Fieldwork was carried out during the summer 2005 and included geomorphological surveys in the study areas, sedimentological logging at selected sites and sampling for cosmogenic exposure dating and OSL dating. The investigated sites included seven in the Lejrelv area and six in the Gåseelv area (Table 1). Positions and elevations were determined by a handheld global positioning system (GPS) device (uncertainty ±10 m). Our elevations appear to differ from those described for the same landforms by Lilliesköld & Salvigsen (1991), but the difference is within the uncertainty of the GPS device. In this study, we consistently use our elevations.

A total of eight samples were collected, six of which are amalgamated pebble samples. Each pebble sample contains 20 clasts of approximately the same size. Sample thickness is equal to the average diameter of pebbles. Samples were collected from flat, well-drained surfaces about 4×4 m in area. Shielding from the surrounding horizon was measured with a compass and clinometer. Samples were processed for ¹⁰Be analysis at the University at Buffalo following procedures modified from Kohl & Nishiizumi (1992) and Briner (2003). About 40 g of clean quartz from each sample was dissolved in batches of 11 and one process blank (Table 3). Known amounts of SPEX CertiPrep (Metuchen, NJ, USA) reference standard ⁹Be carrier solution (1000 ml/l) was added to all samples and to the process blank. ¹⁰Be/⁹Be ratios were measured at the Tandem Laboratory at Uppsala University. The CRONUS-Earth exposure age calculator, version 2.2, was used to calculate model ages (Balco et al. 2008; http://hess.ess.washington.edu/math/). This version of the calculator uses the revised ¹⁰Be decay constant of 5.10±0.26 x 10⁻⁷ yr⁻¹ (Nishiizumi et al. 2007). Ages are presented in Table 3 and have been calculated with a regionally calibrated ¹⁰Be production rate for north-eastern North America (Balco et al. 2009), which overlaps with ¹⁰Be production rates reported from West Greenland (Briner et al. 2012). The scaling scheme by Lal (1991) and Stone (2000) and a constant production rate model (2σ, sea level, high latitude and standard atmosphere) have been used. Correction was made for altitude, shielding by the surrounding horizon and for sample thickness using exponential decrease in nuclide production using a density of 2.66 g cc⁻¹. All ages are calculated with zero erosion. We have chosen to do so because all collected pebbles have smooth surfaces and consist of resistant lithologies such as quartz or quartzite and postglacial erosion is considered to be minimal.

Ten samples of glaciofluvial and glaciolacustrine sediments were collected in opaque plastic tubes that were hammered into dug-out sections. In addition, we have a sample of similar deposits from a nearby location in the Mestersvig delta (Fig. 1; Håkansson, unpubl. data), which serves as a control sample since we have independent age information from that site (see below). The tubes were stored in black plastic bags until opened under subdued orange light in the laboratory at the Nordic Laboratory for Luminescence Dating, Risø, Denmark. The fraction 180–250 µm was extracted and treated with 10% HCl, 10% H₂O₂ and concentrated HF according to standard procedure, to get pure quartz. Large (8 mm) and small (2 mm) aliquots were analysed using a single-aliquot regenerative-dose protocol (Table 2; see. Murray & Wintle 2000, 2003) on automated Risø TL/OSL readers (Risø, Technical University of Denmark; Bøtter-Jensen et al. 2000). Equivalent doses were calculated using the Risø software package Analyst, version 3.24; only aliquots with test dose error <10% and a recycling ratio within 10% of unity were accepted. Environmental dose rates were determined by gamma spectrometry (Murray et al. 1987), and the cosmic ray component was included according to Prescott & Hutton (1994). Water content was measured by weighing the samples with their natural water content, when saturated and when dried (Pusch 1973). A water content value halfway between the water content at the time of sampling and the saturation value was used for age calculations.

Four radiocarbon ages from or near our sites were retrieved from the published literature to provide independent chronological control for our ¹⁰Be and OSL ages and to add to our reconstruction (Table 2). Three of these samples (Lu-647, -712 and -764) are from Kong Oscar Fjord, north of our study areas (Fig. 1). Sample Lu-764 gives the age of the local marine limit in Fleming Fjord close to the outer coast (35–60 m a.s.l., Håkansson 1974; Hjort 1979). Samples Lu-647 and Lu-712 (Håkansson 1973, 1974) were taken close to Mestersvig and provide constraints on the deglaciation of this site. Sample K-1916 (Funder 1978) provides a minimum age of the deglaciation of southernmost Klitdal and likely a close age of the marine limit in the inner Hurry Fjord, which we have sampled for OSL at site 13 (Fig. 2).

The radiocarbon ages have been calibrated by us in the OxCal 4.1 program using the IntCal09 calibration curve (Bronk Ramsey 2009; Reimer et al. 2009) after correcting the marine samples for a reservoir age of 550 years (Hjort 1973).

On the southern side of the Umingmakelv and Draba Sibiricaelv valleys, there are discontinuous lateral meltwater channels at levels between ca. 500 and ca. 200 m a.s.l. (Fig. 2a, 3a). The channels dip 1–1.5° towards the west.

Sites 1 and 2 are on segmented ridges, ca. 20 m tall, found at ca. 300 m a.s.l. in the Umingmakelv Valley and at ca. 290 m a.s.l. in the Lejrelv Valley (Fig. 2, Fig. 3a, b). The ridges form arc-shaped features in the valleys and their surfaces are covered mainly by local sandstone cobbles, but a few crystalline and quartzite erratics <0.3 m in diameter occur. Some of the sandstone clasts are ventifacts. These ridges are interpreted as moraines, and they have subsequently been cut by the river (Fig. 2a, 3a–c). Crystalline and quartzite erratics are also scattered on the Skansen (site 3; Fig. 3d, e) and Centralbjerg mountain plateaus (site 4; Fig. 3f).

Sites 5 and 6 are river sections situated at two of several flat-topped landforms that occur in the Lejrelv Valley on levels between ca. 300 and 250 m a.s.l. along the Lejrelv River (Fig. 2a, 4a; see Lilliesköld & Salvigsen 1991). These sites are situated inboard of the moraine ridge (site 1). At site 5, the landform's top surface is at 295 m a.s.l. and sediment thickness is ca. 8 m. Sediments show a coarsening-upward succession from ripple-laminated sandy silt and sand, over planar cross-bedded sand to massive gravelly sand, finishing with a gravel lag on top (Fig. 4b). The beds dip up to ca. 15° towards the south-east, out of the valley. Site 6 at 250 m a.s.l. exhibits a sediment thickness of ca. 10 m, most of which was difficult to access due to scree cover. In the upper part of the sediment succession, the beds dip ca. 10° towards the south-east, i.e., out of the valley (Fig. 4c). They are dominated by ripple-laminated and planar parallel-laminated sand, with laminated silt at the base of the logged section.

Silty–sandy sediments are exposed in a river section between ca. 240 and 250 m a.s.l. at site 7 (Fig. 4d) in the area between the Lejrelv and Umingmak valleys and the Umingmak and Skansen mountains (Fig. 2a). This site is situated outboard of the moraine ridges at sites 1 and 2. Near the base of the sedimentary section, clayey and silty massive beds alternate with ripple-laminated sand. Upwards the succession is dominated by planar parallel-laminated and ripple-laminated sand, with a slight coarsening-upward trend. The beds dip ca. 10° towards the south-west, towards the south and the Umingmakelv Valley.

The finer, laminated sediments that dominate the successions at sites 5–7 are interpreted as being deposited by low-density turbidity currents along gentle delta slopes. The thin layers of clay and silt at site 7 were deposited during periods of breaks in sediment input from the delta slopes, likely in a deeper part of a lake. The massive gravelly sand at the top of sites 6 and 7 is interpreted as a supra-aquatic deposit, likely formed through alluvial processes, and which has later been affected by deflation to form a lag on the present surface.

Based on the geomorphology and the sediments, the landforms at sites 5–7 are interpreted as glaciolacustrine deltas. The sediment successions represent infilling of glacial lake basins in the Lejrelv Valley and, on the other side of a bedrock threshold, in the intervalley area (Fig. 2a). For lakes to form in these areas, glaciers from Liverpool Land must have dammed the present day mouths of the Umingmak and Lejrelv valleys. However, as shown by the palaeocurrent measurements, the main sediment input was not from these glaciers but from the head area of the river Lejrelv.

Sites 8 and 9 are situated on a prominent wide flat-crested ridge (ca. 130 m a.s.l.) that stretches across the Gåseelv Valley (Fig. 2b, Fig. 5). Its surface is covered mostly by local sandstone but crystalline pebbles and cobbles are common, and a few small crystalline and basalt boulders are scattered on the surface. The ridge is interpreted as a moraine. Funder (1990) mapped another moraine at ca. 400 m a.s.l. on the southern valley side, but in this study we note that this ridge is a bedrock structure.

Site 10 in the innermost Nathorst Valley is a sediment accumulation with a slightly elongated shape and its highest point at ca. 180 m a.s.l. The landform has an uneven top surface, which is gently sloping downvalley (Fig. 6a–c). The surface is covered with a lag that contains particles up to 40 cm in size, but most are 2–5 cm in diameter. A sediment section in the lower, southern part of the landform is dominated by sub-horizontal planar parallel-laminated sand. Ripple-laminated sand becomes increasingly common upwards in the sequence. Palaeocurrent directions are variable, but are mainly from the surrounding mountain slopes. The sediments below the logged section, which are largely covered by scree, appear to be gravelly. The logged sediments are interpreted as deposited by running water, likely in rather shallow channels on a glaciofluvial fan. This interpretation is supported by the palaeocurrent directions from the surrounding mountain slopes and by the gently sloping top surface of the landform. The landform may be related to a similar feature on the western side of the valley (Fig. 6a).

Downvalley of site 10, there is a flight of flat-topped landforms at levels from ca.160 to 110 m a.s.l. (Fig. 2b, 6a, d–f; see Lilliesköld & Salvigsen 1991). The sediments in these landforms are represented by a short section in one of the middle-level deltas, site 11, at 154 m a.s.l. (Fig. 6d–e). The beds, which are mainly of planar parallel-laminated medium and coarse sand, dip ca.15° towards the south, out of the valley. A few beds with ripple-laminated fine–medium sand and massive silt also occur. Below the logged section, there was coarse sand which appeared to be cross-bedded. A fossil ice wedge was found in finer sediments close to the top. Similar sediments were observed at the mouth of the valley at site 12, 110 m a.s.l. (Fig. 6f). The sediments are interpreted to have been formed mainly by high- and low-density turbidites flowing along a subaqueous slope. The flat-topped landforms in the Nathorst Valley are interpreted as glaciolacustrine deltas in lakes dammed by a Liverpool Land glacier. The highest delta was deposited as the glacier front was at or close to the Gåseelv moraine. The successively lower delta levels between sites 11 and 12 represent a lowering of the water level in the ice-dammed lake as the glacier retreated from the moraine.

The lowermost landform described is site 13, located at the mouth of the Gåseelv Valley at 33 m a.s.l. (Fig. 2b, 7a). The sediments show a coarsening-upward trend from massive silt and sand over planar parallel-laminated sand to clast-supported massive gravel (Fig. 7b). The sandy beds dip ca.12° towards the east and the Hurry Fjord. The sediment succession is interpreted as representing a transition from turbidites and gravity flows on a delta slope to a glaciofluvial delta top. The gravel lag at the top is caused by deflation, and soil processes have altered the uppermost decimetres below the ground surface. We suggest that this delta at 33 m a.s.l. represents the marine limit in the inner Hurry Fjord.

A total of eight samples were collected from mountain plateaus and from the crests of moraine ridges in east-facing valleys on eastern Jameson Land and analysed for cosmogenic ¹⁰Be (Table 3).

In the Lejrelv area (Fig. 2a), a quartzite boulder perched at ca. 600 m a.s.l. on weathered sandstone bedrock on the Centralbjerg mountain plateau gives an age of 111.1±7.1 Ky. Two pebble samples from the southern plateau edge of the Skansen mountain at ca. 500 m a.s.l. give ages of 13.3±2.0 and 11.2±2.4 Ky (site 3; Fig. 3d–e; Table 3). Pebble samples from the segmented Lejrelv- and Umingmakelv moraines at ca. 300 m a.s.l. are 6.0±0.5 and 9.5±1.1 Ky, respectively (site 1–2; Table 3).

In the Gåseelv area (Fig. 2b), one single cobble and two pebble samples were collected from the surface of the prominent Gåseelv moraine at ca. 130 m a.s.l. The single cobble gives an age of 4.3±0.5 Ky and ¹⁰Be ages from pebble samples are 8.5±1.2 and 12.1±3.3 Ky (sites 7–8; Table 3).

The quartz has a clear fast signal component but is not very bright; the natural signals integrated over the first 0.8 s average ca. 4500 counts for small aliquots and about 30 000 counts for large aliquots (40–1000 counts/Gy). Tests with known laboratory doses show that the selected protocols are able to recover given doses within 10% (dose recovery ratio 0.91±0.01, n=45). Most results show significantly skewed dose distributions (Fig. 8a) and following the decision protocol of Bailey & Arnold (2006), we use the minimum age model (MAM) with three parameters (Galbraith et al. 1999; Excel macro from Sebastien Huot) to calculate the final dose for these samples. The mean ages range from ca. 10 to 50 Ky (doses 20–90 Gy) and for single samples small and large aliquot ages agree within 2σ (Table 4). The MAM-3 ages have a tighter age span (10–35 Ky), but the agreement between small and large aliquot ages is not as good (Table 4). It should be noted that the proportion of aliquots used for the MAM-3 age calculation is low for some of the small-aliquot samples (p-values in Table 4), which indicates an increased uncertainty of the accuracy.

The position of the quartzite boulder at ca. 600 m a.s.l. on the Centralbjerg plateau (site 4) and the crystalline erratics at ca. 500 m a.s.l. on the Skansen plateau (site 3) indicate minimum thickness of ice when this material was deposited. The lateral meltwater channels, at levels between ca. 500 and ca. 200 m a.s.l., reflect a successive lowering of an ice surface and an eastward retreat of the ice margin, towards Liverpool Land.

The moraines in the Umingmakelv and Lejrelv Valleys (sites 1–2) mark a younger ice-marginal position, situated east of the water divide at ca. 300 m a.s.l. This ice configuration led to a lake being dammed between the Umingmakbjerg, Centralbjerg and Skansen mountains (Fig. 2). As the ice margin retreated further towards the east, ice-dammed lakes formed in the major valleys at successively lower levels (300–250 m a.s.l.). Deltas (sites 5–7) were deposited in these lakes and palaeocurrent directions show that the material in the deltas did not come from the ice in the east, but was transported from surrounding areas (Fig. 2a, 4).

When the ice stood at the moraine in the Gåseelv Valley (130 m a.s.l.), a lake was probably dammed between the ice margin and the water divide 13 km to the west. Glaciolacustrine sediments mapped on the valley floor by Lilliesköld & Salvigsen (1991) support this interpretation (Fig. 2b). A lake was also dammed in the tributary Nathorst Valley to a level at least corresponding to the level of the site 11 delta (ca. 154 m a.s.l.). As the ice margin retreated towards the east, the lake was successively lowered; the lowest delta in the Nathorst Valley is at ca. 110 m a.s.l. When the valley was entirely deglaciated and open towards Hurry Fjord, a delta (site 13) was formed at the marine limit (ca. 33 m a.s.l.).

In the studied valleys and on the adjacent mountain plateaus, only few stable-looking boulders were observed and we have therefore chosen to collect pebble samples each of which is made up of 20 clasts. In a study in Colorado, Briner (2009) compared boulder and pebble ¹⁰Be ages from the same moraines. It was shown that on Colorado moraines the two sample types produce indistinguishable exposure ages and consequently it was suggested that pebble collection on moraine crests can serve as a suitable sample type in some settings.

It has been suggested that the oldest exposure ages from moraines should be regarded as minimum constraining ages of moraine ridge formation, because degradation of landforms leads to exhumation of boulders and clasts at the surface (Putkonen & Swanson 2003; Heyman et al. 2011). Our data set is scattered and exhibits a large range of exposure ages from the moraine ridges on eastern Jameson Land (4.3–12.1 Ky, Table 3). This scatter is most probably due to incomplete exposure of pebbles caused by exhumation of clasts from the moraine and we therefore regard 12.1±3.3 and 9.5±1.1 Ky, the oldest ages, as minimum constraining ages of the Gåseelv moraine (sites 8–9) and the Lejrelv–Umingmakelv moraines (sites 1–2), respectively.

It should also be noted that crystalline and quartzite erratics on Jameson Land can originate either from sources west of Scoresby Sund or from Liverpool Land (Fig. 1). Those of our sampled erratics that are from moraines can by geomorphology be inferred to derive from Liverpool Land (see Fig. 2), but the sampled erratics on the mountains in the Lejrelv area are not related to any landforms indicating from which direction they were deposited. However, previous studies suggest that the last time the Greenland Ice Sheet deposited material on central and eastern Jameson Land was during the penultimate glacial cycle, about 160 Kya (Möller et al. 1994; Håkansson et al. 2009). In contrast, during the last glacial cycle Liverpool Land-based glaciers advanced onto eastern Jameson Land (Adrielsson & Alexanderson 2005) and local ice caps with limited erosion potential covered and shielded central Jameson Land for substantial periods (Håkansson et al. 2009; Håkansson et al. 2011). Due to shielding by non-erosive ice, erratics deposited ca. 160 Kya on Jameson Land will be expected to have exposure ages younger than the initial deglaciation age.

In glacial settings like our study area, incomplete bleaching of sediments can be a problem for OSL dating (e.g., Fuchs & Owen 2008). Based on comparisons between quartz and feldspar luminescence ages and radiocarbon ages, Hansen et al. (1999) showed that incomplete bleaching may be of the order of a few thousand years for Holocene deltaic sediments in this area. In comparison, modern fluvial sediments on Jameson Land seem better bleached with only a ca. 500-year age overestimate (Alexanderson 2007). Although the sediments we sampled in the deltas did not come directly from beneath the glaciers, but rather were transported towards them from the surroundings, the risk of incomplete bleaching should not be overlooked. Typical results of incompletely bleached deposits are skewed dose distributions, age overestimation and stratigraphically scattered ages, all of which can be checked for.

In this study, studies of dose distributions and comparisons with independent age control showed that our samples do suffer from incomplete bleaching. Dose distributions from large aliquots are not very useful for this purpose, since too much averaging occurs between the many grains in large aliquots to reliably recognize different dose populations of grains (Duller 2008), but the data from our small-aliquot measurements show that dose distributions are significantly skewed (Fig. 8). The mean dose therefore does not give a representative value of the true equivalent dose, and thus the age, but rather overestimates it. Our mean OSL ages, both for large and small aliquots, must therefore be regarded as maximum ages of the time of deposition.

To get a more representative estimate of the true dose (age) from skewed distributions, some statistical treatment, such as the MAM (Galbraith et al. 1999), is required. The good correspondence between MAM-3 ages and independent age control shows that the MAM-3 provides reliable ages, at least for small aliquots, despite the uncertainty due to low p-values (Table 4). The OSL ages of the deltas at Mestersvig (11±1.4 Ky) and in Gåseelv (11±1.3, 12±2 Ky; Table 4) agree well with the radiocarbon ages that provide close ages for their formation (11.1–10.2 and 11.2–10.6 calibrated thousand years [cal Ky] B.P., respectively; Table 2). The MAM-3 ages from large aliquots do not show such good correspondence, which is expected due to averaging between grains (see above), and because the number of aliquots per sample is not enough for statistic certainty.

The implication is that mean ages from large aliquots of samples like these may overestimate the true age by up to ca. 20 Ky, although commonly less (average 7 Ky for our 10 samples; Table 4). Mean ages from small aliquots may also overestimate the true age (see Fig. 8a), but statistical methods such as the MAM-3 can be successfully applied to yield good results. Of our OSL ages (Table 4), we therefore find the small-aliquot MAM-3 ages most reliable.