Combined optically stimulated luminescence and radiocarbon dating of aeolian dunes in Arctic Sweden

Multiple parabolic sand dune fields formed in Arctic Sweden after the last deglaciation, facilitated by an abundance of loose glaciofluvial sediment, limited vegetation cover and strong winds. Following initial stabilisation, these dunes underwent repeated reworking after fire events, as evidenced by the presence of buried soils, charcoal layers and redeposited sands in the dune stratigraphy. These reworking events may be driven by wider climate forcing; however, to date, no chronological framework exists for this activity in Sweden. As such, here, we apply quartz optically stimulated luminescence (OSL) dating of Arctic Swedish sand dunes using two dunes at the sites of Vastakielinen and Jorggástat. Resultant double‐SAR (single aliquot regenerative dose) quartz OSL ages are in good agreement with independent ages provided by 14C dating of charcoal fragments recovered from charcoal layers within the dunes, and we conclude that the chosen protocol is generally well suited for dating aeolian reworking of dune sediments in Arctic Sweden. While feldspar contamination limits precise age assignment for initial dune movement, our results nonetheless suggest repeated and long‐lasting aeolian activity in Arctic Sweden throughout the Holocene and, although there are differences in detail, further suggest some general trends in terms of dune stability and reworking over Arctic Fennoscandia.

presence of blowouts and hollows indicates periods of disturbance in some locations. 3,4Furthermore, in many Arctic Fennoscandian dunes, stratigraphic evidence points to repeated phases of aeolian activity and stability during the Holocene (i.e., previous studies 3,5 ).The causes of these events are not fully agreed upon but likely relate to fires causing periodic dune destabilisation, potentially driven by wider climate changes or anthropogenic influence. 3,6As such, these dunes may act as palaeoenvironmental archives, documenting the succession of stabilisation and reactivation throughout the Holocene.A typical dune stratigraphic succession exhibits laminated sands (initial dune formation and stabilisation), palaeosols (stability), charcoal layers (fires) and massive sands (reactivation). 5,7e periglacial environment and deglacial landscape of the last glaciation were largely free of vegetation and as such offered ideal conditions for sand dune formation.][10][11] Most prominent in terms of form are currently stabilised parabolic dunes, 4 which likely developed under the influence of early vegetation establishment.Being tied to the presence of vegetation, this dune type is especially sensitive to changes in its environment, reacting strongly to variations of the local temperature, precipitation, wind regime or water table. 12 utilise these dune records in order to infer wider environmental changes requires independent chronometric dating of multiple, diverse dune archives.4][25][26] Recent semi-automated mapping of Arctic Swedish dunes documents a wide range of simple and complex inland parabolic dune fields in the Swedish Arctic, 4 yet these remain largely unexplored in terms of their age and reactivation.
Luminescence dating allows reconstruction of the timing of the last exposure of sediment to light and therefore the last stages of aeolian activity prior to deposition, ideal for examining aeolian histories.
Thus, here, we aim to develop a broadly applicable quartz optically stimulated luminescence (OSL) dating protocol for Arctic Swedish dunes to address the scarcity of chronometric ages for Holocene dune activity.This is complemented and tested by radiocarbon-dated charcoal fragments, facilitating development of a chronostratigraphic framework for phases of dune stability and activity and, therefore, paving the way to understanding the Arctic Swedish sand dune response to Holocene climatic fluctuations.Prior to this, here, we review previous independent dating of Arctic Fennoscandian dunes.

| DATING OF AEOLIAN DUNES IN FENNOSCANDIA
Research conducted during recent decades on dunes in the wider region (Arctic Finland, Arctic Sweden and southern/central Sweden) has provided some insights into the chronostratigraphy of Fennoscandian dunes (Figure 1), but questions regarding the cause and timing of dune stabilisation and reactivation remain, and the range of well-dated dune sites is unlikely to be sufficient to fully constrain the causes of reactivation. Seppälä 21 and Lundqvist and Mejdahl, 26 though, place the stabilisation up to 3.5-4 ka after deglaciation (c. 10 ka), depending on the area (Figure 1), and based on infrared stimulated luminescence (IRSL) ages from Finnish Lapland sand dunes, Clarke and Käyhkö 19 suggest an even longer phase of primary deposition, well into the mid-Holocene, c. 5-6 ka after deglaciation.This wide range of estimates suggests that stabilisation may not have been a simultaneous event throughout Lapland, but rather dependent on the F I G U R E 1 Ages of stratified and redeposited (massive) sand dune layers, as well as cover sands in Arctic Sweden (green), Arctic Finland (yellow) and southern and central Sweden (red), as reported in published research and the present study.Two ages each from Alexanderson and Fabel 13 and Alexanderson and Bernhardson 14 represent the mean of multiple ages from one sample published in the respective publications.regional vegetation pattern.Sites with favourable conditions for early vegetation may have stabilised shortly after deglaciation, while others remained active for up to 2 ka longer. 3,5,19ter the initial stabilisation, evidence for alternating phases of activity and stability throughout the Holocene is provided by welldeveloped soil horizons buried under reactivated sand layers, often separated by charcoal bands. 21Again, there is distinct variability in estimates of the timing of reactivation events.Notably, in Finland, differently vegetated sites seem to show diverse Holocene histories, with pine woodland sites showing much more evidence for repeated activity in relation to forest fire events than those in the birch woodland or tundra zones. 2,3,5Despite this variability, some patterns emerge when the data are examined as a whole (Figure 1).For example, age clusters for episodes of aeolian activity occur around 7-7.5 ka as well as during the past few centuries for all areas.The latter episode was likely triggered somewhat simultaneously, possibly by climatic deterioration during the Little Ice Age. 2,5,7In several dune fields, aeolian activity from this episode appears to still be ongoing, evidenced by large active hollows and blowout surfaces.Furthermore, publications from the Finish Arctic provide evidence for furtherpotentially localised-aeolian activity in between the 7-7.5-ka and present-day events. 2,3th North Atlantic climate variability and human activity, especially in connection with fire activity, have been implicated in these reworking phases. 3,6,27,28Yet, forest and tundra fires are rather localised events, and the intensity of a single fire may vary locally.Consequently, a straightforward correlation between charcoal layers of different dune fields or even within the same dune field or dune can be difficult, 3,21 and local triggers of fire and dune reactivation may vary.As such, it is critical to carry out much more extensive age dating work on a range of dunes over a wide area, if any wider forcing agents are to be deduced.In particular, Arctic Swedish dunes have seldom been dated, with Lundqvist and Mejdahl 26 providing the only exception.Their thermoluminescence (TL) and early OSL ages from laminated sands from a range of dunes in northern Sweden suggest initial dune formation immediately following local deglaciation (c. 10 ka).
However, with the study focussing on only one sample per site and being conducted prior to key developments in luminescence dating that have substantially improved precision and accuracy, interpretation of the ages is limited, especially because some ages suffer from substantial underestimation.

| LUMINESCENCE CHARACTERISTICS OF SWEDISH DUNE SANDS
The luminescence characteristics of Swedish quartz samples have been evaluated by previous studies. 29,30Luminescence signals of quartz separates from Sweden are generally characterised by rather low intensities, at times with limited fast component. 30This correlates well with the underlying bedrock and can thus be traced back to the short transport history and igneous origin of the material. 30,31Dunes in Arctic Sweden are generally preserved over the Svecokarelian basement, which exhibits relatively moderate signal strengths but variable fast component dominance. 30Low sensitivity typically results in large equivalentdose (D e ) uncertainties and therefore in age estimates of low precision, which, below a certain level, prevent dating altogether.Due to the low signal intensities, often 'large aliquots' comprising several hundred grains are used for dating, resulting in a higher precision with average errors of 4%-7%. 24,29,30Even though large aliquots do not allow for proper detection of incomplete bleaching, aeolian sediments can be expected to be well bleached even at high latitudes as transport likely occurs when ground is unfrozen during lighter seasons. 19 addition to potential low sensitivity and weak fast components, Alexanderson and Murray 29 report feldspar contamination for some quartz samples of southern-central Swedish dune sand.The influence of the infrared (IR)-sensitive feldspar traps on the green/blue stimulated luminescence is, however, typically successfully reduced by the implementation of a double-SAR (single aliquot regenerative dose) protocol 32,33 during measurement.Further considerations in luminescence dating of these dunes include a low thermal pre-treatment of younger samples and higher preheating for older samples, as well as the use of early background subtraction to isolate the fast component where it is not dominating. 13,29Despite these challenges, quartz from central and southern Swedish dunes has been shown to be generally suitable for OSL dating, especially if sourced from the Dala sandstone.However, it is unclear whether this is also the case for dunes further north in Arctic Sweden, derived directly from sediments generated from Svecokarelian or Caledonian bedrock.

| STUDY SITES AND SAMPLING
Two dune fields located close to the modern treeline were investigated in our study: Vastakielinen (67 44.511 0 N, 21 38.737 0 E, 306 m asl, just north of the town of Vittangi) and Jorggástat (68 17.004 0 N, 21 26.131 0 E, 434 m asl, approximately 70 km further north) (Figure 2).Both fields host parabolic and to a lesser extent transverse dunes, with large lunate dunes oriented towards the south-east (SE) characterising the Vastakielinen dune field, while smaller hairpinshaped dunes oriented towards the east-south-east (ESE) define the Jorggástat field. 4Judging from the dominant dune orientation, topographic controlled north-west (NW) winds, which today are common for the area, shaped the dunes, with destabilised sandy eskers to the NW of the dune fields being the likely sediment source. 4With rapid deglaciation from north-east (NE) to south-west (SW), Vastakielinen and Jorggástat were ice free at approximately 10 and 10.1-10.2cal ka BP, respectively, 37 and vegetation likely became established soon after. 3,19At present, the Jorggástat site is vegetated mainly by shrubs and sparse standing mountain birches, while Vastakielinen, located further away from the treeline, is characterised by mature pine trees and a thick ground cover of lichen and smaller shrubs.Age estimates for dune formation and reactivation are entirely lacking, with the noteworthy exception of a TL age of 7.6 ± 0.7 ka as proposed by Lundqvist and Mejdahl 26 for the stratified sands of one of the Vastakielinen dunes.
Here, we examined the spatial variability of dune stratigraphy through the digging of multiple pits throughout a variety of dunes at both sites and then conducted sampling of representative sections for independent dating.Dug pits and sections displaying typical signs of reactivation-loose massive sand above the stratified core, buried palaeosols and charcoal horizons-were then chosen for sampling.The base of each sampled profile was indicated by the presence of stratified sands, which are widely recognised as representing the original dune movement following deglaciation (i.e., previous studies 2,3 ).
Luminescence samples were taken from the stratified sands, massive sands and palaeosols, as well as from the massive sand unit exposed right beneath the surface, while charcoal fragments for radiocarbon dating were taken from charcoal bands that bracket massive sand units.
Eight OSL and one 14 C samples were retrieved from one representative section in the Vastakielinen dune chosen for investigation, sampling all representative units down to the stratified sands at 90-cm depth.At Jorggástat, OSL samples were taken from four positions in a representative section, and 14 C samples were also collected from two buried charcoal horizons.OSL samples were retrieved using steel cylinders of 4-5 cm in diameter and placed into opaque plastic bags directly after extraction.In addition to the OSL samples, further material was taken for measurement of the water content and dose rate at each OSL sample depth.For 14 C dating, $50 g of material was extracted from all layers with expected charcoal presence.All OSL samples were measured at the luminescence laboratory of the Justus-Liebig-University Giessen, Germany, while 14 C sample preparation and measurement was conducted at the Tandem Laboratory, Uppsala University, Sweden.

| Luminescence dating
Sample preparation took place under amber light conditions using a standard preparation procedure including wet sieving of the samples to the desired grain size of 90-200 μm, treatment with diluted F I G U R E 2 (A) Location of the Vastakielinen and Jorggástat sites within the Kiruna municipality of Norrbotten (50-m DEM 34 ) and in (B) and (C) shown in the context of dune locations and Quaternary deposits identified by the Geological Survey of Sweden (SGU) 35 and modified from Stammler et al. 4 Blue dune polygons were accepted from the geographic object-based image analysis (GEOBIA)-classified potential dune dataset (grey) published by Stammler et al. 4 on the basis of the underlying elevation grid 36 with red dots indicating dune locations provided by SGU. 35D) and (E) provide close-ups of the two dunes sampled in the present study with elevation derived again from Lantmäteriet.36 All data are projected in SWEREF 99 TM.hydrochloric acid (HCl) and hydrogen peroxide (H 2 O 2 ) (both 10%) to remove carbonates and organic compounds and repeated density separation in a solution of lithium heteropolytungstate (LST Fastfloat) with densities of 2.68 and 2.63 g cm À3 to extract quartz.The quartz fraction was further etched in 40% hydrofluoric acid (HF) acid for 45 min to remove the alpha-irradiated outer rim of the grains and any remaining contaminants and, following this, was washed in 10% HCl and once again sieved (dry) to 90-200-μm grain size.For measurement, the quartz grains were fixed to stainless steel cups using silicone oil.
A double-SAR protocol comprising infrared (875 ± 80 nm) and green-light (525 ± 25 nm) stimulation (Table 1) was performed on a Freiberg Instruments Lexsyg OSL Smart reader equipped with a calibrated (calibration using LexCal 2014 3.0 Gy-0017) 90 Y/ 90 Sr-beta source providing a dose rate of $0.13 Gy s À1 . 38,39Luminescence signals were detected by a Hamamatsu H7360-02 photomultiplier, equipped with combinations of a 3-mm Schott-BG 3 and a 5-mm Delta-BP 365/50 EX-Interference filter during green stimulation, as well as a Schott-BG 39 and a 3.5-mm AHF-Brightline HC 414/46-Interference filter during infrared stimulation.Samples were initially screened using a standard SAR protocol.However, contamination by feldspar within the quartz fraction of some samples necessitated IRSL signal resetting prior to green stimulation, and further measurements were conducted using the double-SAR procedure [40][41][42] outlined in Table 1.For dose recovery tests, aliquots were first bleached in the reader at 125 C for 75 s, followed by β-irradiation to a known dose close to the expected D e .Subsequently, the D e was determined using a double-SAR protocol with different preheat/cutheat temperature combinations of 160/160, 180/160, 200/180, 220/200, 240/220 and 260/220 C. Preheat plateau tests were conducted using the same temperature combinations.In addition to screening, dose recovery and preheat plateau tests, all samples were subjected to gradual heating to 380 C with a heating rate of 5 C s À1 to investigate the sample-specific TL response in the context of the identified feldspar contamination.As samples J1-6 and J1-59 did not show any OSL response during screening measurements, they were additionally examined at high given doses, for which three aliquots of both samples were bleached, and the bleached OSL response was recorded.Subsequently, the response of aliquots to regenerative beta-doses significantly higher than the expected natural dose (up to 62.5 Gy) was measured using the double-SAR procedure, to determine their general luminescence sensitivity.
For D e determination, OSL signals were recorded for 50 s at 125 C. Natural, regenerative and test dose signals were derived from the initial 0.5 s after subtracting a late background signal based on the last 10 s of stimulation.Early background subtraction, due to the dim nature of the signals, was not possible as it led to excessive rejection rates.A single saturating exponential fit was used to define dose-response curves.Rejection criteria were set to 10% deviation from unity for the recycling ratio, to 10% for the test dose error and to 5% for the recuperation level.The threshold for the maximum palaeodose error was raised to 15% due to high rejection rates at 10% (see Section 7.3).For D e determination and statistical handling, various functions provided by the R package 'Luminescence' [43][44][45] were used.Due to the mostly unskewed nature of the resulting D e distributions, low overdispersion values and methodological transparency, the weighted mean was utilised for the final D e calculation.Sample specific dose rates for OSL age calculation were determined by combined alpha-and beta-counting using a μDose system [46][47][48] at the Giessen Luminescence Laboratory.Following the preparation procedure described by Kolb et al., 46 the sample material was dried at 105 C for 24 h, gently crushed as well as homogenised and finally pulverised in a ball mill (Retsch M400; 29.5 Hz for 45 min).For each sample, 3 g of pulverised material was mounted on sample carriers of 70-mm diameter, which were stored in gas-tight measurement containers.The cosmic-ray contribution was derived according to Prescott and Hutton. 49Based on measurements of the range and variability of in situ water contents, the sedimentological properties and geomorphic settings, sample specific water contents were estimated to range from 5% to 13% and assigned an uncertainty of 5%.Environmental dose rates and final ages were determined using the Dose Rate and Age Calculator for trapped charge dating (DRAC version 1.2 50 ), set to utilise the conversion factors of Guerin et al. 51 ; the alpha-and beta-grain size attenuation factors of Brennan et al. 52 and Mejdahl, 53 respectively; the beta-etch depth attenuation factors of Bell 54 ; and an assumed thickness of the etched layer of 10-20 μm.
T A B L E 1 Double-SAR (single aliquot regenerative dose) sequence based on Roberts and Wintle, 33 as used in this study. Step

| Radiocarbon dating
To add independent age control to the OSL ages, as well as additional information about the timing of reactivation events, charcoal samples from Vastakielinen (V3-10) and Jorggástat (J1- Prior to the determination of the 14 C content in the accelerator, the washed and dried material acidulated to pH 4 was combusted and graphitised in an AGE3™ system, 55 where a tiny fraction of the obtained CO 2 was injected to an online coupled IRMS (isotope-ratio mass spectrometry) to determine the natural mass fractionation Radiocarbon ages were obtained with a MICADAS™ tandem accelerator 56,57 and calibrated using IOSACal v0.5.1. 58| RESULTS To further investigate this signal, all samples were subjected to OSL-IR depletion ratio tests. 59The signals of  Abanico plots showing the distribution of equivalent doses for all samples can be found in Figure S1. 60r the entire sample batch, most D e distributions appear weakly to moderately skewed (skewness between À0.

| Dose rates
Uranium (U), thorium (Th) and potassium (K) concentrations determined in the μDose system are presented in Table 3. U values are slightly higher in Vastakielinen than in Jorggástat, ranging from $1.25 ppm for V3-6 to $2.30 ppm for V3-51.Similarly, the Th contents vary largely within the samples, but there is again a tendency towards higher values at Vastakielinen (up to 6.24 ppm) compared to Jorggástat (up to 3.07 ppm).The K concentrations show little variation, neither within nor between the two sites, ranging between $2.08 and 2.70%.Environmental dose rates calculated with DRAC (see Table 2) are in the range of $2.5-3.6 Gy ka À1 at Vastakielinen and $2.8-2.9Gy ka À1 at Jorggástat.

| Luminescence ages
The luminescence dating results are summarised in Table 2

| Radiocarbon ages
Results of the radiocarbon dating are presented in Figures 6 and 7,

| Signal intensity and luminescence sensitivity
Unfavourable OSL-characteristics like low luminescence sensitivity or few contributing grains are often connected to igneous origin or short recycling history of the grains, 29,31 both of which likely hold true at the studied sites.Thus, the quartz from the study area could be expected to be challenging for luminescence dating, and there is relatively little empirical data from this region to suggest otherwise. 30spite this, screening measurements executed on larger aliquots of 7 mm diameter reveal sufficiently high OSL signals for further investigation and D e determination for all samples except J1-6 and J1-59 (see Figure 3 for examples).
Dosing J1-6 and J1-59 to higher than naturally expected betadoses revealed that even though a slight increase in the signal could be identified with increased dose, this signal was too dim to allow for construction of a reproducible growth curve (Figure 3).The signal intensity of J1-6 does not seem to be influenced by any IR sensitive contaminants, as no IRSL signal could be identified during the measurements.However, J1-59 does show an atypical TL response (Figure 4), possibly due to some contamination.A combination of young age (for J1-6) and low sensitivity likely explains the low natural signals.All Jorggástat dunes show signs of extensive recent reworking in the form of multiple active blowouts, so a recent age of J1-6, the uppermost sample, is feasible.[48] Sample no.U (ppm) Th (ppm) K (%) (Figure 6), likely representing initial dune movement prior to stabilisation, so here the weak luminescence signal is dominantly a function of low quartz sensitivity.This may suggest that reworking of stratified dune core sands during Holocene dune reactivation events acts to increase luminescence sensitivity in this quartz, resulting in stronger signals in reworked homogeneous sand units overlying stratified sand dune cores.In any case, due to this low sensitivity and (for J1-59) unusual TL curve, the two samples were excluded from further measurements.

| Contamination
The atypical TL curves exhibited by V3-58, V3-90 and J1-59 should be free from feldspar grains, which is also supported by microscopic examination of a number of aliquots.However, these measures are not able to remove feldspar microinclusions from the quartz, 64 which, since sample preparation was executed most carefully, are thus deemed to be the source of the infrared contribution.
Even though most samples seem not to be affected much by any feldspar inclusions, as other depletion ratios remained within unity, we find that a dominance of feldspar in 2 out of 10 samples merits the employment of an adapted SAR protocol.Thus, to remove any unwanted feldspar influence on the OSL signal, all equivalent dose measurements were conducted using a double-SAR procedure (Table 1).This is often used for polymineral samples, but also employed by Alexanderson and Murray 29 and Alexanderson and Fabel 13 for Swedish quartz in cases where the presence of feldspar is detected.Judging by good recycling and recuperation for most aliquots, as well as by successful dose recovery and preheat plateau tests (Figure 5), this protocol seems to produce reliable equivalent doses for most samples.Where the feldspar contribution is very high, however, as with samples V3-58 and V3-90, even IR stimulation prior to the OSL measurement might not be able to ensure an OSL signal originating entirely from the quartz component (i.e., previous studies [64][65][66] ).Ages obtained for these samples (Table 2, Figure 6) should therefore be treated with caution, especially that of V3-58, where dose recovery at any tested PH/CH temperature combination was not successful (Figure 5).

| Equivalent dose and age calculation
All rejection criteria were initially applied with standard thresholds (10% for the recycling ratio, test dose and palaeodose errors and 5% for the recuperation ratio), but it was necessary to increase the palaeodose error threshold to 15% due to high rejection rates for V3-51, J1-23 and J1-34.As mean D e remained nearly unchanged with this adaption, we regard it as unproblematic.Fortunately, most aliquots did not exceed recycling and recuperation thresholds, as well as the adapted palaeodose error for the majority of samples.Where recuperation values were larger than 5% (individual aliquots of V3-6, V3-58 and V3-90), this can be traced back to low signal intensity, where the natural and zero dose L x /T x values are similar.This can result from either young age or low residual post-IR signals and can also be regarded as unproblematic. 41,67th low rejection rates, the equivalent dose estimates from accepted aliquots of V3-6 to V3-51, J1-23 and J1-34 (Table 2) can therefore be considered especially reliable, with sensitivity changes properly monitored and corrected for in these samples and thermal transfer at an acceptably low level.Samples V3-58 and V3-90, however, show poor dose recovery, high rejection rates for recycling and recuperation, are generally dim, and the influence of feldspar F I G U R E 7 Calibrated radiocarbon age probability distributions for charcoal samples analysed at the sample sites using IOSACal v0.4.1 and atmospheric data from Reimer et al. 61 Note that the ages shown on the right are the raw uncalibrated input ages.
T A B L E 4 Sample data for radiocarbon samples.microinclusions is not reliably removed.Thus, good reproducibility and reliability of the D e is unlikely.In addition, only a comparatively small number of aliquots were accepted and overdispersion was high, making the resultant age even more problematic.
D e distributions (Figure S1) of all accepted aliquots for each sample provide further information.Almost all distributions appear to be unskewed if statistical outliers are ignored.This is further supported by the proximity of mean and median D e , and low overdispersion values.Despite this being a sign of well-bleached, undisturbed samples, the shapes of the depicted distributions may be somewhat misleading due to averaging effects of large numbers of grains in large aliquots.Fortunately, aeolian quartz grains even from high latitude environments are typically well bleached, promoted by extended sunlight hours during warmer summer months as well as snow cover and/or frozen ground inhibiting aeolian activity during the dark season. 68,69The aeolian sands from the present study area should thus be fairly well bleached, supported also by findings obtained by Clarke and Käyhkö 19 for Arctic Finnish dune sand.
However, the D e distributions attained for V3-58 and V3-90 (Figure S1) contain too few accepted aliquots (Table 2) to make similar interpretations, while the presumably persistent influence of feldspar further complicates the interpretation.Assuming the OSL signal from these samples is still partly due to feldspar contamination (Figure 4), anomalous fading could for example result in an underestimate of the palaeodose.Despite this, the age of V3-90 appears much more reliable than V3-58, fitting rather well into the stratigraphic sequence within its large age uncertainties and exhibiting much lower overdispersion values than V3-58 (Table 2).The age could, however, be older than that accounted for within the error range.

| Chronostratigraphy and independent age control
The reliability of the ages and therefore the robustness of the chosen protocol is further assessed by their specific stratigraphic context, comparison with radiocarbon ages from charcoal bands, and connection to the existing chronostratigraphic framework of Fennoscandian dune ages.
At Vastakielinen, the ages are in stratigraphic order, ranging from youngest to oldest within errors, if V3-58 is ignored, as discussed above (Figure 6).Initial dune formation, as indicated by the oldest reliable age in the profile in the stratified sands (V3-51: 10.28 ± 0.53 ka), confirms a formation age around the proposed timing of deglaciation at 10 cal ka BP. 37 The two stratified sand ages above overlap and indicate ongoing dune movement until at least 8.1 ± 0.4 ka.A welldeveloped podzol indicates a subsequent phase of dune stability, with its timing constrained by luminescence ages from V3-26 providing a minimum age for the start of the stable phase (6.7 ± 0.23 ka) and V3-15 with a maximum age for its cessation (3.65 ± 0.18 ka).The well-developed nature of the podzol soil also suggests a long period of stability, as in Arctic Sweden such soils require at least 500 years to form due to weak soil formation rates. 3 The exact timing of cessation of the initial dune formation before this first stable phase is hard to constrain at Vastakielinen, as the podzol-overprinted stratified sands are less well defined towards the E horizon of the podzol soil, which itself shows no visible signs of stratification.The age of sample V3-26 (6.7 ± 0.23 ka), within the E horizon and therefore the more massive part of the lower sand complex, however, does not overlap with the ages below, indicating the potential existence of renewed sand deposition after initial dune formation, but before the first intensive phase of soil formation and dune stability.
The first homogenous sand unit (V3-15) above the stratified sands suggests an age of 3.65 ± 0.18 ka for the timing of initial dune reworking after the first stable phase (Figure 6).This is followed by a second stable phase, indicated by another well-developed podzol profile overprinting the massive sand unit.The timing of this dune stability is constrained as occurring between 3.65 ± 0. At Jorggástat, the unavoidable exclusion of two samples from the chronostratigraphic dataset largely limits full assessment of the timing of dune activity and stability.Still, the existing ages of 7.83 ± 0.39 ka (J1-34) and 7.94 ± 0.39 ka (J1-23) in the massive sand units immediately above the stratified sands overlap within uncertainty (Figure 6) and suggest sand movement at the time.However, the apparent absence of soil formation or a charcoal horizon between these massive sands and underlying stratified sands make it unclear if this unit represents ongoing initial dune formation or a reactivation event.
Newly established vegetation coinciding with ongoing dune formation may have led to massive sand formation even during initial dune formation, 21

| Wider dune formation and reworking patterns in Arctic Fennoscandia
The published ages from dunes in Arctic Fennoscandia and southerncentral Sweden are summarised in Figure 1.Our new results from Arctic Sweden confirm that initiation of sand transport in Fennoscandia seems to have generally been coincident with the deglaciation of the area, at least from central Sweden northwards (Figure 1).The duration of the initial dune building phase is more complex, however.
At Vastakielinen and Jorggástat, initial dune activity lasted at least until 7-8 ka, and a similar timing is also reported by Seppälä 21 and Lundqvist and Mejdahl 26 for Arctic Fennoscandia.However, some other studies in Arctic Finland 2,3,5 suggest a longer phase of initial dune movement (potentially to 5 ka) that is spatially variable, dependent on vegetation type.However, this is not clearly shown in the data here, nor in the compilation of ages in Figure 1.In southern and central Sweden, cessation of initial dune activity seems to have taken place within 1-2 ka of deglaciation (i.e., previous studies 13,24 ), although some dune ages are several thousands of years younger, and may imply that restricted ongoing dune activity could occur under favourable local conditions. 14In any case, initial dune movement (and reworking phases) seems much more extended in Arctic Fennoscandia.
The timing of stabilisation and reworking phases of dunes in Fennoscandia seems even more complex from the available data, although some general patterns do emerge (Figure 1).For example, age clusters for episodes of aeolian activity occur around 7-7.5 ka, as well as during the past few centuries for all areas.However, evidence from the Finnish Arctic for massive sands at Kuttanen (4.42 ± 1.06 ka), close to the Swedish border.Furthermore, neither site shows evidence for a forest fire as the triggering event.However, there is no evidence for the 2.97 ± 0.57 ka event at Kuttanen in the Vastakielinen dune, showing that even though both dunes seem to have experienced repeated redeposition, partially at similar times, their activity cannot generally be assumed to coincide.The radiocarbon age from the upper charcoal band at Jorggástat coincides with the OSL age of a massive sand unit at Vastakielinen, 3.5 ka, and broadly overlaps with the 2.97 ± 0.57 ka sand deposition event at Kuttanen. 3Charcoal horizons and reactivation events in the studied dunes show varying degrees of correspondence to similar features in other dunes in Arctic Fennoscandia, but the phase of activity over the last few hundred years can be seen throughout the region, even if it does not seem to have started at the exact same time in all dunes (Figure 1).
Overall, the differences between the ages of stabilisation and reactivation of the dunes here, and in other studies (Figure 1), imply caution in interpretation of ages in terms of denoting wider forcing, until more extensive chronologies can be developed.However, it is clear that alternating repeated reworking and stabilisation phases are much more extensive in dunes in Arctic Fennoscandia, than those in central and southern Sweden.This may be due to the denser, more extensive vegetation cover in more southern sites.Interestingly though, coarse silt loess sediments in Värmland in central Sweden also suggest prolonged aeolian activity well into the Middle Holocene, 70 in contrast to the majority of dune ages from the region, 13 implying that aeolian activity throughout Fennoscandia may be episodic in the Holocene.
In general, two main mechanisms have been proposed to explain the timing of dune movement in Arctic Fennoscandia.First, summer temperature and precipitation patterns, which are predominantly controlled by Atlantic Meridional Overturning Circulation (AMOC) as well as the position of the westerlies, are believed to control fire dynamics over Fennoscandia. 6The abundant charcoal horizons underlying most dune reactivation layers suggest that forest or tundra fires played a major role in driving repeated aeolian episodes by inducing large scale or total damage to vegetation, which recovers only slowly in the Arctic environment, giving the opportunity for renewed aeolian activity. 11,21Fire-rich episodes are inferred to occur especially during weakening of the AMOC, resulting in regional cooling over the North Atlantic and southward redirection of the westerlies.The resultant intensification of high-pressure systems over Scandinavia can lead to dry and cool conditions, increasing the likelihood of fire activity. 6ternatively, human impact in the form of land clearance and herding practices 27,28 has been proposed as a trigger of dune reactivation.
Especially the use of fire by early inhabitants to the region to increase the supply of reindeer lichen is believed to have resulted in the increase of local fire activity, as well as a dramatic change in the vegetation cover through grazing pressure, both of which may potentially have triggered wind erosion. 7,27hile the uncertainties discussed above currently preclude fully testing these possible forcing mechanisms in driving dune movement in Arctic Fennoscandia, some observations can be made.Notably, the latest episode of dune movement is widespread, and coincides with climatic deterioration during the Little Ice Age. 2,5,7In several dune fields, aeolian activity from this episode appears to still be ongoing, evidenced from large blowout surfaces.However, this activity also coincides with increased human activity, which has been evoked to explain similar recent dune reactivation in southern and central Sweden. 14The brief phase of dune activity between 7 and 7.5 ka over Arctic Fennoscandia (Figure 1) occurs at a time when the region was experiencing the highest temperatures of the pre-industrial Holocene, 71,72 while initial dune movement occurred during cooler climates.Evidence for general dune stability in the region between 7 and 4 ka does coincide with the longer period of peak Holocene warmth in Fennoscandia, 72,73 potentially suggesting a causal link.
However, testing this requires considerably more extensive and detailed age dating of dunes in the region than has currently been undertaken.

| CONCLUSION
Here, we apply a quartz OSL dating protocol using a double-SAR procedure that allows for reliable dating of most quartz sand samples for two aeolian dune sites in Arctic Sweden.The reliability of the protocol is reduced only in samples that show a significant and dominating contribution of IR sensitive traps to the OSL signal.The IR sensitivity is likely connected to feldspar microinclusions within the quartz crystals that the sample preparation and measurement procedure is unable to remove entirely.Samples recovered from stratified sands in the dune core show low quartz luminescence sensitivity, and this combined with the feldspar inclusion signals limits precise age assignment for initial dune movement and first stabilisation.Nonetheless, our results show that the two investigated dunes likely started forming immediately after deglaciation at 10-10.5 ka, ongoing to at least 7-8 ka.
They also show evidence for repeated and long-lasting aeolian activity throughout the Holocene, interspersed with periods of stability, particularly between 7 and 4 ka, broadly coincident with peak warmth in the region.While there are considerable differences in the detail between the chronostratigraphy of dunes in Arctic Fennoscandia, the new OSL ages generally support the overall trends of aeolian activity in Fennoscandia, and are in agreement with the independent age control provided by 14 C ages for the sites.Together, our new ages and published ages from dunes in Arctic Finland may suggest a coupling to Holocene climate variability, however, much more extensive age investigation of individual dune fields and over a wide area is required to properly test this.Our protocol provides a means to achieve this.
cycle, i = 0 and D 0 is the natural dose.b Aliquot cooled to 60 C after heating.c Aliquot cooled to 40 C after stimulation.d Aliquot cooled to 50 C after stimulation.e Cutheat at 20 C less than preheat temperature (except for a preheat temperature of 160 C and 260 C where the cutheat temperature is 160 C and 220 C, respectively).f L i and T i are derived from the initial optically stimulated luminescence (OSL) signal (Channels 1-5) minus a background (Channels 401-500).

6. 1 |
OSL dating6.1.1 | Luminescence signalsTesting of both 3-and 7-mm aliquots showed that while 3-mm aliquot signals were very dim, all samples except for J1-6 and J1-59 exhibited sufficiently bright luminescence signals for luminescence measurements on 7-mm aliquots.Exposing J1-6 and J1-59 to regenerative beta-doses (up to 62.5 Gy) significantly higher than the expected natural dose (high-dose testing) yielded only a negligible increase in luminescence when compared to the bleached sample (Figure3A).As such, the two samples were excluded from further quartz OSL analyses.Luminescence signals from other samples are sufficiently bright (e.g., V3-26; Figure3B), allowing for further quartz luminescence analyses to be performed.Atypical TL curves for quartz were seen in some samples, notably V3-58 and V3-90, indicating quartz signal contamination.To investigate this TL behaviour, both the natural signal during TL testing and the signal from the preheat after the first regenerative dose were inspected.Atypical TL responses could be identified for V3-58, V3-90 and J1-59 (Figure4), where especially the 110 C peak remained elevated through higher temperatures, again justifying the exclusion of J1-59 from further measurements.
samples V3-6 to V3-51, J1-23 and J1-34 remained close to 10% of unity, while the signals of V3-58 and V3-90 were significantly depleted after IR exposure (ratios of 0.64 and 0.23, respectively), indicative of a significant contribution of IR-sensitive signal components.6.1.2| Dose recovery and preheat plateau tests Dose recovery tests (Figure 5A,C) were carried out for V3-6 to V3-33, V3-58, J1-23 and J1-34 for a range of preheat temperatures, showing the best dose recovery mostly for 180, 220 and 240 C at Vastakielinen and 160, 180 and 200 C at Jorggástat (cutheats tracking 20 C lower).Equivalent dose (D e ) preheat plateau tests (Figure 5B,D) were able to reaffirm the suitability of some of these temperature combinations, together indicating the best double-SAR performance using preheat temperatures of 180 C for younger samples (V3-6, V3-15) and 240 C for older samples (V3-26 and below) at Vastakielinen and 200 C for samples at Jorggástat.Thus, sample-specific preheat and cutheat temperatures (PH/CH) were set to 180/160 C and 240/220 C for F I G U R E 3 (A) Optically stimulated luminescence (OSL) decay curve exhibited by a representative aliquot of J1-59 as a result of laboratory beta-irradiation (regenerative dose) significantly higher (62.5 Gy) than the expected natural dose during high-dose testing.The response to 0 Gy of laboratory irradiation represents the OSL signal of the sample in bleached condition (background signal).Due to the dim response, J1-59 was excluded from dating.(B) Natural OSL decay curve of a representative aliquot of V3-26 (palaeodose $19 Gy).Vastakielinen (V3-6 to V3-15 and V3-26 to V3-90, respectively) and 200/180 C for Jorggástat.Sample V3-58, unfortunately, did not recover the dose well at any given temperature (Figure 5C) but was able to produce a D e preheat plateau around 36.5 Gy, showing a particularly low scatter at 240 C (Figure 5D).6.1.3| Equivalent doses Results of D e determination and sample-specific measurement information are provided in

44 and 1 )
, with two exceptions (V3-51: 1.23 and V3-58: 1.63) (FigureS1).If statistical outliers exceeding 1.5 Â IQR (interquartile range) are excluded, all distributions can be considered unskewed (skewness between À0.24 and 0.33), except for V3-51, with a skewness of 0.8.Weighted mean and median values lie within less than 5% of each other, except for V3-58 (74.96%).For V3-58 and V3-90, however, the relevance of this difference is unclear due to the low number of aliquots passing the acceptance criteria.D e values show an increase in weighted mean and median with profile depth, if V3-58 and V3-90 are not considered.The absolute standard error increases with D e , ranging from <0.01 Gy (V3-6) to 0.66 Gy (V3-51) at Vastakielinen and 0.42 to 0.44 Gy at Jorggástat (J1-34 and J1-23, respectively).Only V3-58 and V3-90 show absolute standard errors of >1 Gy, with 4.23 Gy (V3-58) and 4.22 Gy (V3-90).F I G U R E 4 TL response of V3-51 (A), V3-58 (B), V3-90 (C) and J1-59 (D) during the preheat after irradiation with the first regenerative dose (continuous sample heating of 5 C s À1 ).The dotted line is located at 110 C, the maximum temperature at which a strong TL peak (commonly termed the 110 C peak) can be expected for quartz samples.(A) provides an example of an expected response for pure quartz within the shown temperature range, while (B), (C), and (D) visualise atypical responses identified within the sample batch.F I G U R E 5 (A) Example of the results of a double-SAR (single aliquot regenerative dose) dose recovery test on V3-33 using a given dose of 22.1 Gy.Temperatures of 220 and 240 C with the best dose recovery ratios are shaded brown.The respective preheat plateau test (B) shows only a small scatter at 240 C. (C, D) Example of the double-SAR dose recovery (C) and preheat plateau (D) test of V3-58 with a given dose of 27.3 Gy used for the dose recovery test.Where available, all depicted points represent the mean of up to three measured aliquots, and error bars indicate 1σ standard deviation.

Figure 4 )
indicate the contribution of other minerals to the quartz signal.A TL response typical for quartz is expected to exhibit strong single peaks at 95-110 C, 200-220 C and 305-325 C.62 However, for these samples the TL signal remained elevated up to higher temperatures during heating.A mineral that is abundant in the studied dunes as observed during density separation, and also reported as a F I G U R E 6 Site stratigraphy alongside luminescence and radiocarbon ages and sampling locations of profiles V3 at Vastakielinen and J1 at Jorggástat.Ages in red indicate dating results considered less reliable in connection with low acceptance rates and substantial feldspar contamination (see text for details).contaminantby Alexanderson and Murray,29 is feldspar.Even a few single grains of feldspar within a large quartz aliquot may dominate the OSL response, as feldspar signals can be up to an order of magnitude greater than those of quartz.59,63As opposed to quartz, which is mainly stimulated with light in the visible wavelength spectrum, feldspar additionally responds to stimulation in the infrared.Thus, low OSL-IR depletion ratios calculated for V3-58 and V3-90 confirm feldspar as the likely contaminant of the samples (see Section 6.1.1).This in combination with extremely low post-IR OSL signals during double-SAR measurement indicates a large part of the OSL signal of these samples originates from IR sensitive traps dominating the quartz component.However, since preparation measures were taken to isolate the quartz fraction (density separation and HF etching), all samples

Table 4 and
Figure S2.While the charcoal samples were not species identified, the δ 13 C and an ocular inspection indicate, however, that they were of the species Pinus and therefore are almost certainly Pinus sylvestris.The acid precipitated SOL-fraction was almost not observable, indicating negligible influence of humic acids from the soil.
away, or even within the same dune field, unless a chronostratigraphic framework is developed placing events (for example reactivations or phases of stability) within the same timeframe.If the wider causes of dune formation, stabilisation and reactivation within a larger study area are to be inferred, numerical ages from multiple different dunes and in different stratigraphic levels are needed, but some degree of local signal will be inevitable in any single dune.In this context, the present study aims to expand the existing Fennoscandian dune chronostratigraphic framework to Arctic Sweden by developing a robust protocol for quartz OSL dating and providing reactivation ages for two separate dune areas.
which might support the possibility that these ages are However, the fire associated with this charcoal band may not have triggered the aeolian episode indicated by the uppermost massive sand at Jorggástat, as the OSL sample from this sand (J1-6) above the charcoal horizon is likely much younger than the charcoal itself, as indicated by the weak luminescence characteristics of the quartz as discussed above, and the comparatively weak soil development having taken place within the unit.
3nit at Vastakielinen (V3-15; 3.65 ± 0.18 ka), indicating the timing of initial dune reworking after the first stable phase, coincides well with a single age provided by Matthews and Seppälä3 2,3and from the sites here provide evidence for further-potentially localised-aeolian activity between these phases.That said, between 7 and 4 ka, only very little evidence for aeolian activity exists from dunes throughout Arctic Fennoscandia (only Kotilainen 2 and Matthews and Seppälä 3 ), including those here, suggesting minimal dune movement at this time.The age of the first homogenous sand