Testing the accuracy of quartz OSL dating using a known-age Eemian site on the river Sula, northern Russia

Page 1

ARTICLE IN PRESS

1871-1014/$ - se

doi:10.1016/j.qu

�CorrespondE-mail addr

Quaternary Geochronology 2 (2007) 102–109

www.elsevier.com/locate/quageo

Research paper

Testing the accuracy of quartz OSL dating using a known-age Eemiansite on the river Sula, northern Russia

A.S. Murraya,�, J.I. Svendsenb, J. Mangerudb, V.I. Astakhovc

aNordic Laboratory for Luminescence Dating, Department of Earth Science, University of Aarhus, Risø National Laboratory, DK-4000 Roskilde, DenmarkbDepartment of Geoscience and Bjerknes Centre for Climate Research, University of Bergen, Allegtaten 41, N-5007 Bergen, Norway

cGeological Faculty, St. Petersburg University, Universitetskaya 7/9, St. Petersburg 199034, Russian Federation

Received 9 April 2006; accepted 10 April 2006

Available online 14 June 2006

Abstract

Quartz optically stimulated luminescence (OSL) forms the basis for the chronology of Weichselian ice advances in Arctic Eurasia

developed over the last few years. There is almost no age control on this chronology before 40 ka, except for some marine sediments

correlated with marine isotope stage (MIS) 5e on the basis of their palaeofauna. Results from more southern latitudes have shown that

dose estimates based on quartz OSL and the single aliquot regenerative (SAR) dose procedure may underestimate the age of MIS 5e

deposits. Here we use the same method to date well-described marine sediments, thought to have been deposited during the very

beginning of the Eemian interglacial at �130 ka, and exposed in two sections on the river Sula in northern Russia. Various quality-

control checks are used to show that the OSL behaviour is satisfactory; the mean of 16 ages is 11272 ka (s ¼ 9 ka). This represents an

underestimate of �14% compared to the expected age, a discrepancy similar to that reported elsewhere. In contrast to SAR, the single

aliquot regeneration and added (SARA) dose procedure corrects for any change in sensitivity during the first OSL measurement. The

SARA results are shown to be �10% older than those from SAR, confirming the geological age estimate and suggesting that SAR ages

may underestimate older ages (larger doses), despite their good performance in the younger age range.

r 2006 Elsevier Ltd. All rights reserved.

Keywords: OSL dating; Eemian; MIS 5e; Quartz; Luminescence dating; Accuracy; SAR; SARA

1. Introduction

The last time the Barents–Kara Ice Sheet advanced ontomainland Russia was during the Middle Weichselian(Valdaian) beyond the limit of 14C dating (Svendsenet al., 2004). In the large European project ‘‘QuaternaryEnvironment of the Eurasian North (QUEEN)’’ thechronology of the glacial history, and indeed otherstratigraphic events, was almost entirely based on opticallystimulated luminescence (OSL) ages (QUEEN, 1999, 2001,2004). From this chronology, it was concluded that thenorthern rim of the Eurasian continent was inundated atleast twice during the Early/Middle Weichselian by theBarents–Kara Ice Sheet. The precision and accuracy ofthese OSL ages is thus of considerable importance. An age-

e front matter r 2006 Elsevier Ltd. All rights reserved.

ageo.2006.04.004

ing author. Tel.: +4546 77 46 77; fax: +4546 77 56 88.

ess: [email protected] (A.S. Murray).

independent test of internal consistency between sitesspread over hundreds of km and with widely differentpresent-day dose rates was obtained when samples from anumber of sections in Lake Komi beach sediments allyielded ages in the range 80–100 ka (Mangerud et al.,2004), but this did not test the accuracy of these ages. Someattempt has been made to test the method against marineisotope stage (MIS) 5e deposits, using individual samplesand sites during the QUEEN project (e.g., Mangerud et al.,1999), but these were neither intensive nor systematic.Murray and Olley (2002) have summarised a comparison

between quartz single aliquot regenerative (SAR) OSL agesand independent age control. From their data, it wasconcluded that there is no significant systematic differenceover the entire age range considered (up to about 600 ka).However it is also known that, at the individual grain level,one can observe sensitivity-corrected natural signals that lieabove the SAR growth curve (Feathers, 2003; Yoshida

Page 2

ARTICLE IN PRESS

Fig. 1. Location map and Sula 22 section, with sampling positions and

SAR OSL ages. The sediment log (composite) from Sula 22 is from

Ulvedal (2003). (Log) refers to column 1, Table 1.

A.S. Murray et al. / Quaternary Geochronology 2 (2007) 102–109 103

et al., 2000). For these grains, the SAR procedure clearlyoverestimates the equivalent dose (De), and it is notunreasonable to assume that there must be other grainsfor which the natural signals do intersect the growth curve,but at too high a dose. Bailey (2004) has suggested a modelto explain this, but no significant testing has yet beencarried out. On the other hand, close examination of thedata presented by Murray and Olley (2002) suggests thatfor samples correlated to MIS 5e, there is a tendency forthe OSL ages to lie below the expected age. Murray andFunder (2003) undertook a case study at such a site, whichthey argued must have been deposited between 128 and132 ka. The SAR OSL age was 11976 ka, which is justconsistent (within 2 standard errors) with the expected agerange, but only when all estimates of systematic uncer-tainty were included. Schokker et al. (2004) report a relatedinvestigated at a site on the lower Rhine. Stokes et al.(2003) also report a comparison of five OSL ages with anoxygen isotope model age curve, and other independentage control, and again there is a tendency for their OSLages to systematically underestimate the expected ages, butonly by �10%; this was not considered significant by thoseauthors. In our opinion, there seems little doubt that SARages perform well when compared with independent agecontrol over the 14C age range, but it is certainly possiblethat SAR underestimates ages by �10% around MIS 5e.This paper investigates this hypothesis by considering afurther case study from Arctic Russia.

2. Site description and sampling

On the river Sula (67100.00N, 50120.20E), a northerntributary to the Pechora River, we have dated two sections(21 and 22) with well-exposed, foreshore marine sediments,described in Mangerud et al. (1999). A crucial point is thatthese sections contain an in situ mollusc fauna indicatingcoastal water several degrees warmer than at present day;these must certainly indicate a warm interglacial. Theseand related sites have, for decades, been ascribed to the so-called Boreal transgression that has, in turn, been acceptedas correlated with the Eemian in the Russian literature(e.g., Devyatova, 1982; Lavrova, 1949; Lavrova andTroitsky, 1960). Based on a detailed correlation withwestern European pollen stratigraphy by Funder et al.(2002), we consider the sediments should have an age ofabout 130 ka (beginning of the interglacial), and followingMurray and Funder (2003) an uncertainty on this age of72 ka is adopted.

At the base of site 22 (Fig. 1), there is a dark silt and clay(poorly exposed during our visits) containing a rather coolmollusc fauna, which may possibly date from the LateSaalian deglaciation. The well-exposed sandy unit beginswith a thin gravel in tabular foresets containing pairedMytilus edulis. The gravel is interpreted as a foreshorefacies and is overlain by a generally upward fining sandwith a parallel increase in bioturbation. The fauna isdominated by large in situ individuals of Arctica islandica

and with a few Cerastoderma edule and Zirphaea crispata.The marine formation is cut by an erosional unconformitythat is, in turn, overlain by Weichselian age sand and ablack lacustrine clay (Fig. 1). The marine sand, from whichall samples reported here were collected, was deposited inshallow water and during a short period, almost certainlyless than 5000 years.The site Sula 21 lies 4 km upstream of Sula 22, but the

sediment sequence here is not as thick as at site 22. Thecorrelation with site 22 appears secure. Both sites indicatethat the corresponding Eemian shoreline was about 50mabove present sea level, which is explained by isostaticdepression produced by the large Saalian (MIS 6) ice sheetthat probably covered the entire region.

3. Measurement facilities and methods

All OSL measurements were made on a Risø TL/OSLreader model DA12 or 15 (Bøtter-Jensen et al., 2000)equipped with blue light stimulation (470730 nm;�50mWcm�2). Photon detection was through 7.5mm ofU340 glass filter. Samples obtained in 2002 were taken inopaque plastic tubes driven into a cleaned face in thesection. The older samples (log numbers 1–3 and 6 [Sula 22]and 4–5 [Sula 21] in Table 1) were collected by digging thesediment from a recently cleaned section into black plasticbags. This was certainly not as good method as latter, butcare was taken that the sand should not be exposed to light.

Page 3

ARTICLE IN PRESS

Table

1

Summary

ofdosimetry,

Demeasurements

andluminescence

ages

Log

no.

Sample

no.

Field

no.

Depth

(m)

238U

(Bqkg�1)

226Ra

(Bqkg�1)

232Th

(Bqkg�1)

40K

(Bqkg�1)

Sat.

(%)

Obs.

(%)

Dose

rate

(Gyka�1)

SAR

De(G

y)

nSAR

age

(ka)

SARA

De

(Gy)

nSARA

age

(ka)

1942513

93-25/2

8n.a.

n.a.

n.a.

n.a.

27*

51.5070.07

16074

19

10776

169712

48

109710

2952503

94-0146

7n.a.

n.a.

n.a.

n.a.

27*

41.3270.06

13977

20

10577

174718

48

129715

3952504

94-0147

5n.a.

n.a.

n.a.

n.a.

27*

81.3970.07

14876

18

10677

126710

48

9179

4952507

y94-0084

5n.a.

n.a.

n.a.

n.a.

27*

71.4570.07

18078

20

12478

15778

48

10979

5952509

y94-0097

6n.a.

n.a.

n.a.

n.a.

27*

31.4770.07

16675

18

11377

201724

48

144719

6002501

94-0149

5.5

774

5.570.3

5.970.3

36078

27*

31.1870.04

14076

21

11877

146710

48

123710

7H22545

02-4013

3.3

774

5.670.3

6.470.3

28277

24

15

1.0270.07

10374

19

10178

112710

48

111712

8H22546

02-4014

4.3

373

5.370.2

6.270.2

37077

25

51.1670.06

13074

20

11177

146710

48

128711

9H22547

02-4015

5.3

874

4.370.6

5.070.5

330720

28

81.0970.06

9976

30

9077

134711

48

127713

10

H22548

02-4016

6.2

974

8.870.3

9.370.4

30678

28

51.1470.05

14577

20

12879

130711

48

119712

11

H22549

02-4017

7.0

674

4.870.3

5.770.2

33778

28

61.1170.05

12078

23

10879

125715

44

118716

12

H22550

02-4018

7.8

272

4.370.4

4.570.3

315714

27

41.0470.05

10677

23

10378

116712

42

119714

13

H22551

02-4019

8.6

572

5.3570.16

5.0570.18

27875

24

21.0070.04

102711

23

102712

136712

46

146715

14

H22552

02-4020

9.4

772

5.9970.18

6.7170.17

32776

28

51.1170.05

12674

20

11376

146717

48

141718

15

H22553

02-4021

10.2

1274

5.570.6

5.970.5

210716

26

30.8670.05

8775

24

10178

12476

48

159713

16

H22554

02-4022

11.0

574

6.170.3

5.2870.03

22876

24

50.8970.04

8973

32

10076

94710

48

117714

Allsamplesfrom

Sula

22exceptthose

marked

y,whichare

from

Sula

21.Thegammaspectrometry

calibrationetc.isdescribed

inMurrayet

al.(1987).Conversionfactors

from

activityconcentrationsto

dose

rate

are

taken

from

Olley

etal.(1996).Dose

ratesin

thefirstfivesamplesare

basedonfieldgammaspectrometry

andlaboratory

betacounting;‘n.a.’—

notavailable.Saturatedwatercontent(those

marked

*are

estimated)corrections,calculatedcosm

icraycontributionsandaninternalquartzdose

rate

of0.06Gyka�1are

included

inthedose

rate

data.Theuncertainties

are

estimatedstandard

errors,and

nisthenumber

ofindividualaliquots

contributingto

De.

A.S. Murray et al. / Quaternary Geochronology 2 (2007) 102–109104

Page 4

ARTICLE IN PRESS

10 Stimulation time, s

10 20 30

2x104

3x104

(0.

16 s

)-1

A.S. Murray et al. / Quaternary Geochronology 2 (2007) 102–109 105

On return to the laboratory, the light-exposed material atthe end of each tube was retained for dose rate and watercontent analysis; the inner portion was wet-sieved torecover the 180–250 mm grain size. This was then cleaned,etched in the usual manner (HCl, H2O2 and HF) and testedfor feldspar contamination using infrared (IR) stimulation.No significant IR- (compared to blue) stimulated signalwas detected in any sample. De determination used a SARprotocol (Murray and Wintle, 2000, 2003) and the initial0.8 s integral of the OSL signal, less a backgroundestimated from the last 10 s of stimulation.

Dose rate calculations are based on high-resolutiongamma spectrometry (Murray et al., 1987). Calculated(Prescott and Hutton, 1994) cosmic ray dose rates rangefrom, at the top of the Sula 22 profile, 0.14Gy ka�1 to, atthe bottom, 0.07Gy ka�1. Natural and saturated watercontents were measured in the laboratory. For calculationof total dose rates, we have assumed that the sedimentswere saturated by water or ice throughout the burialperiod; this is because Pleistocene permafrost started todegrade in this area only by �8 ka 14C BP (Tveranger et al.,1995), and patches of thin permafrost still occur nearby.Nevertheless, we must recognise that the Sula 22 section islocated in the valley slope; during some of the time whenthe sediments were not frozen, it is possible that the upperpart of the sequence was located above the ground watertable. Five of the earlier samples taken in 1994 and 1995(log numbers 1–5 in Table 1) are exceptions to the above, inthat dose rates are based on field gamma counting andlaboratory beta counting, and saturated water contents(needed only for beta dose rate calculations) are assumed,based on the measurements on the remaining samples.

4. Radionuclide concentrations

All available radionuclide concentrations (log no. 6–16in Table 1) are summarised in Table 1 and Fig. 2. Althoughthere are peaks in the 226Ra and 232Th concentrations at�6m depth, these do not contribute significantly to thetotal dose rates, which are dominated by 40K at this site.

Bq.kg-1

0 2 4 6 8 2 4 6 8

Dep

th, m

4

6

8

10

Bq.kg-1

0

Bq.kg-1

0 150 300

Gy.ka-1

0.0 0.5 1.0

226Ra 232Th 40K Totaldoserate

Fig. 2. Radionuclide concentrations and total dose rates at Sula 22.

The average total dose rate is about 1Gy ka�1, with arange of only �20% about this value. There is no evidencefor gross radioactive disequilibrium between 238U and226Ra, although the large uncertainties on the 238Umeasurements limit the usefulness of this comment.

5. Luminescence characteristics

Fig. 3 (inset) presents an example (using sample H22546)of a regenerated decay curve for this material (naturalcurves were indistinguishable from regenerated). The decaycurves are typical for quartz, and appear to be dominatedby the fast component (Bailey et al., 1997; Jain et al., 2003).A representative growth curve is also shown in Fig. 3; thisis well represented (solid line) by the sum of two saturatingexponential curves (shown separately as dashed–dot–dotlines). The sensitivity-corrected natural signal intercepts thegrowth curve at about 55% of the saturation OSL value; itis interesting to note that, at the corresponding dose, thefirst exponential component of the growth curve is fullysaturated (D0 ¼ 44Gy), whereas the second component isonly at 30% of its saturated value (D0 ¼ 450Gy).The lack of dependence of De on preheat temperature is

shown in Fig. 4a,b, together with the correspondingrecycling ratio and recuperation values. All results areinsensitive to variations in preheat temperature; recupera-tion values may show a slight systematic trend, butnevertheless do not exceed 1.5% of the De. A preheattemperature of 260 1C was selected for further work. Asummary of all recycling ratios and recuperation valuesfrom all aliquots used for De estimation in this study isgiven in Fig. 5; the great majority (80%) of recuperationvalues lie below 2%, and the mean recycling ratio (inset) is1.01470.004 (n ¼ 263), with a relative standard deviation

Laboratory dose, Gy

0 200 400 600 800 1000

Cor

rect

ed O

SL

0

2

4

6

8

Natural

De=168 GyH22546

104

OS

L,

Fig. 3. SAR growth curve for one aliquot of sample H22546 (Table 1, log

8). Recycling and recuperation points are shown as unfilled symbols. The

best fit solid line is the sum of the two exponential growth curves shown as

dashed–dot–dot lines. A typical regenerated OSL decay curve is shown

inset.

Page 5

ARTICLE IN PRESS

Measured/given dose ratio

0.6 0.8 1.0 1.2 1.4

Fre

quen

cy

0

5

10

15

20

25

30mean = 1.017±0.012 (n=95)

Fig. 6. Summary dose recovery data for all aliquots (Table 1, log 6–16).

Vertical line is drawn at mean value.

De,

Gy

0

100

150

H22552; Cut heat 160°C

Preheat temperature, °C150 200 250 300

Rec

yclin

g R

atio

0.0

0.5

1.0

Rec

uper

atio

n, %

N

0

1

2mean 1.021±0.005

(a)

(b)

Fig. 4. (a) Preheat plateau for sample H22552 (Table 1, log 14; 3 aliquots

at each temperature); (b) recycling and recuperation for the same aliquots.

Recuperation, %N0 2 4 6 8 10

Fre

quen

cy

0

10

20

30

40

50

(n=272)

Recycling ratio0.8 1.0 1.2

Fre

quen

cy

0

25

50

75

Mean = 1.014

±0.004

(n=263)

Fig. 5. Summary of all available recuperation and (inset) recycling data

from this study (vertical line is drawn at mean value).

A.S. Murray et al. / Quaternary Geochronology 2 (2007) 102–109106

(RSD) of �6%. Finally, Fig. 6 presents a summary of 95dose recovery measurements for the last 11 samples inTable 1, with a mean of 1.01770.012 and a RSD of 12%.It is clear from all these summary statistics that our SAR

protocol is able to measure a laboratory dose bothaccurately and precisely. It now remains to be testedwhether it can do the same for a dose absorbed in nature.

6. Luminescence ages

All the mean estimates of De from 16 samples aresummarised in Table 1. This data set includes the fivesamples for which dosimetry is based on field gamma andlaboratory beta counting; these have been presentedpreviously by Mangerud et al. (1999). Some values of De

(and thus ages) in Table 1 are significantly different fromthe earlier published values; the previous De values werebased on very few aliquots (between 3 and 7) and all havebeen completely remeasured to make them consistent withcurrent practice.The dependence of De on dose rate is shown graphically

in Fig. 7. The two variables are highly correlated(R2 ¼ 0:89) and the slope of the line is 113 ka, or 112 kaif the line is forced through the origin. For comparison, thelines representing the expected age range are showndashed; only two or three of the data points are consistentwith the expected slopes. The age data are presented as ahistogram (inset to Fig. 7); they are normally distributed,and the mean age (weighted or unweighted) is 11272 ka, or77 ka if a further 5% systematic uncertainty is included(Murray and Funder, 2003). We conclude that the mean

Dose rate, Gy.ka-1

0.0 0.5 1.0 1.5

Equ

ival

ent d

ose,

Gy

0

50

100

150

200

slope=113 ka

slope=128-132 ka

Age, ka100 120 140

Fre

quen

cy

2

4

6

8

10expected

age range

Fig. 7. Relationship between equivalent dose and dose rate for all

samples. The solid regression line is not forced through the origin. The

expected age range is shown as two dashed lines. A histogram of all ages is

shown inset.

Page 6

ARTICLE IN PRESS

Laboratory dose, Gy

-100 0 100 200 300

Mea

sure

d do

se, G

y

0

100

200

300

400

De = 146±10 Gy

Slope = 0.88±0.04

0 2 4 6 8 10

Nor

m. D

e

0.8

0.9

1.0Integration time, s

Fig. 8. Typical SARA data set (sample log 8), showing the doses

measured using SAR plotted against the doses added in the laboratory.

Inset: effect of changing integration interval on equivalent dose. Dose

estimates for all aliquots for all samples were normalised to the value

obtained by summing channels 1 and 2 (point at 0.32 s).

A.S. Murray et al. / Quaternary Geochronology 2 (2007) 102–109 107

SAR luminescence age underestimates the expected age of130 ka by up to �14% (depending on the duration of thetransgression).

There are various possible explanations for such anunderestimate: (i) Dose rate effects: The laboratory doserate is 4109 times that in nature, and a combination ofcharge competition and trap stability effects could meanthat the laboratory-generated growth curve does not applyto the natural signal. However Bailey (2004) predicts thatthis effect should lead to an age overestimation, rather thanthe underestimation we find here. (ii) Variations in dose

rate: It is assumed that the dose rate has been constantthrough time. If the sediment was deposited out of secularequilibrium, or if, after deposition, it behaved as an opensystem with respect to radioactive nuclides, then thisassumption may be incorrect. Murray and Funder (2003)argued that this was the explanation for two of their 22ages appearing as outliers in their coastal marine Eemiansite on the south-east coast of Denmark. Even with thosetwo excluded, their mean OSL age underestimated by�9%. We are less concerned by this possibility here;despite a factor of two variations in measured dose rate,the relative standard deviation in the ages is only �8%,and there are no outliers in our age distribution. (iii)Inaccurate estimation of the intensity of the fast component:The initial OSL signal is often contaminated by a thermallyless stable slow (S3) component (Singarayer and Bailey,2003; Jain et al., 2003). If this slow component contributessignificantly to the initial intensity of the OSL signal, thendose underestimation may result when comparing thenatural signal (in which this component has thermallyfaded) with a regenerated signal (in which it does not havetime to fade). This would result in a decrease in De withstimulation time. However, significant contaminationwould be surprising. The lifetime of the slow (S3)component is about 1000 times that of the fast component(Jain et al., 2003; Singarayer and Bailey, 2003) and it isclear from the decay curves (e.g. inset to Fig. 3) that theinitial intensity of the slow component can only be a smallfraction of that of the fast. This, coupled with backgroundsubtraction using the last 10 s of a total of 40 s stimulation(5S3 lifetime), makes it unlikely that significant contam-ination has occurred in these cases. The preheat plateautest (Fig. 4) is specifically intended to isolate a thermallystable signal; nevertheless, it must be recognised that ourdata are not sufficiently sensitive to exclude a 14%difference between the De at the selected preheat of260 1C and at the maximum of 300 1C (the observeddifference in De at these two temperatures is 5714%). Thisissue is considered further in Section 7. (iv) Changes in

sensitivity during first stimulation: Murray and Wintle(2000) pointed out that SAR protocols such as ours cannotdetect changes in sensitivity (from whatever cause)occurring during the preheating or stimulation of thenatural OSL signal; nevertheless, they were able to showthat such changes were not significant for two samples withmuch smaller De than those considered here. They

suggested that for single aliquot work, the only reliableargument for the absence of this effect was agreement withknown age. However there is a procedure for detectingsuch changes if multiple aliquots are used—the singlealiquot regeneration and added (SARA) dose procedure(SARA; Mejdahl and Bøtter-Jensen, 1994). This approachis employed in Section 8.

7. Testing for dependence of De on integration interval

The only reliable way to test for slow-componentcontamination of the fast component is (i) by fittinglinearly modulated OSL data to separate the variouscomponents, and then determining the De value using onlythe fast-component intensity or (ii) by using an instru-mental approach to separate the fast component usingpreferential IR bleaching of the fast component at elevatedtemperature (Jain et al., 2004). However, there is noestablished routine approach suitable for application to the42000 OSL signals used here. A less conclusive but morepractical approach is to examine the variation of De withintegration time—as the integration time is shortened, thesignal is dominated more and more by the fast component.To investigate this dependence, all OSL decay curves fromthe later analyses (log numbers 6–16 in Table 1) werereanalysed using varying initial integration intervals, 1–2channels, 1–5, 1–10, etc. (each channel corresponds to0.16 s). Then all the De values from all aliquots, derivedusing the various integration times, were normalised toeach other by the results obtained using the shortestintegration interval, and these normalised results averagedover all samples. The results (inset to Fig. 8) were fittedwith a single exponential decay. From these data, the

Page 7

ARTICLE IN PRESSA.S. Murray et al. / Quaternary Geochronology 2 (2007) 102–109108

5-channel (0.8 s) integration used in this work under-estimates by o3% the value of De that is obtained whenthe integration interval is extrapolated to zero. While notconclusively proving that the slow (S3) component doesnot affect our results significantly, the consistency of this3% with the variation observed from the preheat plateau(Fig. 4) gives us confidence that any such effect is small.

8. Allowing for initial sensitivity change—comparing SAR

and SARA De estimates

In the SARA procedure (Mejdahl and Bøtter-Jensen,1994), groups of aliquots from a single sample are eachgiven a different added dose before any other treatment.The apparent dose in each aliquot is then measured, andthe measured dose plotted against the laboratory dose(Fig. 8). In our implementation of this procedure we haveused SAR to measure the individual aliquots. Murray(1996) modelled this procedure, and discussed the assump-tions required for these data to lie on a straight line, which,when extrapolated back to the laboratory dose axis, wouldintercept at De. Any sensitivity change during the firstpreheat and OSL measurement is reflected in the slope ofthe regression; in the absence of sensitivity change, this is ofunit slope. Fig. 8 illustrates this process with a data set of48 aliquots from log number 8 in Table 1. The data areclearly consistent with the fitted straight line (of slope0.8870.04), and the intercept on the horizontal axis is146710Gy (c.f. SAR 13074Gy). This process has beenrepeated for all 16 samples, and the SARA De and ages arelisted in Table 1. The mean slope is 0.9570.02 (n ¼ 16) andthe mean ratio of SARA to SAR De is 1.0970.04 (n ¼ 16).

Including a systematic uncertainty of �5% gives anaverage SARA age of 12478 ka (n ¼ 16), completelyconsistent with the expected age. The typical uncertaintyon individual ages is �13 ka, to be compared with theobserved standard deviation of the 13 ages of 18 ka; it canbe deduced that there may be an unexplained residualdispersion of about 10% in the ages. Nevertheless,sensitivity change in the first measurement of these samplesappears to contribute significantly to the 14% discrepancybetween SAR ages and the expected age.

9. Discussion and conclusions

Most (but not all) of the testing of the reliability ofquartz single aliquot regenerative (SAR) optically stimu-lated luminescence (OSL) dating has been undertaken bycomparison with 14C or with other independent datingmethods with a younger range of applicability (e.g.,Murray and Olley, 2002). The application reported in thispaper, of SAR dating to an northern Eurasian Eemiandeposit, appears to support the suggestion that SAR maybegin to underestimate the true age to some degree forsamples older than about 40 ka, with a discrepancy ofabout 10% at �130 ka. From the work reported here, someof this underestimate probably arises from changes in

sensitivity during the measurement of the natural signal,and some may arise from contamination of the quartz fast-component OSL signal by a less-stable slow component(although this is less certain). However, it would be unwiseto accept these two mechanisms as the only sources ofsignificant systematic uncertainty in our OSL ages; opensystem radionuclide behaviour and inaccurate estimates oflong-term water contents are two obvious potential sourcesof significant systematic error that cannot readily even beidentified, far less quantified. More case studies arerequired before we can state confidently that (i) SARunderestimates are to be expected in this age range, and (ii)the single aliquot regeneration and added does (SARA)approach is more accurate. Even if this proves to be true, itis not clear that SARA should be used instead of SAR insuch older material; because of the need to add laboratorydoses on top of the natural dose, SARA can only beapplied to samples where the natural dose lies at a smallfraction of the saturation limit. Moreover, SARA is a time-consuming multiple aliquot procedure. It provides a light-intensity weighted-average value of equivalent dose (De),and precludes any study of the distribution of dose within asample; it cannot be applied to single grains. Nevertheless,SARA remains the only established measurement proce-dure that automatically corrects for sensitivity change inthe first measurement.

Acknowledgements

This work is a contribution to the research projectsPECHORA II (Paleo Environment of the Russian Arctic)and ICEHUS (The Ice Age development and humansettlement in northern Eurasia) both financially supportedby the Research Council of Norway. Support from theNordic Centre of Excellence programme of the JointCommittee of the Nordic Natural Science ResearchCouncils is also gratefully acknowledged.

Editorial handling by: R. Grun

References

Bailey, R.M., 2004. Paper I—simulation of dose absorption in quartz over

geological timescales and its implications for the precision and

accuracy of optical dating. Radiation Measurements 38, 299–310.

Bailey, R.M., Smith, B.W., Rhodes, E.J., 1997. Partial bleaching and the

decay form characteristics of quartz OSL. Radiation Measurements

27, 123–136.

Bøtter-Jensen, L., Bulur, E., Duller, G.A.T., Murray, A.S., 2000.

Advances in luminescence instrument systems. Radiation Measure-

ments 32, 523–528.

Devyatova, E.I., 1982. Late Pleistocene Natural Environment and Its

Impact on Human Population in the North-Dvina Basin and in

Karelia. Institute of Geology, Petrozavodsk (in Russian).

Feathers, J.K., 2003. Single-grain OSL dating of sediments from the

Southern High Plains, USA. Quaternary Science Reviews 22,

1035–1042.

Funder, S., Demidov, I., Yelovicheva, Y., 2002. Hydrography and mollusc

faunas of the Baltic and the White Sea-North Sea seaway in the

Page 8

ARTICLE IN PRESSA.S. Murray et al. / Quaternary Geochronology 2 (2007) 102–109 109

Eemian. Palaeogeography, Palaeoclimatology, Palaeoecology 184,

275–304.

Jain, M., Murray, A.S., Bøtter-Jensen, L., 2003. Characterisation of blue-light

stimulated luminescence components in different quartz samples: implica-

tions for dose measurement. Radiation Measurements 37, 441–449.

Jain, M., Murray, A.S., Bøtter-Jensen, L., Wintle, A.G., 2004. A single-

aliquot regenerative-dose method based on IR (1.49 eV) bleaching of the

fast OSL component in quartz. Radiation Measurements 39, 309–318.

Lavrova, M.A., 1949. On the interglacial marine transgressions of the

Pechora region. Uchonye zapiski, Leningrad University, Series

Geography 124, 14–51 (in Russian).

Lavrova, M.A., Troitsky, S.L., 1960. Interglacial transgressions in

northern Europe and Siberia. In: Reports of Soviet Geologists to the

XXI session of the International Geological Congress, vol. 4,

pp. 124–136 (in Russian).

Mangerud, J., Svendsen, J.I., Astakhov, V.I., 1999. Age and extent of the

Barents and Kara ice sheets in Northern Russia. Boreas 28, 46–80.

Mangerud, J., Jakobsson, M., Alexanderson, H., Astakhov, V., Clarke,

G., Henriksen, M., Hjort, C., Krinner, G., Lunkka, J.-P., Moller, P.,

Murray, A., Nikolskaya, O., Saarnisto, M., Svendsen, J.I., 2004. Ice-

dammed lakes and rerouting of the drainage of Northern Eurasia

during the last glaciation. Quaternary Science Reviews 23, 1313–1332.

Mejdahl, V., Bøtter-Jensen, L., 1994. Luminescence dating of archae-

ological materials using a new technique based on single aliquot

measurements. Quaternary Science Reviews (Quaternary Geochronol-

ogy) 13, 551–554.

Murray, A.S., 1996. Developments in optically transferred luminescence

and photo-transferred thermoluminescence dating: application to a

2000-year sequence of flood deposits. Geochimica et Cosmochimica

Acta 60, 565–576.

Murray, A.S., Funder, S., 2003. OSL dating of a Danish coastal marine

deposit: a test of accuracy. Quaternary Science Reviews 22, 1177–1183.

Murray, A.S., Olley, J.M., 2002. Precision and accuracy in the optically

stimulated luminescence dating of sedimentary quartz. Geochronome-

tria 21, 1–16.

Murray, A.S., Wintle, A.G., 2000. Luminescence dating of quartz using an

improved single-aliquot regenerative-dose protocol. Radiation Mea-

surements 32, 57–73.

Murray, A.S., Wintle, A.G., 2003. The single aliquot regenerative dose

protocol: potential for improvements in reliability. Radiation Mea-

surements 37, 377–381.

Murray, A.S., Marten, R., Johnston, A., Martin, P., 1987. Analysis for

naturally occurring radionuclides at environmental concentrations by

gamma spectrometry. Journal of Radioanalytical and Nuclear

Chemistry 115, 263–288.

Olley, J.M., Murray, A.S., Roberts, R.G., 1996. The effects of

disequilibria in the uranium and thorium decay chains on burial dose

rates in fluvial sediments. Quaternary Geochronology 15, 751–760.

Prescott, J.R., Hutton, J.T., 1994. Cosmic ray contributions to dose rates

for luminescence and ESR dating large depths and long-term time

variations. Radiation Measurements 23, 497–500.

QUEEN Special Volume, 1999. Boreas 23 (4), 281–535.

QUEEN Special Volume, 2001. Global and Planetary Change 31 (1–4),

1–474.

QUEEN Special Volume, 2004. Quaternary Science Review 23 (11–13),

1225–1511.

Singarayer, J.S., Bailey, R.M., 2003. Further investigations of quartz

optically stimulated luminescence components using linear modula-

tion. Radiation Measurements 37, 451–458.

Schokker, J., Cleveringa, P., Murray, A.S., 2004. Palaeoenvironmental

reconstruction and OSL dating of terrestrial Eemian deposits in the

southeastern Netherlands. Journal of Quaternary Science 19,

193–202.

Stokes, S., Ingram, S., Aitken, M.J., Sirocko, F., Anderson, R.,

Leuschner, D., 2003. Alternative chronologies for Late Quaternary

(Last Interglacial–Holocene) deep sea sediments via optical dating of

silt-sized quartz. Quaternary Science Reviews 22, 925–941.

Svendsen, J., Alexanderson, H., Astakhov, V., Demidov, I., Dowdeswell,

J., Funder, S., Gataullin, V., Henriksen, M., Hjort, C., Houmark-

Nielsen, M., Hubberten, H., Ingolfson, O., Jakobsson, M., Kjær, K.,

Larsen, E., Lokrantz, H., Lunkka, J., Lysa, A., Mangerud, J.,

Matiouchkov, A., Murray, A., Moller, P., Niessen, F., Nikolskaya,

O., Polyak, P., Saarnisto, M., Siegert, C., Siegert, M., Spielhagen, R.,

Stein, R., 2004. Late Quaternary ice sheet history of Northern Eurasia.

Quaternary Science Reviews 23, 1229–1271.

Tveranger, J., Astakhov, V., Mangerud, J., 1995. The margin of the last

Barents–Kara ice sheet at Markhida, northern Russia. Quaternary

Research 44, 328–340.

Ulvedal, P., 2003. Stratigrafi og sedimentologi fra siste interglasiale-

glasiale syklus langs Sula-elven i Nord-Russland (Stratigraphy and

sedimentology from the last interglacial-glacial cycle along the river

Sula in North-Russia). Cand. Scient. Thesis in Geology (in Norwe-

gian), Dept. of Geology, University of Bergen, 100pp.

Yoshida, H., Roberts, R.G., Olley, J.M., Laslett, G.M., Galbraith, R.F.,

2000. Extending the age range of optical dating using single ‘super-

grains’ of quartz. Radiation Measurements 32, 439–446.