Dependence of ice-core relative trace-element …eosterberg/images/Koffman et al., 2014...Dependence...

Dependence of ice-core relative trace-element concentration onacidification

Bess G KOFFMAN12 Michael J HANDLEY1 Erich C OSTERBERG3 Mark L WELLS4

Karl J KREUTZ12

1School of Earth and Climate Sciences University of Maine Orono ME USAE-mail besskoffmanmaineedu

2Climate Change Institute University of Maine Orono ME USA3Department of Earth Sciences Dartmouth College Hanover NH USA

4School of Marine Sciences University of Maine Orono ME USA

ABSTRACT To assess the role of methodological differences on measured trace-element concentrations

in ice cores we developed an experiment to test the effects of acidification strength and time on dust

dissolution using snow samples collected in West Antarctica and Alaska We leached Antarctic samples

for 3 months at room temperature using nitric acid at concentrations of 01 10 and 100 (vv) At

selected intervals (20 min 24 hours 5 days 14 days 28 days 56 days 91 days) we analyzed 23 trace

elements using inductively coupled plasma mass spectrometry Concentrations of lithogenic elements

scaled with acid strength and increased by 100ndash1380 in 3 months Incongruent elemental dissolution

caused significant variability in calculated crustal enrichment factors through time (factor of 13 (Pb) to

80 (Cs)) Using snow samples collected in Alaska and acidified at 1 (vv) for 383 days we found that

the increase in lithogenic element concentration with time depends strongly on initial concentration

and varies by element (eg Fe linear regression slope =166 r=098) Our results demonstrate that

relative trace-element concentrations measured in ice cores depend on the acidification method used

KEYWORDS ice and climate snow and ice chemistry

INTRODUCTION

Ice-core records provide evidence of past changes inatmospheric chemistry and transport linked both to global-and regional-scale climate variability (eg Mayewski andothers 1994 Kreutz and Sholkovitz 2000 Fischer andothers 2007) and to anthropogenic pollution and land-usechanges (McConnell and others 2002 2007 Vallelonga andothers 2002 Osterberg and others 2008) on a range oftimescales Impurities in glacial ice come from the de-position of particulate and soluble aerosols on the snowsurface which in turn are sourced from mineral dustvolcanic eruptions sea spray biological emissions andanthropogenic pollution The physical and chemical proper-ties of these impurities allow detection using a range ofanalytical tools The analysis of trace elements usinginductively coupled plasma mass spectrometry (ICP-MS)has become widely used for reconstructing past changes inthe atmospheric burden of pollutants such as Pb (egMcConnell and others 2002 Vallelonga and others 2002Hong and others 2004 Osterberg and others 2008 Grossand others 2012) and the deposition of important bioactivemetals such as Fe (eg Edwards and others 2006 Gaspariand others 2006) In addition changes in the composition ofdust signaling changes in its source have been inferred usingICP-MS measurements (Kreutz and Sholkovitz 2000 Gab-rielli and others 2010)

Ice-core samples are prepared for ICP-MS analysis bymelting and acidification and are then either injecteddirectly into the mass spectrometer (McConnell and others

2002) or analyzed in individual vials in both cases usuallywithout filtration (Knusel and others 2003 Osterberg andothers 2006) Acidification of ice-core samples is necessaryin order to maintain sample stability during analysisMethods for acidifying samples vary in terms of the mineralacid used acid strength and acid contact time Samples arecommonly acidified to 1 (vv) nitric acid whether asdiscrete samples (eg Osterberg and others 2006) or as aliquid stream for continuous injection (eg McConnell andothers 2002) though a weaker 05 nitric acid addition hasalso been used (eg Krachler and others 2005 2008 Grottiand others 2011) Hydrochloric acid has been used toacidify samples but is less common (Edwards and others2006) Perhaps the greatest source of methodologicaldifference is the acid contact time Continuous injectionICP-MS allows for an acid contact time of seconds tominutes while discrete samples may be acidified for anylength of time Although few ice-core studies have statedexplicitly how long samples were acidified prior to ICP-MSanalysis (table 1 of Rhodes and others 2011) it is apparentthat 24 hours is a common length of acidification (Krachlerand others 2005 Osterberg and others 2006)

A study using synthetic dust samples has suggested thatacidification time has a significant impact on measuredtrace-element concentrations (Rhodes and others 2011)but no systematic test has been conducted on actual snowor ice-core samples Considering the variety of techniquescommonly used in ice-core trace-element analysis weconducted a 3month experiment to evaluate trace-elementdissolution from atmospheric dust in West Antarctic surfacesnow samples at nitric acid concentrations of 01 10 and100 (vv) The goal was to evaluate the effects of acidstrength and time on dust dissolution and the impact that

Journal of Glaciology Vol 60 No 219 2014 doi 1031892014JoG13J137

Present address LamontndashDoherty Earth Observatory Columbia University

Palisades NY USA

103

differential dissolution rates could have on calculatedcrustal enrichment factors In addition we used a timeseries of snow-pit samples collected from Denali NationalPark Alaska USA and measured twice (383 days apart at1 (vv) nitric acid) to investigate the concentrationdependence of measured increases in trace-element con-centrations through time Our study expands the work ofRhodes and others (2011) by testing the effects of both acidstrength and time on real environmental samples fromAlaska and Antarctica two important ice-core drillinglocations Dust particles deposited in the snowpack at bothsites are likely to have experienced significant weatheringrelative to their source rock particularly during atmospherictransport (Cwiertny and others 2008) Our results aredirectly relevant to ice-core studies of trace-element de-position In addition we evaluate the concentrationdependence of relative changes in four trace-elementconcentrations (Al Ca Fe and S) measured over 1 year ofacidification Our results imply that ice-core time-seriesrecords developed after relatively short acidification periodslikely underestimate both the magnitude and the variabilityof trace-element concentrations potentially leading to biasin paleoclimate calibration work which in turn couldimpact interpretations of past climate variability based ontrace-element proxies

METHODS

Sample collection WAIS Divide Antarctica

Snow samples were collected from a 26m snow pit dug in2008 at the West Antarctic ice sheet (WAIS) Divide ice-coresite (794688 S 1120868W 1766masl) within the clean-air sector of camp Once the snow pit was excavated onewall was cut back an additional 20 cm using a plasticshovel cleaned with Citranox detergent and Millipore Milli-Q deionized water The wall was subsequently scrapedback an additional 4 cm using plastic scrapers that hadbeen decontaminated using nitric acid and Milli-Q waterSampling personnel wore Tyvek clean suits and doublepairs of gloves gloves were replaced whenever theybecame torn Snow samples were collected using aspecially built pure Ti sampling tool which had beenwashed with Citranox detergent and Milli-Q water Eachacid-washed sample bottle was double-bagged for storageand shipment using plastic Ziploc bags to reduce contactwith the environment Samples were shipped frozen to theUniversity of Maine and stored in a freezer until meltingand analysis

Experimental set-up

We combined Antarctic snow from several sample bottlesand melted it in an acid-cleaned capped Teflon1

perfluoroalkoxy (PFA)-coated bottle using a warm waterbath Once the snow was fully melted we homogenized itby shaking and inverting the bottle then poured the waterinto three additional PFA bottles (110mL each) We thenadded Optima grade nitric acid (Fisher Scientific) to eachbottle resulting in volumetric concentrations of 01 10and 10 acid respectively We evaluated possible sorptionof metals to the walls of the bottle used for melting thesnow by swishing the bottle with a 10 nitric acid solutionMetal sorption was negligible (05 of initial measuredconcentrations)

We analyzed samples after 20 min 24 hours 5 days14 days 28 days 56 days and 91 days At each analysis timewe poured five replicates from each bottle (to avoid potentialcontamination from a pipette) into acid-cleaned 4mLpolypropylene (PP) vials for analysis Samples were eitheranalyzed immediately or frozen until later analysis Weanalyzed samples using a Thermo Finnigan Element IIinductively coupled plasma sector-field mass spectrometer(ICP-SFMS) with an ESI Apex sample introduction system anda PFA-ST nebulizer Instrumental settings for the ICP-SFMSare given in Table 1 Procedural blanks (Milli-Q water) weretreated identically and analyzed in parallel (eg 01 10 and100 (vv) Optima grade nitric acid concentrations) Allbottles were shaken periodically (about once a week) andprior to subsampling and were kept in the dark at roomtemperature All sample handling was conducted in a class1000 clean room under a laminar flow High EfficiencyParticle Air (HEPA) bench Unfortunately because we pouredoff replicates (rather than using a pipette for volumetricaccuracy) we ran out of sample earlier than planned andwere not able to quantify recovery rates using hydrofluoricacid (HF) digestion

All five replicates were included in the calculations ofmean concentration and standard deviation the error barsshown are one standard deviation (1SD) of the meanconcentration for each analysis time Because melted snowand ice samples include both dissolved and suspendedimpurities they are inherently heterogeneous therefore thestandard deviation of each set of replicates likely representsreal sample variability We subtracted the average of fiveblank replicate measurements (for each acid treatment ateach analysis time) from each mean sample concentrationto correct for potential analytical inconsistencies Blankconcentrations remained well below the sample concen-trations throughout the experiment (over an order ofmagnitude for most elements Table 2) Data from this studywill be made publicly available through the US NationalSnow and Ice Data Center (NSIDC)

Sample collection and treatment Denali NationalPark Alaska

Snow samples were collected from a 4m pit dug at KahiltnaPass Denali National Park in 2008 (630768 N1511748W 2957masl Campbell and others 2012)Samples were collected at 5 cm resolution using trace metalclean techniques (as described above) and were stored inacid-washed bottles Snow samples were shipped frozenthen melted and acidified to 1 (vv) with Optima gradenitric acid under a HEPA bench Samples were analyzedusing ICP-SFMS at two time intervals spaced 383 days apartand were stored at room temperature in the interveningtime All sample handling was conducted in a class 1000clean laboratory

Table 1 ICP-SFMS instrument conditions and measurement par-ameters

Forward power 1280WCool gas 16 Lminndash1

Auxiliary gas 095 Lminndash1

Sample gas 08 Lminndash1

Additional gas 02 Lminndash1

Sample uptake 100mLminndash1

Number of scans 15

Koffman and others Trace-element dependence on acidification104

RESULTS AND DISCUSSION

Effects of acid strength and acidification time onmeasured concentrations

We tested the effects of two variables acid strength andacidification time on the measured concentration of traceelements in West Antarctic snow samples While bothvariables had a positive effect on the measured concen-tration we observed that the magnitude of concentrationchange depended on the element analyzed (Fig 1) Ingeneral increases with time were greater for the 01 acidtreatment because the initial concentration was lowcompared with the 10 and 100 acid treatments Elementsthat increased minimally (ie lt100) relative to the initialconcentration for all acid treatments includeMg S K Ca andCs Elements that increased substantially with time (gt100after 1 month) for all three acid treatments include Al MnFe Cu Cd La Ce Pb and U We also observed largedifferences among the acid treatments particularly early inthe experiment For example from 20 min to 24 hours Alincreased by 7 10 and 20 respectively for the 01 10 and100 acid treatments From 20 min to 5 days Al increasedby 32 62 and 82 respectively and after 3 months Alconcentrations increased by 345 374 and 397 for the threetreatments Although the relative differences among acidtreatments decreased with time Al concentrations at3 months were still significantly different among the three

treatments (21 018 25 011 and 30011mg Lndash1respectively errors given are 1SD) Elements generallyassociated with soil dust (eg Al Ti V Fe Mn La Ce Pr)

were consistently leached more effectively by stronger acidat the end of the 3month experiment concentrations of theseelements were gt1SD different from each other (eg Fe andMn dissolution curves in Fig 2) In contrast more solubleelements such as K and Ca reached equivalent concen-trations among the three acid treatments after 2ndash4 weeks (egCa dissolution curve in Fig 2)

Our second observation concerns the shape of theconcentration profile Figure 2 shows the four dominantpatterns we observed The Ca curve reveals that acid strengthis the dominant factor affecting its measured concentrationduring the first 2 weeks after which point the concentrationsbecome statistically indistinguishable K Cs and As show asimilar pattern (not shown) In addition we observe that forthe 1 and 10 acid treatments the Ca concentration

increases by 1mg Lndash1 from 20min to 24 hours then

decreases by 15mg Lndash1 at 5 days after which it increases

by 05mg Lndash1 to its final value We infer that some of the Caprecipitated and redissolved during this interval The Cd Feand Mn curves show a common feature of the concentrationprofiles of most elements a change in slope that occurs afterabout 2 weeks to 1 month While the Cd and Mn concen-trations appear to be reaching a steady state the slope of the10 acid Fe concentration curve suggests that at 3 monthsFe is still actively leaching into solution Al has a similarpattern (not shown) The concentrations of all other dust-borne elements appear to have a decreasing slope from 1 to3 months suggesting that their concentrations are approach-ing a maximum acid-leachable fraction This difference inslope between Fe and Al and other lithogenic elements may

Table 2 Instrumental detection limit (IDL) and mean of 35 blank measurements with 1SD error Blanks acidified to 01 10 and 100 (vv)nitric acid respectively all values reported in ng Lndash1 The number of reported decimal places reflects the analytical precision associated withmeasurement of each element

Analyte Instrumental detectionlimit

01 acid blank 10 acid blank 100 acid blank Factor of difference

27Al(MR) 678 669 394 550328 731465 2575As(MR) 009 016 009 017012 025023 15138Ba(LR) 0161 0666 0321 08470505 09320618 2744Ca(MR) 1012 4251 2076 37891819 51332987 16111Cd(LR) 0032 0125 0104 00730052 00930048 18140Ce(LR) 0007 0204 0543 00980069 00760034 1959Co(MR) 0074 0096 0058 00940046 00920049 1352Cr(MR) 0122 0414 0221 04650238 05160223 103133Cs(MR) 0006 0043 0029 00470030 00530033 863Cu(MR) 2391 2926 2662 32842721 31872478 3556Fe(MR) 95 195 131 241228 214102 6439K(HR) 1409 ltIDL ltIDL ltIDL gt8139La(LR) 0024 0049 0028 00630033 00550028 1624Mg(MR) 196 863 826 732519 10241643 3555Mn(MR) 0453 0463 0274 05340255 06610285 48208Pb(LR) 0065 0436 0364 04590343 04990324 29141Pr(LR) 0005 0017 0016 00180015 0014010 1832S(MR) 1264 2801 2667 24822094 26582342 2688Sr(LR) 022 050 028 042028 062040 4847Ti(MR) 0496 4588 3394 49802577 48713119 31238U(LR) 0006 0009 0003 00080004 00120004 1251V(MR) 0093 0300 0144 04470334 03820265 766Zn(MR) 6142 2894147436 2790631427 19855 15776 11

LR indicates low-resolution mode (mm=300) MR indicates medium-resolution mode (m=m =4000) and HR indicates high-resolution mode

(m=m =10 000) for the ICP-SFMSInstrumental detection limit calculated as 3SD of ten water blanksFactor of difference is average sample concentration divided by average blank concentration for each analysis time averaged for the seven analysis times The

reciprocal of this number indicates the fraction of the sample concentration represented by the blank concentration

Koffman and others Trace-element dependence on acidification 105

be explained by results from an East Antarctic trace-element

study Grotti and others (2011) used 045 mm filtration and

two acid treatments (05 (vv) nitric acid and an HF

digestion) to evaluate particulate vs dissolved and acid-

leachable vs total fractions of trace elements in surface snow

They found that 80ndash100 of Fe and Al was contained within

the particulate fraction while the particulate fraction of the

lithogenic elements Co Cr Mn Pb and V ranged from

Fig 1 Bar plot showing percent increase in elemental concentration relative to the 20min analysis for selected acidification times for eachacid treatment Snow-pit samples collected in Antarctica

Fig 2 Concentrations of Ca Cd Fe and Mn with acidification time for each acid treatment measured in snow-pit samples from Antarctica


20 to 70 (Grotti and others 2011) Thus the prolongedincreases of Fe and Al concentrations in our samples likelyreflect continuing dissolution of relatively refractory alumi-nosilicate and oxide minerals in dust

Our results are consistent with those from other studiesAlthough they used a different type of acid (hydrochloricadded to reach pH 19) Edwards and others (2006) foundthat Fe concentrations leaching out of dust in snow samplesfrom Law Dome Antarctica increased by a factor of threewithin the first month of acidification and continued toincrease for up to 3 months Rhodes and others (2011) foundthat elemental dissolution from four different crushed rockstandards (JG-2 granite Nod-P-1 ferromanganese noduleBHVO-2 basalt and W-2 dolerite) acidified at 1 (vv)nitric acid was both time- and mineral-dependent Over thecourse of 2 months at room temperature Al and Mnincreased by 350ndash5000 and 100ndash500 respectivelyamong the four rock standards (Rhodes and others 2011)Elemental dissolution rates varied substantially Mn concen-trations in the JG-2 leachate approximately doubled in thefirst 3 weeks then remained constant In contrast Alconcentrations changed at a relatively consistent rate overthe course of 12 weeks increasing by a factor of 4 over thistime interval (Rhodes and others 2011) In leachate from theNod-P-1 standard Al increased by a factor of 3 in the first12 hours then increased at a slowing rate until around2 weeks after which its concentration remained stable Thustrace-element dissolution depends on the mineralogy of therock standard used a finding that has direct implications forice-core studies because the composition of dust deposited atice-core sites changes as dust sources and transport changethrough time (eg Marino and others 2004 Revel-Rollandand others 2006 Bory and others 2010)

The significant increases observed in trace-elementconcentrations as a function of acidification time havedirect relevance to the different ICP-MS methodologiescurrently employed by the ice-core community As statedearlier the major difference among studies is the use ofcontinuous vs discrete ICP-MS sample analysis and theresulting differences in acid contact time Although we didnot analyze samples within seconds to minutes of acidifica-tion our results show that the concentrations of mostelements measured after 20min of acidification did notrepresent those achieved after 3 months (Fig 1) Knusel andothers (2003) undertook a direct comparison of convention-ally cut and decontaminated ice-core samples with samplesmelted continuously and analyzed using a continuous-injection ICP-MS They found that the continuous-injectionmethod achieved similar concentrations of relatively solubleelements such as Na Mg and Sr but measured significantlylower concentrations of Al V Fe As Rb Cd Sb Cs Ba LaCe Pr Nd Sm Pb Bi Th and U compared with theconventional method They attributed the lower concen-tration results of these lithogenic elements to reduced acidcontact time (Knusel and others 2003) It is clear that resultsobtained using these two approaches are not comparablebecause of the inherent differences in the duration ofacidification prior to analysis

Effect of variable elemental dissolution on calculatedcrustal ratios and crustal enrichment factors

Lithogenic elements such as Al Ce Fe La and Mn arecommonly used as crustal dust tracers (eg McConnell andothers 2007) In turn the enrichment of various trace

elements (such as Pb Bi and As) relative to their ratio to agiven lithogenic element (or elements) in the upper contin-ental crust (UCC) is used to infer the presence of non-crustalsources or of changes in dust provenance through time(Gabrielli and others 2005 Osterberg and others 2008Marteel and others 2009 Dixon and others 2013) Thecrustal enrichment factor EFc is defined as RsampleRUCCwhere R is the ratio of the element of interest to a givenreference element such as Al Although Reimann and DeCaritat (2000) argued that enrichment factors have limitedutility in environmental geochemistry they remain a widelyused tool within the ice-core community

To explore how differential dissolution rates affect calcu-lated crustal ratios and enrichment factors we evaluated alltrace elements in relation to Al Ce Fe La and Mn Whileour analysis suggests that many elements have sources otherthan continental dust (ie marine and volcanic aerosols)some elements do approach their UCC ratio after 1 monthsuggesting that with sufficient acidification time they may beused as dust indicators Figure 3 shows changes in theelemental ratios to Al through time for each acid treatmentfor Cs Ba La Ce Pr and K along with their ratio in the UCC(Wedepohl 1995) While all these elements approach theirUCC ratio to Al within 1ndash3 months (suggesting that they aredust-derived) it is clear that during the first month the ratiochanges dramatically Indeed BaAl LaAl CeAl and PrAlfall both below and above the UCC ratio during the firstweek of acidification Differences in the ratios among thethree acid treatments are most apparent within the firstmonth With time the ratios become more similar amongtreatments and closer to the UCC ratio We found similarresults for elemental ratios to Ce Fe La and Mn (not shown)These results suggest that dust flux estimates derived usingelemental concentrations and crustal ratios will depend onthe duration of acidification prior to analysis and may varydepending on the relative solubility of different minerals

In order to see how incongruent element leaching mightaffect interpretations about possible anthropogenic pollutionsources we calculated crustal enrichment factors of Cs PbAs Cr Zn and U using Al as a reference element and foundthat they all decreased with acidification time by a factor of13 (Pb) to a factor of 8 (Cs) (Fig 4) In addition theenrichment factors of Pb As Cr Zn and U showed increaseswithin the first month likely as a result of higher solubilityrelative to Al As Al continued to leach into solution theseenrichment factors all decreased to their final values asmeasured at 3 months

The differences in enrichment factors that we observedwith acidification time can be compared with those meas-ured during both preindustrial times and recent decades Wefind that the observed change in EFc (As) a factor of 26 isthe same as that observed on glacialinterglacial timescalesin the European Project for Ice Coring in Antarctica (EPICA)Dome C ice core This was interpreted to suggest additionalnon-dust sources or a possible shift in dust source area ortransport parameters (Gabrielli and others 2005) In con-trast EFc (Pb) measured in the Mount Logan ice coreCanada is a factor of 15 above natural background(Osterberg and others 2008) while we observed changes inEFc (Pb) of a factor of 14 Thus while our results suggestthat enrichment factors calculated within the first month ofdissolution are likely to be erroneously high they wouldlikely not be of the same order as those resulting fromanthropogenic pollution


Our results are supported by additional experimental andfield measurements Rhodes and others (2011) found thatdifferent rates of element dissolution from crushed rockstandards led to very different ratios relative to the UCC ratiothrough time They calculated crustal enrichment factors fora hypothetical dust created from four rock standards andleached at 1 nitric acid for 1 week and found that EFc (U)decreased by a factor of 2 using Al as a reference element(Rhodes and others 2011) This is comparable with thechange in EFc (U) that we observed over 3 months Grottiand others (2011) determined that compared with a full

acid digestion lt20 of Fe and Al contained in Antarcticdust dissolved in a 05 nitric acid treatment Theycautioned against the calculation of crustal enrichmentfactors using these elements unless fully digested

One approach to minimize possible enrichment factorvariation due to incongruent dissolution of trace elements isto use multiple reference elements (Osterberg and others2008) Given our results that lithogenic elemental ratiosapproached unity with the UCC ratio and that calculatedenrichment factors of trace metals remained relativelyconstant after 1 month we suggest that samples to be used

Fig 3 Ratios of selected elements to Al with acidification time for each acid treatment measured in snow-pit samples from Antarctica UCCratios from Wedepohl (1995)

Fig 4 Calculated crustal enrichment factors of selected elements in Antarctic snow-pit samples using Al as a reference element for eachacid treatment


for the calculation of enrichment factors are allowed to sitacidified at room temperature for at least 1 month prior toanalysis However it must be kept in mind that incompletedissolution of minerals will likely lead to some degree ofoverestimation of enrichment due to incongruent leachingtherefore complete digestion using HF is the best approachwhen the intent is to calculate crustal enrichment

Concentration dependence of measured elementalconcentration increase

The Denali snow-pit data allow us to observe concentrationchanges over an extended time period (383 days) and toevaluate the effect of initial elemental concentration on theobserved concentration increase during this time intervalWe focus on four elements with different solubilities S CaAl and Fe Al samples were treated with 1 (vv) nitric acidFigure 5 shows the measured concentrations of these fourelements with depth for both analysis times The S curvesare indistinguishable while the Ca curves are very similarbut with some noticeable increases in concentration at thepeak concentrations (eg at 34 96ndash150 595 and 894 cmdepth) The Al and Fe curves show enhanced concentrationsfor nearly all values with time and larger concentrationchanges within the peaks (as with Ca) Repeat measurementsof instrumental check standards demonstrate the reproduci-bility of these data (Table 3) Figure 6 shows the concen-tration difference between 2009 and 2008 relative to theinitial (2008) concentration The S data remain around 0 and

Fig 5 Concentrations of S Ca Al and Fe in Denali snow-pit samples analyzed 383 days apart in 2008 and 2009

Table 3 Concentrations (mg Lndash1) of internal check standard run with

the Denali snow-pit samples in 2008 and 2009

32S 44Ca 27Al 56Fe

Check standard value 50 50 1 1

Measured concentrations (2008)Analysis 1 4732 5181 105 105Analysis 2 5185 5062 103 109Analysis 3 5035 4943 105 108Analysis 4 4930 4899 101 0997Analysis 5 4853 4906 0974 104Analysis 6 4787 5176 106 106

2008 mean 4920 5028 103 105

2008 SD 168 131 003 003


2009 mean 5020 5072 102 101

2009 SD 140 125 001 002


show no relationship with initial concentration (linearregression slope =002 r=025) The Ca data show a weaklinear relationship between the change in concentration andthe initial concentration (linear regression slope = 017r=074) In contrast the Fe and Al data show strong positivelinear relationships between measured concentration differ-ence and initial concentration (Fe linear regression slope =166 r=098 Al linear regression slope= 175 r=097)While Ca Fe and Al all exhibit a linear relationship betweeninitial concentration and concentration change with time itis clear that the concentration dependence is much greaterfor Fe and Al than for Ca suggesting that these elements arebeing leached from relatively refractory mineral phasesRhodes and others (2011) used three different concen-trations of the W-2 rock standard to look at how leachateconcentration related to dust concentration They foundstrong linear correlations (r = 098ndash100) between dustconcentration and leachate concentration after 12 hoursfor Sr and Ce in agreement with our results

Potential implications for snow and ice-core datainterpretation

The finding that analytical methods affect measured trace-element concentrations is not unique It has been welldocumented in the oceanographic community as marinechemists have sought to make clean reproducible measure-ments of Fe and other trace elements in the upper ocean(Bruland and others 1979 Achterberg and others 2001Bowie and others 2010) and in dust aerosols (Sholkovitzand others 2012) However a quantitative understanding ofhow methods affect measured chemical concentrations inice cores is just beginning to emerge (Knusel and others2003 Ruth and others 2008 Rhodes and others 2011)Because of the significant differences we observed betweenmeasured elemental concentrations at different acidstrengths and times we emphasize that it is essential totreat all samples within individual studies in an identicalmanner Moreover given the variety of methods currentlyemployed in ice-core studies it is unlikely that twolaboratories measuring trace elements in the same samples

would achieve the same concentrations unless they usedidentical methods

To estimate the differences between continuous anddiscrete approaches to ICP-MS analysis of snow andice-core samples we use our 20min acidification data asa proxy for continuous-injection ICP-MS and compare withour 1month data (1 (vv) nitric acid treatments for bothcases) We find that Al increases from 05340096 to

174 0226mg Lndash1 an increase of 232 and Fe increases

from 04020040 to 145 0187mg Lndash1 an increase of260 Pb increases from 4481 0450 to 115051025ng Lndash1 an increase of 157 while EFc (Pb) decreasesby a factor of 14 from 41161 to 29812 SimilarlyAs increases from 1569 0234 to 2481 0335ng Lndash1 anincrease of 158 while EFc (As) decreases by a factor of 26from 1177 100 to 438 47 Thus both elementalconcentrations and calculated crustal enrichment factorsshow significant differences between these two methodo-logical approaches

Based on these results we suggest that trends and generalpatterns (ie locations of peaks and troughs in a time-seriesrecord) should be consistent among different studies whileabsolute concentrations and peak-to-trough amplitudes arelikely to differ significantly depending on the acidificationmethod used Samples to be used for calculation ofatmospheric fluxes and crustal enrichment factors must beacidified for at least 1 month prior to analysis to preventpotentially erroneous interpretations of trace-element dataand researchers should bear in mind that incomplete andincongruent leaching is likely to produce overestimates ofcrustal enrichment Complete digestion using HF offers thebest way to quantify deposition of the full range of traceelements (Correia and others 2003 Grotti and others2011) Incorporating HF digestion even at low temporalresolution in ice-core studies would allow for the quantifica-tion of recovery rates which will be a useful way to compareresults obtained from different laboratories More import-antly quantification of recovery rates would provide somelevel of standardization among studies conducted in differ-ent geographic locations allowing for more accurate

Fig 6 Change in concentration over 383 days of acidification (2009 ndash 2008) vs initial concentration (2008) for S Ca Al and Fe measured ina Denali snow pit Linear regressions are given for each element and the y= x line is plotted for reference


interpretations of spatial and temporal variability in trace-element deposition Given the range of evidence presentedhere we suggest that continuous-injection ICP-MS is not anappropriate tool for the quantitative analysis of Al Ba CdCe Co Cr Cu Fe La Li Mg Mn Pb Pr Sr Ti U V or Zn inice-core samples Further we believe there is a pressingneed for a laboratory intercomparison study within the ice-core trace-element community

Our results also have implications for the use of trace-element concentrations as proxies for past climate variabilityProxy development relies on finding meaningful statisticallysignificant relationships between trace-element concentra-tion records and physical components of the climate systemsuch as zonal wind strength or sea-ice cover (Mayewski andothers 1994 Kreutz and others 2000 Goodwin and others2004 Yan and others 2005 Dixon and others 2012)Relationships are established using observational or climatereanalysis data in the modern era and interpretations areextended into the past using the ice-core proxy data Becausethe commonly used Pearsonrsquos linear correlation coefficientdepends on actual values (rather than ranked values) ourdata suggest that the strength of correlation will depend onthe relative concentrations of trace elements Thus differentacidification methods could lead to potentially different so-called lsquocalibrationsrsquo of ice-core data and therefore differentinterpretations about past climate variability particularly interms of the magnitude of events

CONCLUSIONS

We conducted a 3month experiment to test the effects ofacid strength and acidification time on measured trace-element concentrations and calculated crustal enrichmentfactors in snow samples from West Antarctica In additionwe used snow-pit samples from Alaska to evaluate theconcentration dependence of measured increases in litho-genic element concentrations through time Our analysesbuild on previous work (Knusel and others 2003 Grotti andothers 2011 Rhodes and others 2011) by providing clearevidence that trace-element relative concentrations inenvironmental samples depend on both acid strength andacidification time Our results suggest the followingconclusions

1 Acid strength and acidification time significantly in-crease measured trace-element concentrations leachedfrom impurities in snow samples collected from remotelocations Ice-core trace-element studies should allow atleast 1 month for dissolution of particulate material inorder to achieve a representative acid-leachable fractionAll studies should quantify the recovery rates achievedby their analytical methods using HF digestion

2 Incongruent dissolution of elements leads to widelyvarying elemental ratios relative to their UCC ratiosthrough time Although we observed that many elementsapproached their crustal ratios after 1 month ofacidification we caution that complete congruentelement dissolution can be achieved only through a fullacid digestion Therefore crustal enrichment factorscalculated for samples that have not been digestedshould be interpreted with this caveat in mind

3 Lithogenic element dissolution is linearly dependent ondust concentration though the slope of this relationship

is element-specific and may change with dust lithologyBecause relative trace-element concentrations dependon the acidification method used trace-element proxiesof past climate variability may not accurately representthe magnitude of past variability in the climate system

4 An interlaboratory comparison study is needed withinthe ice-core trace-element community

ACKNOWLEDGEMENTS

This work was supported by US National Science Founda-tion grants ANT-0636740 AGS-1203838 and ARC-0713974and by the University of Maine Dissertation ResearchFellowship to BGK We thank the WAIS Divide ScienceCoordination Office Ice Drilling Design and OperationsGroup the National Ice Core Laboratory Raytheon PolarServices Company and the 109th New York Air NationalGuard for field support in Antarctica We also thank TimothyBartholomaus Thomas Bauska Seth Campbell John Fegy-veresi Shelly Griffin Jonathan Hayden Logan MitchellAnaıs Orsi Mike Waskiewicz and Gifford Wong for fieldand laboratory assistance We thank Nelia DunbarStephen Norton Rachael Rhodes and an anonymousreviewer for providing helpful comments which improvedthe manuscript

REFERENCES

Achterberg EP Holland TW Bowie AR Mantoura RFC andWorsfoldPJ (2001) Determination of iron in seawater Anal Chim Acta442(1) 1ndash14 (doi 101016S0003-2670(01)01091-1)

Bory A and 6 others (2010) Multiple sources supply eolian mineraldust to the Atlantic sector of coastal Antarctica evidence fromrecent snow layers at the top of Berkner Island ice sheet EarthPlanet Sci Lett 291(1ndash4) 138ndash148 (doi 101016jepsl201001006)

Bowie AR Townsend AT Lannuzel D Remenyi TA and Van derMerwe P (2010) Modern sampling and analytical methods forthe determination of trace elements in marine particulatematerial using magnetic sector inductively coupled plasmandashmass spectrometry Anal Chim Acta 676(1ndash2) 15ndash27 (doi101016jaca201007037)

Bruland KW Franks RP Knauer GA and Martin JH (1979) Samplingand analytical methods for the determination of coppercadmium zinc and nickel at the nanogram per liter level insea water Anal Chim Acta 105(1) 233ndash245 (doi 101016S0003-2670(01)83754-5)

Campbell S and 7 others (2012) Melt regimes stratigraphy flowdynamics and glaciochemistry of three glaciers in the AlaskaRange J Glaciol 58(207) 99ndash109 (doi 1031892012JoG10J238)

Correia A and 6 others (2003) Trace elements in South Americaaerosol during 20th century inferred from a Nevado Illimani icecore Eastern Bolivian Andes (6350masl) Atmos Chem PhysDiscuss 3(3) 2143ndash2177 (doi 105194acpd-3-2143-2003)

Cwiertny DM Young MA and Grassian VH (2008) Chemistry andphotochemistry of mineral dust aerosol Annu Rev Phys Chem59 27ndash51 (doi 101146annurevphyschem59032607093630)

Dixon DA and 6 others (2012) An ice-core proxy for northerly airmass incursions into West Antarctica Int J Climatol 32(10)1455ndash1465 (doi 101002joc2371)

Dixon DA and 6 others (2013) Variations in snow and firn chemistryalong US ITASE traverses and the effect of surface glazingCryosphere 7(2) 515ndash535 (doi 105194tc-7-515-2013)

Edwards R Sedwick P Morgan V and Boutron C (2006) Iron in icecores from Law Dome a record of atmospheric iron depositionfor maritime East Antarctica during the Holocene and Last


Glacial Maximum Geochem Geophys Geosyst 7(Q12)Q12Q01 (doi 1010292006GC001307)

Fischer H Siggaard-Andersen M-L Ruth U Rothlisberger R andWolff E (2007) Glacialinterglacial changes in mineral dust andsea-salt records in polar ice cores sources transport anddeposition Rev Geophys 45(RG10) RG1002 (doi 1010292005RG000192)

Gabrielli P and 9 others (2005) Variations in atmospheric traceelements in Dome C (East Antarctica) ice over the last twoclimatic cycles Atmos Environ 39(34) 6420ndash6429 (doi101016jatmosenv200507025)

Gabrielli P and 11 others (2010) A major glacialndashinterglacialchange in aeolian dust composition inferred from Rare EarthElements in Antarctic ice Quat Sci Rev 29(1ndash2) 265ndash273(doi 101016jquascirev200909002)

Gaspari V and 9 others (2006) Atmospheric iron fluxes over the lastdeglaciation climatic implications Geophys Res Lett 33(3)L03704 (doi 1010292005GL024352)

Goodwin ID Van Ommen TD Curran MAJ and Mayewski PA(2004) Mid-latitude winter climate variability in the south Indianand southwest Pacific regions since 1300 AD Climate Dyn22(8) 783ndash794 (doi 101007s00382-004-0403-3)

Gross BH and 6 others (2012) Constraining recent lead pollutionsources in the North Pacific using ice-core stable leadisotopes J Geophys Res 117(D16) D16307 (doi 1010292011JD017270)

Grotti M Soggia F Ardini F and Magi E (2011) Major and traceelement partitioning between dissolved and particulate phases inAntarctic surface snow J Environ Monitor 13(9) 2511ndash2520(doi 101039C1EM10215J)

Hong S and 9 others (2004) Atmospheric heavy metals in tropicalSouth America during the past 22000 years recorded in a highaltitude ice core from Sajama Bolivia J Environ Monitor 6(4)322ndash326 (doi 101039B314251E)

Knusel S Piguet DE Schwikowski M and Gaggeler HW (2003)Accuracy of continuous ice-core trace-element analysis byinductively coupled plasma sector field mass spectrometryEnviron Sci Technol 37(10) 2267ndash2273 (doi 101021es026452o)

Krachler M Zheng J Fisher D and Shotyk W (2005) Analyticalprocedures for improved trace element detection limits in polarice from Arctic Canada using ICP-SMS Anal Chim Acta530(2) 291ndash298 (doi 101016jaca200409024)

Krachler M Zheng J Fisher D and Shotyk W (2008) Atmosphericinputs of Ag and Tl to the Arctic comparison of a high resolutionsnow pit (AD 1994ndash2004) with a firn (AD1860ndash1996) and an icecore (previous 16000 yr) Sci Total Environ 399(1ndash3) 78ndash89(doi 101016jscitotenv200803006)

Kreutz KJ and Sholkovitz ER (2000) Major element rare earthelement and sulfur isotopic composition of a high-elevation icecore sources and transport of mineral dust in central AsiaGeochem Geophys Geosyst 1(11) 1048 (doi 1010292000GC000082)

Kreutz KJ Mayewski PA Pittalwala II Meeker LD Twickler MS andWhitlow SI (2000) Sea-level pressure variability in the Amund-sen Sea region inferred from a West Antarctic glaciochemicalrecord J Geophys Res 105(D3) 4047ndash4059 (doi 1010291999JD901069)

Marino F Maggi V Delmonte B Ghermandi G and Petit JR (2004)Elemental composition (Si Fe Ti) of atmospheric dust over thelast 220 kyr from the EPICA ice core (Dome C Antarctica) AnnGlaciol 39 110ndash118 (doi 103189172756404781813862)

Marteel A and 11 others (2009) Climate-related variations in crustaltrace elements in Dome C (East Antarctica) ice during the past672 kyr Climatic Change 92(1ndash2) 191ndash211 (doi 101007s10584-008-9456-3)

Mayewski PA and 13 others (1994) Changes in atmosphericcirculation and ocean ice cover over the North Atlantic duringthe last 41000 years Science 263(5154) 1747ndash1751 (doi101126science26351541747)

McConnell JR Lamorey GW Lambert SW and Taylor KC (2002)Continuous ice-core chemical analyses using inductivelycoupled plasma mass spectrometry Environ Sci Technol36(1) 7ndash11 (doi 101021es011088z)

McConnell JR Aristarain AJ Banta JR Edwards PR and Simoes JC(2007) 20th-century doubling in dust archived in an AntarcticPeninsula ice core parallels climate change and desertificationin South America Proc Natl Acad Sci USA (PNAS) 104(14)5743ndash5748 (doi 101073pnas0607657104)

Osterberg EC Handley MJ Sneed SB Mayewski PA and Kreutz KJ(2006) Continuous ice-core melter system with discretesampling for major ion trace element and stable isotopeanalyses Environ Sci Technol 40(10) 3355ndash3361 (doi101021es052536w)

Osterberg E and 10 others (2008) Ice-core record of rising leadpollution in the North Pacific atmosphere Geophys Res Lett35(5) L05810 (doi 1010292007GL032680)

Reimann C and De Caritat P (2000) Intrinsic flaws of elementenrichment factors (EFs) in environmental geochemistry EnvironSci Technol 34(24) 5084ndash5091 (doi 101021es001339o)

Revel-Rolland M and 7 others (2006) Eastern Australia a possiblesource of dust in East Antarctica interglacial ice Earth PlanetSci Lett 249(1ndash2) 1ndash13 (doi 101016jepsl200606028)

Rhodes RH Baker JA Millet M-A and Bertler NAN (2011)Experimental investigation of the effects of mineral dust on thereproducibility and accuracy of ice-core trace element analysesChemical Geol 286(3ndash4) 207ndash221 (doi 101016jchemgeo201105006)

Ruth U and 15 others (2008) Proxies and measurement techniquesfor mineral dust in Antarctic ice cores Environ Sci Technol42(15) 5675ndash5681 (doi 101021es703078z)

Sholkovitz ER Sedwick PN Church TM Baker AR and Powell CF(2012) Fractional solubility of aerosol iron synthesis of a global-scale data set Geochim Cosmochim Acta 89 173ndash189 (doi101016jgca201204022)

Vallelonga P Van de Velde K Candelone JP Morgan VI Boutron CFand Rosman KJR (2002) The lead pollution history of Law DomeAntarctica from isotopic measurements on ice cores 1500 AD to1989 AD Earth Planet Sci Lett 204(1ndash2) 291ndash306 (doi101016S0012-821X(02)00983-4)

Wedepohl KH (1995) The composition of the continental crustGeochim Cosmochim Acta 59(7) 1217ndash1232 (doi 1010160016-7037(95)00038-2)

Yan Y Mayewski PA Kang S and Meyerson E (2005) An ice-coreproxy for Antarctic circumpolar zonal wind intensity AnnGlaciol 41 121ndash130 (doi 103189172756405781813294)

MS received 5 July 2013 and accepted in revised form 22 October 2013


differential dissolution rates could have on calculatedcrustal enrichment factors In addition we used a timeseries of snow-pit samples collected from Denali NationalPark Alaska USA and measured twice (383 days apart at1 (vv) nitric acid) to investigate the concentrationdependence of measured increases in trace-element con-centrations through time Our study expands the work ofRhodes and others (2011) by testing the effects of both acidstrength and time on real environmental samples fromAlaska and Antarctica two important ice-core drillinglocations Dust particles deposited in the snowpack at bothsites are likely to have experienced significant weatheringrelative to their source rock particularly during atmospherictransport (Cwiertny and others 2008) Our results aredirectly relevant to ice-core studies of trace-element de-position In addition we evaluate the concentrationdependence of relative changes in four trace-elementconcentrations (Al Ca Fe and S) measured over 1 year ofacidification Our results imply that ice-core time-seriesrecords developed after relatively short acidification periodslikely underestimate both the magnitude and the variabilityof trace-element concentrations potentially leading to biasin paleoclimate calibration work which in turn couldimpact interpretations of past climate variability based ontrace-element proxies

METHODS

Sample collection WAIS Divide Antarctica

Snow samples were collected from a 26m snow pit dug in2008 at the West Antarctic ice sheet (WAIS) Divide ice-coresite (794688 S 1120868W 1766masl) within the clean-air sector of camp Once the snow pit was excavated onewall was cut back an additional 20 cm using a plasticshovel cleaned with Citranox detergent and Millipore Milli-Q deionized water The wall was subsequently scrapedback an additional 4 cm using plastic scrapers that hadbeen decontaminated using nitric acid and Milli-Q waterSampling personnel wore Tyvek clean suits and doublepairs of gloves gloves were replaced whenever theybecame torn Snow samples were collected using aspecially built pure Ti sampling tool which had beenwashed with Citranox detergent and Milli-Q water Eachacid-washed sample bottle was double-bagged for storageand shipment using plastic Ziploc bags to reduce contactwith the environment Samples were shipped frozen to theUniversity of Maine and stored in a freezer until meltingand analysis

Experimental set-up

We combined Antarctic snow from several sample bottlesand melted it in an acid-cleaned capped Teflon1

perfluoroalkoxy (PFA)-coated bottle using a warm waterbath Once the snow was fully melted we homogenized itby shaking and inverting the bottle then poured the waterinto three additional PFA bottles (110mL each) We thenadded Optima grade nitric acid (Fisher Scientific) to eachbottle resulting in volumetric concentrations of 01 10and 10 acid respectively We evaluated possible sorptionof metals to the walls of the bottle used for melting thesnow by swishing the bottle with a 10 nitric acid solutionMetal sorption was negligible (05 of initial measuredconcentrations)

We analyzed samples after 20 min 24 hours 5 days14 days 28 days 56 days and 91 days At each analysis timewe poured five replicates from each bottle (to avoid potentialcontamination from a pipette) into acid-cleaned 4mLpolypropylene (PP) vials for analysis Samples were eitheranalyzed immediately or frozen until later analysis Weanalyzed samples using a Thermo Finnigan Element IIinductively coupled plasma sector-field mass spectrometer(ICP-SFMS) with an ESI Apex sample introduction system anda PFA-ST nebulizer Instrumental settings for the ICP-SFMSare given in Table 1 Procedural blanks (Milli-Q water) weretreated identically and analyzed in parallel (eg 01 10 and100 (vv) Optima grade nitric acid concentrations) Allbottles were shaken periodically (about once a week) andprior to subsampling and were kept in the dark at roomtemperature All sample handling was conducted in a class1000 clean room under a laminar flow High EfficiencyParticle Air (HEPA) bench Unfortunately because we pouredoff replicates (rather than using a pipette for volumetricaccuracy) we ran out of sample earlier than planned andwere not able to quantify recovery rates using hydrofluoricacid (HF) digestion

All five replicates were included in the calculations ofmean concentration and standard deviation the error barsshown are one standard deviation (1SD) of the meanconcentration for each analysis time Because melted snowand ice samples include both dissolved and suspendedimpurities they are inherently heterogeneous therefore thestandard deviation of each set of replicates likely representsreal sample variability We subtracted the average of fiveblank replicate measurements (for each acid treatment ateach analysis time) from each mean sample concentrationto correct for potential analytical inconsistencies Blankconcentrations remained well below the sample concen-trations throughout the experiment (over an order ofmagnitude for most elements Table 2) Data from this studywill be made publicly available through the US NationalSnow and Ice Data Center (NSIDC)

Sample collection and treatment Denali NationalPark Alaska

Snow samples were collected from a 4m pit dug at KahiltnaPass Denali National Park in 2008 (630768 N1511748W 2957masl Campbell and others 2012)Samples were collected at 5 cm resolution using trace metalclean techniques (as described above) and were stored inacid-washed bottles Snow samples were shipped frozenthen melted and acidified to 1 (vv) with Optima gradenitric acid under a HEPA bench Samples were analyzedusing ICP-SFMS at two time intervals spaced 383 days apartand were stored at room temperature in the interveningtime All sample handling was conducted in a class 1000clean laboratory

Table 1 ICP-SFMS instrument conditions and measurement par-ameters

Forward power 1280WCool gas 16 Lminndash1

Auxiliary gas 095 Lminndash1

Sample gas 08 Lminndash1

Additional gas 02 Lminndash1

Sample uptake 100mLminndash1

Number of scans 15




















































32S 44Ca 27Al 56Fe



2008 mean 4920 5028 103 105

2008 SD 168 131 003 003


2009 mean 5020 5072 102 101

2009 SD 140 125 001 002














CONCLUSIONS







ACKNOWLEDGEMENTS


REFERENCES





























































































32S 44Ca 27Al 56Fe



2008 mean 4920 5028 103 105

2008 SD 168 131 003 003


2009 mean 5020 5072 102 101

2009 SD 140 125 001 002














CONCLUSIONS







ACKNOWLEDGEMENTS


REFERENCES























































CONCLUSIONS







ACKNOWLEDGEMENTS


REFERENCES











































Dependence of ice-core relative trace-element …eosterberg/images/Koffman et al., 2014...Dependence...

Documents

Transcript of Dependence of ice-core relative trace-element …eosterberg/images/Koffman et al., 2014...Dependence...