of H. In units I and J the seeds occur quite regularly – and in Unit K especially the Poaceae number increases. The two occurrences of the Draba-type seeds are in the lower part of units I and K (Fig. 10).
The seed maxima in this study are correlated with the LOI increase in units C (lower and upper part), G and K, indicating higher terrestrial biological productivity in these units. There are also seed occurrences in units H (central part) and J, however, that do not show increases in LOI. Fluctuations in seed types and frequencies are interpreted as signals of actual vegetation changes. This study as well as earlier biostratigraphical, palynologically based works (Vorren 1978; Vorren et al. 1988; Alm & Birks 1991; Alm 1993) all show that the vegetation during the LGM was dominated by Poaceae, Papaver and Brassicaceae.
The sediments from units B to K in core C3 were investigated for their bryophyte contents. Only mosses, and no liverworts, were recorded (Fig. 10). The three dominating taxa were Syntrichia ruralis, Aulacomnium turgidum and Bryum spp.
Units H (middle part) and K represent the highest diversity. Almost clean layers of Syntrichia ruralis occur in unit B and C sediment, whereas Bryum spp. form an almost clean layer in the middle of Unit H.
The ecology and distribution of the various mosses (especially Syntrichia ruralis and the Tortula species) found in units B to K indicate a cold and dry, continental climate and basic soil allowing discontinuous vegetation cover with much disturbance. However, there are also indications of moisture-demanding vegetation, such as Bryum calophyllum (occurring near brooks and on banks at lakes). There are also species typical of wet soil surrounding small brooks, which is to be expected, considering the local topography. The present geographical distribution of the mosses is primarily Holarctic, although with a northern and mountainous tendency. The moss assemblage could be characterized as an impoverished version of the zonal vegetation type on patterned ground in the northernmost subzone A of Canada, namely the polar desert or Papaver radicatum subzone (Walker et al. 2011), as is the vascular seed assemblage. Parducci et al. (2012) arrived at a similar conclusion based on i.a. cpDNA analysis of core C1.
Several bones of the arctic bird little auk (Alle alle) were found in the cores (Elverland et al. 2007), in particular in units C and G (c. 20.5 and 18.5 cal. ka BP) and in the basal part of Unit K (15.0 cal. ka BP). The little auk is an arctic species that at present breeds on eastern Baffin Island, Greenland, Jan Mayen, Svalbard, Iceland, Franz Josef Land, Novaya Zemlya and Severnaya Zemlya. Two sub-species of the little auk are recognized: the nominate race A. a. alle and the significantly larger A. a. polaris. The latter inhabits Franz Josef Land, whereas the nominate race inhabits the rest of the breeding range, including Svalbard (http://www.npolar.no/en/species/litle-auk.html). The fossil bones found are relatively large, indicating that they may derive from A. a. polaris.
An earlier finding of a well-preserved vertebrae of stout (Mustela erminea) at a stratigraphic level corresponding to Unit C was reported by Fjellberg (1978). Presently, M. erminea has a wide distribution including arctic and alpine environments. It occurs in, for example, the Papaver radicatum desert of northernmost Greenland and the Canadian Arctic archipelago. As indicated by Fjellberg (1978), the presence of this carnivorous animal indicates the presence of other animals as well. The prey of stout consists mainly of small rodents.
Several studies have been carried out to reconstruct the shoreline displacement on Andøya (Undås 1967; Andersen 1968; Møller & Sollid 1972; Bergstrøm 1973; Møller 1985; Fjalstad & Møller 1994; Fjalstad 1997). Lake Nedre Æråsvatn (35 m a.s.l., Fig 1C) contains marine sediments from 22.2 to 18.7 cal. ka BP, corresponding to units (B?), C, D, E and F in Lake Endletvatn. Our present results show that the middle part of Unit C in Endletvatn (36 m a.s.l.) was deposited in a marine environment. Thus, we can now bracket the age of the highest relative sea level after the final deglaciation (the marine limit, ML) on Lake Endletvatn to be between 21.0 and 20.3 cal. ka BP, peaking at around 20.7 cal. ka BP (Fig. 11). Clark et al. (2009) found a rapid 10-m rise in sea level from the LGM lowstand sometime between 19 and 20 cal. ka BP. Possibly the ML on northern Andøya reflects this sea-level rise.
Endletvatn provides a minimum altitude of 36 m for the relative sea level. The entire sequence of Lake Øvre Æråsvatn is lacustrine, showing that the maximum sea level since the deglaciation has been lower than 43 m a.s.l. (Alm 1993). The present outlet of Endletvatn is to the east but was probably dammed by the glacier during the LGM. Thus, it is reasonable to assume that Lake Endletvatn drained to the north or west across passes 38 m a.s.l. Fjalstad & Møller (1997) describe a section just north of Lake Endletvatn containing laminated low-angle sandy crossbeds, which they interpret as shore sediments. The relative sea-level altitude for these sediments is given as ∼38 m. These sediments were dated to ±18.0 ka by photoluminescence and thermoluminescence dating techniques. Thus the sediments were deposited within the age range of Unit C. We conclude that the ML on northern Andøya is between 36 and 38 m, and that this occurred c. 20.7 cal. ka BP.
The younger part of the sea-level curve has been adjusted to a date from Lake Storvatnet (26 m a.s.l., Fig 1B) where gyttja just above the marine isolation contact gave an age of 17.0 cal. ka BP. Furthermore, peat at 10 m depth offshore Andenes (Fig. 1B) gave an age of 11.1 cal. ka BP (Fjalstad & Møller 1994).
The sedimentary environment seems to have been quite stable throughout the nearly 7000 years, from 22.2 to 14.5 cal. ka BP, represented by the LGM sequence in Endletvatn (units B to K). The sedimentation rate is fairly modest (0.5 mm a−1), and the sediments do not show any signs of flooding events. We infer this to mirror a dry climate with perennial frozen ground. This is also the general picture provided by the fossil record.
The fossil bird bones indicate an arctic environment. If it is correct that the fossil bird bones derive from A. alle polaris, mean July temperatures of as low as +2°C may have prevailed (based on mean July temperatures in the present habitat of A. alle polaris on the archipelago Franz Josef Land).
We compare our results of the fossil moss assemblages with studies by Walker et al. (2011) of the modern zonal vegetation in the arctic Canadian archipelago. In their polar desert zone (arctic subzone A, reflecting a mean July temperature between 0 and 3°C), Syntrichia ruralis occurs in 90% of the relevees and Aulacomnium turgidum in 48%. However, these mosses may also occur in other subzones. Aulacomnium turgidum occurs in its greatest quantities in the southernmost subzone E, and Syntrichia ruralis also occur in subzones B and C in smaller quantities. The combination of Syntrichia ruralis and Aulacomnium turgidum, which are the two dominating taxa here, is unique for the polar desert zone.
Subzone B of Walker et al. (2011), also called the Dryas zone, reflects a July mean temperature of between 3 and 5°C. Common to the northernmost subzones A and B are the mosses Sanionia uncinata and Polytrichastrum alpinum. The total absence of those mosses in the 22.2–14.5 ka sediments may indicate an even harsher climate during the Andøya Late Weichselian polar deserts than in the modern Canadian polar deserts. It should be noted here that the Allerød sediments of Andøya are rich in those two moss species.
Referring to the general assemblage of mosses within units B–K, it seems likely that the mean July temperatures at Andøya in the period 22.2–14.5 cal. ka BP were closer to 0 than to 3°C. Owing to the variations in lithology and biostratigraphy, we infer that the climate has been variable within this narrow limit. The seed production seems to mirror climatic oscillations, pointing to units C, G and I–K as representing the more optimal phases. The mosses indicate the middle of Unit H as a potentially favourable period, starting with a moist phase (Bryum spp. maxima) and optimally with a very dry phase, as evidenced by Syntrichia ruralis and the Tortula species.
Vorren (1978) inferred two ‘warm’ periods, the Endletvatn thermomers 1 and 2 (ET1 and ET2), experiencing July mean temperatures between 8 and 10°C. Evidently ET1 can be correlated with units B (22.2 to 21.4 cal. ka BP) and (C?), and ET2, having an age between 19.3 and 18.1 cal. ka BP (Alm 1993), correlates with units E, F and G (19.5–18.1 cal. ka BP). ET1 was inferred from pollen spectra in the lowermost part of Endletvatn containing Betula nana, Ericales, Rubiaceae and Scorpidium scorpiodies, indicating a climate of middle to low arctic type (Vorren 1978). In Øvre Æråsvatn, a climatic amelioration is indicated by an increase in pollen concentration between 21.8 and 21.4 cal. ka BP (Alm 1993). Nedre Æråsvatn shows no indication (Vorren et al. 1988). ET2 was inferred from pollen of Apiaceae and cf. Melampyrum and from moss fossils of Sphagnum papillosum in Endletvatn (Vorren 1978). In the present material there are no similar indications of such high mean July temperatures. Because the present material was taken from larger samples where the risk of contamination should be negligible, we suspect that the 1978 data probably suffered from reworked fossils and/or contaminated samples. Reworked fossils may derive from reworked interglacial sediments.
In conclusion, throughout the period between 22.2 and 14.5 cal. ka BP the mean July temperature fluctuated between 0 and 3°C. The warmest periods occurred during the deposition of units C, G middle H and I–K; that is, from 21.4 to 20.1, from 18.8 to 18.1, around 17 and from 16.4 cal. ka BP onwards. The coldest periods are represented by units D, F parts of H and I (Fig. 12). The shifts between the different climatic regimes were rapid – probably occurring within less than a couple of decades.
Did trees live on Andøya during the LGM?
Parducci et al. (2012) conclude that results from macrofossil and cpDNA analysis indicate the presence of a polar desert or open pioneer vegetation community from c. 22 000 cal. a BP. Furthermore, they suggest that tundra herb diversity increased with climatic warming c. 15 000 cal. a BP. This is in general accordance with our results. They also claim that Picea and Pinus grew on Andøya c. 22 000, 19 200 and 17 700 cal. a BP, but do not explain how this polar desert could host Pinus and Picea. We suggest that their finding of traces of pine and spruce is due to reworked sediments. As mentioned above, the radiocarbon dates of bulk material also indicate older carbon in the sediments, which probably derives from older interglacial/interstadial plant remains.
Another intriguing finding has emerged in the last decade. Kullman (2006, 2006,2008, http://www.kullmantreeline.com/empty_31.html) claims that Betula pubescens s.l. grew on Andøya 20 130 cal. a BP. In a study on Stavedalen, 60 m a.s.l. (Fig. 1B), he describes wood remnants protruding 5–10 cm above the surface of a Sphagnum hummock. One piece of wood, a root from Betula pubescens, 40 cm in length and 8 cm in diameter was retrieved (Kullman 2006). We did not find any indication in our material of B. pubescens during this time. Furthermore, the nature of the finding site led us to seriously doubt that the age determination of the tree remains is correct. Thus we conducted field studies at Kullman's finding site, indicated on the photo in Kullman (2006). After a thorough search, one remnant of a B. pubescens tree root was found at the spot he indicates. The root contained minerogenic matter between the growth rings. A well-cleaned sample gave a radiocarbon age of 101±35 14C a BP (UBA-20621). The two most likely explanations to this age discrepancy are: (i) that we have not dated the same root, or (ii) that the dated root of Kullman contained old carbonaceous material.
The LGM glaciation history
Vorren et al. (1988) and Vorren & Plassen (2002) reconstructed the LGM glacial events on the northern part of Andøya as well as in the adjoining Andfjord. We have refined this reconstruction of the glacial history based on the results of the present study and on other results that have emerged in the last three decades from northern Andøya. If our assumption that the base of UnitB represents the start of the deglaciation is correct, the result from Lake Endletvatn seems to corroborate the lake Nedre Æråsvatn record. Two 10Be exposure ages from boulders on the Kjølhaug Moraine and the Endleten Moraine (Fig. 1C), respectively, also seem to corroborate the Lake Nedre Æråsvatn record. The Kjølhaug Moraine has an apparent age of 22.1±2.2 10Be ka, and the exposure age of the boulder on the Endleten Moraine is 20.0±2.1 10Be ka.
Dahlgren & Vorren (2003) and Rørvik et al. (2010) studied the glaciation history using data retrieved from the continental slope just south and west of the Lofoten islands (Fig. 2). Both works indicate that the start of the LGM occurred c. 26.0±0.5 cal. ka BP, and that several oscillations of the glacier front occurred during the LGM.
Early LGM (c. 26.0–c. 23.5 cal. ka BP). –
This might be the period when the LGM ice sheet was thickest on Andøya, according to Lambeck et al. (2010). However, exposure dates indicate that the ice sheet did not cover the mountain areas on Andøya (Nesje et al. 2007). The glacier may have had an extent as indicated in Fig. 12 (see below). For comparison, it is interesting to note that on the western shelf off Svalbard the glacier front reached its maximum, the shelf edge, between 24 and 23.5 cal. ka BP (Jessen et al. 2010).
23.5±0.5 cal. ka BP. –
Unit II in Lake Nedre Æråsvatn (Vorren et al. 1988) reflects a period when this lake and its surroundings were deglaciated and the lake was inundated by the sea. Based on the fossils found in Unit II, a middle to low arctic climate was inferred to prevail during this period (Vorren et al. 1988). Olsen et al. (2001) reconstructed nine glaciation curves across Norway from the inland area to the coast and shelf. They define a Trofors interstadial (25–20.2 cal. ka BP). The Trofors interstadial probably comprises more than one deglaciation period (interstadial). The occurrence of animal bones and calcareous concretions from a cave at Kjøpsvik indicates that the ice front receded to the inner fjord areas during this period (Lauritzen et al. 1996).
23–22.2 cal. ka BP. –
During this period Lake Nedre Æråsvatn was overrun by a glacier. Vorren et al. (1988) discussed the extent of the glaciers on northern Andøya during this period, namely during the period just before the final deglaciation of this area. The extent of the glaciers reconstructed in Fig. 12 may reflect the situation during this period. This reconstruction is in conflict with the basal bulk dates of silty gyttja from Lake Øvre Æråsvatn (26.1–25.8 cal. ka BP; Alm 1993) and the exposure dates from Lake Store Æråsen (45–36 ka; Nesje et al. 2007). However, the upper sediment in Unit II in Lake Nedre Æråsvatn is highly deformed and consolidated, indicating that the ice deforming the sediments must have had a substantial thickness. Furthermore, a lateral till terrace about 100 m a.s.l. along the mountain side of Røyken (Fig. 1B) probably marks the margin of the ice sheet during this period. If this is correct, it implies that Lake Øvre Æråsvatn and Stor Æråsen experienced a glacial cover during this period. Alternatively, if the basal dates in Øvre Æråsvatn are correct, this lake may have been covered by glaciers during the early LGM and have been deglaciated as early as c. 26.0 cal. ka BP. Then, the reconstruction (Fig. 13) could illustrate this period, implying that the glaciation after Unit II was deposited had a more restricted extent. We favour the first alternative.
22.2–18.7 cal. ka BP. –
Early during this period, the western margin of the Andfjorden ice stream receded to the Kjølhaugen Moraine and shortly thereafter to the Endleten Moraine or beyond (Fig. 1C). Vorren & Plassen (2002) tentatively correlated the Kjølhaug Moraine with the Bjerka Moraine in outer Andfjorden, and the Endleten Moraine with the Egga-II event when the ice stream reached the shelf-break. Møller et al. (1992) recovered shell fragments of Mya truncata at Bleik (Fig. 1B) giving an age of c. 21.0 cal. ka BP, indicating that the Bleik area was deglaciated at this time. Furthermore, Dahl et al. (2010) suggest, based on OSL dating of beach ridges 20 m a.s.l. on Bleik, a limited ice extent and seasonally open water around 19 ka, and suitable conditions for rock glacier formation close to the present sea level between 19 and 15 ka.
Few data are available that can reveal the behaviour of the Andfjorden ice stream later in this period. The rather stable position of the shoreline for 3000 years (c. 22.2 to 18.5 cal. ka BP; Fig. 12) could indicate that the Andfjorden ice stream was relatively stable during this period. On the other hand, IRD data presented by Rørvik et al. (2010) occurring at the end of Unit C and Unit E time point to periods with down-melting and subsequent drawdown and calving of the ice stream in Trænadjupet–Vestfjorden. Possibly this occurred in Andfjorden as well.
Paasche et al. (2007) indicated that local glaciation occurred on southern Andøya between 21 050 and 19 100 cal. a BP. We are reluctant to accept their conclusion, as it is based on assumptions about the modern equilibrium line altitude (ELA) that are in conflict with existing glaciers, and consequently on incorrect estimates concerning the magnitude of the depression of the ELA. Rather, this local glaciation occurred during the Younger and Older Dryas, as found for adjoining areas (e.g. Andersen 1968; Rasmussen 1984).
After 18.7 cal. ka BP.
The final drawdown and break-up of the Andfjorden ice stream in the outer and middle reaches occurred c. 17.8 cal. ka BP (Vorren & Plassen 2002). Shortly after the break-up, the Andfjorden ice stream re-advanced/halted at the Flesen Moraine in inner Andfjorden, and later at the innermost part of Andfjorden between 16.6 and 15.1 cal. ka BP (Vorren & Plassen 2002).
This study of Lake Endletvatn on northern Andøya has provided more detailed and precise dates to give a better understanding of the sedimentary environment and of the floral, faunal, climatic and glacial development in the Andøya–Andfjorden region during the LGM.
The structures and texture of the sediments indicate a low-energy environment with a mean sedimentation rate of 0.5 mm a−1 and perennially frozen ground in the surroundings through the LGM. The sediment influx was from small brooks and aeolian activity as well as from intrabasinal organic production that varied over time.
Brown seaweed occurs in abundance between 21.0 and 20.3 cal. ka BP, peaking around 20.7 cal. ka BP, indicating that Lake Endletvatnet (36 m a.s.l.) was transgressed by the sea for a short period during the LGM. The maximum relative sea level was 38 m a.s.l.
The climate proxies indicate a dry, high arctic climate (i.e. July mean temperatures between 0 and 3°C) between 22.2 and 14.5 cal. ka BP. The highest July mean temperatures were from 21.4 to 20.1, from 18.8 to 18.1, around 17 and from 16.4 cal. ka BP onwards. The shifts between the different climatic regimes were rapid – probably occurring over one or two decades.
Given a dry, high arctic climate from 22.2 to 14.5 cal. ka BP, we find it implausible that Pinus, Picea or Betula should have existed during parts of the LGM on Andøya as suggested by Parducci et al. (2012) and Kullman (2006, 2006,2008).
We infer the LGM glaciation history in the Andfjorden–Andøya region to be as follows: c. 26–c. 23.5 cal. ka BP: possibly maximum extent, but the glaciers did not override the mountains. 23.5±0.5 cal. ka BP: northern Andøya was deglaciated. 23–22.2 cal. ka BP: northern Andøya was glaciated. Maximum extent of the LGM glacier might have occurred during this period. 22.2–18.7 cal. ka BP: during an early stage of this period the western lateral margin of the Andfjorden ice stream receded to the Kjølhaugen Moraine and shortly thereafter to the Endleten Moraine. The Andfjorden ice stream possibly experienced two recessions later during this period. After 18.7 cal. ka BP: the final drawdown and break-up of the Andfjorden ice stream commenced c. 17.8 cal. ka BP, after a period of down-melting. Re-advances/halts of the Andfjorden ice stream produced the Flesen Moraine c. 17.6 cal. ka BP, and occurred in innermost Andfjorden between 16.6 and 15.1 cal. ka BP.
Funding provided by the University of Tromsø and the Research Council of Norway through the Norpast and the SPONCOM projects is gratefully acknowledged. We offer our sincere thanks to Ellen Elverland, who did much of the LOI, LOW analysis and otherwise participated in the laboratory work; to Jan Fredrik Tønnesen and Terje Bargel, Geological Survey of Norway, who supervised the ground-penetrating radar measurements; to H. Christian Hass at the Wadden Sea Research Station of the Alfred Wegener Institute for Polar and Marine Research in List, Germany, who carried out the granulometric analyses; to Derek Fabel, Department of Geographical and Earth Sciences, University of Glasgow, UK, who carried out the two exposure age determinations; to Jan Petter Holm, Department of Geology, University of Tromsø, who helped with the figures, to Det norske oljeselskap ASA for financial support for the XRF core-scanner measurements; to Julie H. Velle for laboratory assistance, and to reviews by Jan Mangerud and the reviewers Atle Nesje and Jochen Knies for constructive comments as well as to the Editor-in-Chief, Jan A. Piotrowski.