J Ollmal cifG'laclolog)I, r ol. 42. S o. 141, 1996

Itnpurity influence on nortnal grain growth in the GISP2 ice core, Greenland

R . B. ALLEY A:,\D G. A. VVOO DS

Ear/h s,.)'stem Science Center and D e/Jar/mmt of G'eosciell(Ps. The PeJlns)lil'allia Slate [;-Ilil'ersi{j', CI//z'ersl[J' Park . Penll.~l'lvallia 16802. C .S. A .

ABSTRACT. ] nte rcept a na lys is o f a pproxima te ly bi-yea rl y ve rti ca l thin sec ti ons from th e upper pa rt o f th e G ISP2 ice core, centra l G reenl a nd , shows th a t g ra in-size ra nges increase with inc reasing age . This d emonstra tes th a t somethin g in th e ice affects g ra in-g rowth ra tes, a nd th a t gra in-size ca nnot be used di rect ly in pa leo th e rmometry as has bee n proposed. C o rrela ti on of g ra in-grow th ra tes to c hemi cal a nd iso topic da ta indi ca tes slower grO\\-th in ice with hi g he r impurit y concentrat ions, a nd es pec ia ll y slow gro""th in "fores t-fire" layers conta ining ab und a nt a mm onium ; howe \·e r. th e im p ur ity/g ra in-growth rela ti ons a re qui te noisy. LiLl Ic co rre la ti on is fo und between g row th ra te and isoto pi c compositi o n o f ice .

INTRODUCTION

I n th e iso th erm a l firn a nd sha ll o\\' ice of cold ice shee ts, g ram grow th is dri ve n b y th e energy a nd CUr\ 'Li ture o f g ra in bo unda ri es (" soap-b ubbl e g row th " ), a nd (l\'e rage g ra in c ross-sec ti ona l a rea A increases linea rl y \\'ith age t acco rding to

A = An + K t (1)

where Ao is th e initi a l m'C rage g ra ll1 c ross-sect iona l area a nd 1\- sho\\-s an Arrh e ni us depend ence o n tempera ture a nd may depe nd on o th e r [acto rs (Gow, 1969 ). U nusua ll y hi g h so lubl e a nd in so lubl e im p urit y co ncentrati o n s contribute to 510\\- g ro w th rates in co ld ice (red uced 1(; Gow a nd \ \ ' il lia mso n , 1976 ).

I ce fro m the \\-isconsina n rth e las t g lac ia l peri od ) in co res from D ome C a nd \' ostok, A nta rc ti ca, fa ll s in th e d ep th ra nge of "soa p- b ub ble" gra in grow th but exhibits ano m a lo usly sma ll gra i n-sizes com pa red to o\'erly i ng H olocene ice e\'C n a ft e r co rrect ion fo r th ei I' differe n t tempe ra ture histo ri es using the Arrh enius d ependence o f g ra in-g row th ra te o n te mpera ture (Du va l a nd Loriu s, 1980; Pe tit a nd o th e rs, 1987 ). Poss ible ex pl a na ti ons fo r th ese sma ll g ra in-siz ies ine lude redu ced J( in Eq ua ti on ( I ) owing to drag eflcc ts o f th e hi gher (but still quite low ) impurity concentra ti o ns in \\ 'isconsin a n ice (Alley a nd o th ers, 1986a, b) o r o wing to a "m em o ry" of th e lower \\' isconsi na n firnifi ca t io n temperatu re i m printed on th e ice stru c ture a t densiti es less th a n 550 kg m :l (roughl y th e upper 10- 20 m ) a nd th e n " quenched " (P e ti t a nd o th e rs, 1987 ) .

Beca use of th e cl ose co rrela ti o n o f firn ifi cati o n te mpe ra ture a nd impurity loadin g in \ Vi sconsina n ice from centra l E as t Anta rct ica, th e field da ta d o not a ll o w sepa ra ti o n of th ese po tcn ti a l con tro Is. V a ri ous phys ica l

argume llls ha\T bee n prese nted (All ey a nd o th ers, 1988; Pet i t a nd o th e rs, 1988; P a te rson, 199 1). I n pa rti cul a r , D LI\'a l a nd Lo rius ( 1980 ) and Peti t a nd o th e rs ( 1987, 1988 ) a rg ued th a t th e so lubl e and inso lubl e impurity concentra ti o ns of H oloce ne a nd \ \' isco nsi n a n ice fro m inland sites in A nta rcti ca a nd G reenl a nd a rc too 10\\- to

ha\T a ITe-c ted g ra in-growth ra tes measura bl y. ] n co nt ras t , All ey a nd o th e rs ( 1986a, b , 1988 ) a rgued th a t d rag effec ts from a g i\'e ll impuri ty can va ry by ord ers o f m agllitud e depenclin g- o n th e sta te of th e impu rity in the ice, a nd th a t signi fica nt impurit y-d rag effec ts arc proba ble in ice-shee t ice fo r likel y impurit ), di stributions.

J ouze l a nd o th ers ( 1987 ) a nd Petit a nd o th e rs ( 1987 ) used th e ass um ed rela ti on be twee n [irnifica ti o n tempera­tu re a nd g row th ra te as a pa leo th erm omete r fo r Anta rcti c cores. If im p urity effec ts a rc sig nifi ca nt o r d o min a nt, o r if a tempel'a tulT " memory" d oes no t a ffec t g ro \\- th ra tes, th en g ra in-size is no t a n a ppro pria te pa leothe r mome ter.

Seaso na l cycles a ll ow a pa rti a l tes t o f' thi s a rg um ent. All ice fro m a d epositiona l yea r in the uppel- pa rt of an ice shee t has ex pe ri enced th e sa m c stra in hi sto r y, a nd th e sa me di age ne ti c-tcmpera tul-e history bcl o \\' th e uppe r meter o r two, but seasona l va ri a ti ons in impurity loadings a rc as la l-gc as, or la rge r tha n, th e fcw-fold differences bet\\Te n \Vi sconsina n a nd H o locene ice fro m centra l Eas t Antarc ti ca.

If g ra in-g ro \\·th rates \\' ithin a yea r diffe r, then so methin g o th e r tha n a m e m o ry of di age n e ti c tem per­a tu re is a ffec tin g th e g row th ra tes, and gra in-size cann ot be use d direc tl y in pa leo th e rm ome tr y . Cor re lat io n betwee n g ro wth ra te a nd impurity loadin g th en wo uld sugges t im pu ri ty-d rag effec ts, and corre la ti o n betwee n growth ra te a nd iso topi c composition mi g ht sugges t some memo ry of d epos iti ona l condi tions or di age nes is in th e fi rs t yea r o r so, o r a tru e iso to pi c erIee t. No ti ce, howC\'er , th a t such a stud y cann o t d irec tl y address th e P e tit and

255

Journal cif Glaciolog}!

others (1987) hypothes is of ex istence of a tempera tu re memory, beca use a ll of th e samples will ha \'e experienced essenti a ll y the same diageneti c-temperature hi sto ry as they a pproac hed th e hypothes ized "quenching" level in the fim ; we ca n onl y rest whether effe cts other than a temperature memory a re acti ve .

METHODS

Th e G ISP2 ice co re was drill ed during th e summers of 1989- 93 at 72.6° N, 38 .5° W , 3200 m ele\'a ri on, 28 km west of th e summit of the Greenland ice sheet, by the Pola r Ice Coring Office . Mean annua l (20 m) tempera ­ture is a bout - 31°C, and accumula tion is a bour 240 mm ice year I (Alley and K oci , 1990; Alley and others, 1993 ) . Dating used here is from Alley and others (1993 ) a nd \1eese a nd others (1994). A full suite of ph ysical­properties studies was conducted on the co re, including meas urem ent of c-ax is fabric s a nd g ra in-sizes on horizon ta l and 10 cm vertical thin sec tions, densiti es a nd sonic ve lociri es (Gow a nd others, 1993; Anandakrishna n and others, 1995; unpubli shed information from A.J . Gow, 1996). H ere, wc fo cus on a single experim en t to

address controls on norma l gra in growth. T ypically with in days of retri eving co res, verti cal thin

scc tions werc cut a long the sides of selec ted core sec tions using a band-saw . (For the region of very brittle ice be twe en 680 and 1370 m, some sec tion s were c ut immediately and some a fter a yea r of storage at the G [SP2 site at - 30°C to a llow this brittl e ice to r("lax ; no differences in the grain structure were obseryed after the storage, except th a t the ice fractured into small pi eces when worked soon a fter cores were re tri eved, but was easier to work with after storage.) After the back of each ice sample was sanded Oa t, th e ice was affixed to a g lass pl a te using eith cr cya noacryla te a dh esive or sli g ht warm ing of the p late followed by " freezing-on" (we observed no differences in gra in structure between th ese two techniques, but th e cya noacrylate gave sec tions that were eas ier to observe because it avo ided the " bubbled­ou t" a i r trapped between the ice and glass of some deeper frozen-on samples) . Th e ice then was mi crotomed to an ap propri a te thi ckness (roughl y 0.5mm ) for petrographic study. The sa mples were photograph ed between cro sed pol a ri zers. Sec tions as long as 18cm were prepared , w ith use of consecutive sec tions to extend continuous coverage to 50 cm in many regions. Sample widths typically were 8- 10 cm.

Grain-sizes were measured using the mean-intercept techniqu e. In this, a tes t line is drawn on the section , a nd the length of line di\'ided by the number of grains crossed gives the grain-size. From stereologica l experience a nd theory (see Underwood , 1970, ch. 4; Alley, 1987a, b ), as long as the grain shapes do not change grea tl y wi th depth , any measure of th e mean gra in cross-sectional a rea will be related to the squa re of the mean intercep t length by a cons tan t geometrical factor of order I which correc ts for the sec tion effec t (m a ny grains a re cut nea r an edge ra th er th an through th e center; Gow, 1969) and for the geometry of th e sampling (see U nd erwood , 1970, tab les 4 . 1 and 4 .2) . We used th e geo m e tri cal facto r for monosized spheres foll owing nd ervvood (1970) , as

256

discussed in Alley (1987 b) . Grain cross-sec tio ns remained nea rl y eq ua n t and convex throughout the upper part of th e core th a t wc consider in deta il , although with a slight tendency to become Oa ttened with elonga tion parallel to the ice-shee t surface.

Use of mea n in tercept ra ther than some a real measure (e.g. Cow, 1969; Alley, 1987a, b ) is preferred here because of the ease of measuremen t a nd the high stratigraphic resolution . Ambiguity would be introduced to thi s stud y if we were forced to choose whether a given grain O\'er­lapping a stratigrap hi c boundary " belonged " to one of the two laye rs or to both. A layer-parallel test line ca n always be assigned to a layer, and meas ures the grain-size in th a t layer. Pre\·ious use of thi s m ean-intercept rechniqu e in glaciology includes that of Thorsteinsson and others (1995; see also Alley, 1987a, b) .

:\fIost of th e ohse rvations were made by one of us (G.A .W. ) and are detailed in Woods ( 1994) . \'\' e meas ured gra i n-size wi th i n i ndi vid ual s tra ti gra phi c laye rs by using horizontal tes t lines only. VI/ e placed a test line typicall y eye ry 2 mm a long the co re, a nd test lines sa mpled typica ll y 30 grain s or more. Thus, a sample of 100 gra in crossings is obtained typicall y every 6- 8 mm along th e core. Grain-sizes we re ave raged over 2- 3 cm for comparison to chemi cal and isotopic da ta, as desc ribed below.

Grain-g rowth studies such as Cow (1969) and Duval and Lorius (1980) are based on the average g ra in-size, Ji, in a sample contain ing a wide range of grain-sizes. Meas urem ent of a la rge enough number of grains is required to a llow accurate de termination of the average grain-size, as di scussed in Und erwood (1970, p. 12- 18; see also summary in Alley, 1987b, p. 66- 68 ) . T ypi ca ll y, average g ra in-sizes based on measuremen t of a few tens to a few hundreds of grain a re suffici ent to de termine Ji, with in a few per cent or less .

Major ions and ca ti ons were measured by th e Glacier R esea rch Group, University of New H ampshire (e.g . Mayews ki a nd others, 1993 ) . Impurity concen trations are similar to or a li ttle lower than typical val ues ci ted by Petit and others (1987 , 1988 ) for other inla nd sires (e.g. a few ppb to a few tens of ppb Na + for GISP2, vs 20-40 ppb N a +, ci ted by Pe tit and others ) . Sta ble-iso top ic compositio ns were measured by the Quaternary Iso tope Laboratory , U niYersi t y of \V as hingto n , and at th e Institute for Arcti c a nd Alpine Research , University of Colorado (Grootes and others, 1993 ). Verti cal sampling in tervals for chemistry a nd istopes were typically 2- 3 cm.

RESULTS

Grain-size data (mean inte rcept ) are shown in Figure I for the enlire core (vV oods, 1994). Woods (1994) iden­tified four grain-growth regimes:

(I) Grain-size increase consistent wi th Equ a tion (1), th e linea r area/age model of Gow (1969) , in the nearl y iso therm a l ice of th e upper 700 m (0- 3200 yea rs old; see Fig . 2 and Wood s, 1994).

(2) Cons tan t grain-size from 700 to 1678 m d epth (3200-II 600 years, with 11 600 years being th e end of the Younger Dryas co ld even t) .

12r-----.-----.------.-----.-----,------n

Regime 4 10

~ 8 § ~R~e~g~im=e~I_,--~R~eg~im~e=2--,_--~R~e~ym~· =e~3~--_1

" . ~

.~

" 500 1000 1500 2000 2500 3000

Depth (m)

Fig . 1. Jfeall/zori;::olllal illlmepl qj'graills l'S de/Jilt ill lite GISP2 ice core, Jrom fl'oods (1994) . T he JOllr gra ill­gr01L'ih regimes are described ill Ihe 1nl alld ill l, 'oods ( 1994 ), alld overlie Jine-grail1ed si/I)' ice al Ihe bed. l re plol meall illlerce/)I ra/her Iltcm cross-secliollal area to allow beller dis/Jh~J' of lite coarse graills Il ear the bed.

(3) Often sma ll e r g ra in-size bu t \Vi th \'a ri a ti ons th a t co rrelat e direc t! y wi th isotopi c com position (deposi­ti o n tempera ture ) a nd il1\ 'e rsel y with impurit y loadings in a nd befo re th e Yo unger Dryas e\·ent.

(4 ) \' ery la rge gra i ns in wa rm er ice nea I' th e bed (sce a lso Go,,' and oth ers, 1993; Thorste insson and o th ers, 1995; basa l temperatures reach a s hig h as 9a e nea r th e bed ; C uffe)' a nd o thcrs, 1995 ) .

Th ese four zones o f clean glac ier ice m 'erli e \"Cr \' fin e­gra ined , \'isibl y silly basa l ice . Th e bo und ary be twee n regim es I a nd 2 pro ba blv is gradu a l, a nd th a t bcrwee n zo nes 3 and 4 appea rs interfingered (sec a lso Thorste ins­son a nd o th ers, 1995 ) .

" 'oods ( 1994 ) interpre ted the la rge g ra in-sizes in hi s d ee pes t zo ne as represen tin g a nn ea ling associa ted wi th hig h basal temperatures (Go\\" a nd Willi amson, 1976; Bu clcl and J ac ka , 1989 ) , the ice-age zo ne as reOec ti ng som e impurit y or o th e r effec ts on g ra in-g rowth or g ra in­subdi\'ision ra tes (Lang" 'a y a nd o th ers. 1988; Pa terson , 199 1; Thorsteinsson a nd o th ers, 1995 ), the consta n t g ra in-sizes in th e o ld e r pa rt o f" th e Ho loee ne as representing gra in subdi"i sion by polygo niza ti on a t th e sam e ra te th a t g ra ins a re consum ed by normal g ra in g rowth (All ey and o th ers, 1995) a nd th e upper pa rt as representing norma l "'soap-bubble" g ra in growth. In th e d eeper zones, small g rai n -sizes may i nd ica te slow grai n g rowth or ra pid g ra in subdi\'ision, so wc res tri c t our a tt e nti on to th e upper fe\\' hundred m e ters \I·here g rain g rowth occurs with o ut g ra in subdi vision.

Fig ure 2 shows a m o re d eta il ed \·ie\\" of da ta ri"om thi s regio n of norm a l g ra in g ro \\·th (regim e I ). Noti ce from Fig ure I tha t gra in g ro,,·th continu es to a bout 700 m or 3200 years; by omitting fro m Fig ure 2 samples in th e old e r pa rt o f regime I , it mig h t a ppear th a t gr a i n growth stops so m ewha t ea rli er beca use o fOu ctu a ti o ns in the da ta . Th e tra nsition from regime 1 to regim e 2 does a ppear gradu a l, howe' T r, as polygo ni za ti o n becomes more importa nt with in creasing depth. \V e be li C\'e, based o n inspection of th e sampl es for strain sha d ows, etc., that our a nalyses in reg im e I a re confined to ice a bo\"C d epths of signifi ca nt po lygo ni za ti on.

AlIe.J1 alld ll 'oods: II7I/J lIri~) ' i17j711 e11ce Oil grain growlh

Depth (m) 100 200 300 400 500 600

35 I I I I I I

30 ~

~

'8' 0 0

0

25 § 8

120 I I §

: t i I '" ~ c

'Ol (5

15 -r

10

8

.... ~ i I +

i J o

o 500 1000 1500 2000 2500 3000 Age (years)

Fig. 2 . . \leall gra in rross-secliollal area vs age III Ihe 1l0rmaL-graill-growllt .::.olle (regime / ) of Ihe C'ISP2 core. The al'erage graill-si::.e , A,jor eaclt 8 10 cm 10llg lesl lille is show lI as a /)oilll ; aLl lesl lines ( rollgh(v 250 /)er sam/Jle) ill each a/J/Ho limalelj' 2)'ear Ihill seclioll are showlI allhe same age. Regressioll lilies are sho1.{'" '/or aLl 0./ Ihe dala, and Jo r Ihe slllal/esl alld l/i e /({Igesl a1'erage graill-sizes ill each 2)'ear illlm'al. The /l n /- dee/Jer 2}ear Lhill secliolls. al lite 10/) 0./ regime 2. are .Iomew/wl coarser Ihall Ihose sholt'lI here (see Fig. / ) .

Fig ure 2 sho,,'s th a t th e a \'e rage gra in-size (a rea ) a nd th e ra n ge o f" g rain-sizes in a sa mple in crease \\'ith increasing age, and thu s a lso \\"ith d epth. Each point represent s a test line o f ro ug hl y 10 cm o r 30 gra ins. \" a ri o us o th er plottin g co n\"C nti ons (fo r exa mple, a \'e[­aging all test lin es in 3 c m lengths a lo ng th e core, o r roughly 500 gra ins) yield a simila r result , so this is not a sampling artifa ct. \\'ithin ice ri'om a 2 yea r peri od , th e ra nge be t,,"Ce n the a\"(' rage g ra in-sizes in th e coa rses t­gra ined and fin es t-gra ined la ye rs increases \\'ith increas­ing d epth and age .

!\ clea r diffi culty in es tim a ting a \'erage g ra in-grO\\"lh rat e fro m a \ 'e rage grain-size in a la ve r is that one cannot tell \\'he th e r a gi\Tn laye r in a dee per sa mple bega n as a coa rse-g ra in ed , medium -g rained or fin e-g ra ined laye r nea r th e surface. The two yo unges t samp les shown in Fig ure 2 a rc from snow pits in the upper 2 m (upper 3 years; d e nsit y ~ 330 kg m 3) a nd from 39.5- 40 m d epth in th e core (d ensit y ~690kg m \ th ey thus brac ket th e end o f" ra pid firnifi ea ti o n a nd the d epth a t \\'hi ch th e " temperature mcmory" is acquired in th e Petit a nd o th ers ( 19 8 7, 198 8 ) mod e l (d ensity ~ - 5 0kg m 3,

ac hi e\'ed at ~ 1 5 m d epth ) . The awragf' g ra in-size in th ese yo ungest sampl es is sm a ll compa red to th e old er ones in Fig ure 2, and th e range of a \"C rage g rain-sizes in the younges t sa mpl es is narro w . \\'1" thus do not introduce grea t erro r by using th e intercept o f" th e regression line throug h a ll th e da ta in Fi g ure 2 as th e initi a l grailHize, All , in calcu la ting grain-g rowth ra tes fo r d ee per samples, a nd we h a \'e done so in subsequ ent calcula ti o ns.

Th e uncertaint), in g ro vvth rate introduced by this lack of kn owledge of sta rtin g g ra in-size can be assessecl by drawing a line from a ny point of interes t to th e top a nd to

th e bo ttom of' th e a rra y o f grain-sizes for th e two sha llowes t samples. This unce rtainty clea rly d ec reases as age , d e pth and g ra in-s ize in crease, and is nearl y

25 7

J oum al cif GLaciology

insig nifi cant fo r th e deep er samples in regime I, whieh is wh y we concentra ted sampling in th e d eep er part. N o la rge clim atic changes h ave occ urred thro ugh this tim e (e.g . G rootes a nd others, 1993; i\ r eese a nd o th ers, 1994), so th e sta rtin g gra in-size is unlikely to ha \'e cha nged sig nifi cantl y.

\ V e fo rm ed correla tion ta bles between grain-growth rate, sta ble-isoLO pic compos ition , concentra tions of solu­ble m aj or ions (Mg2 + , Ca2+ , Na + , K +, Cl , NH4 +,

N0 3 , SO l2 ), and LO ta l so luble ions (bo th by weight and by number or molec ul es) fo r those thin sec ti ons fo r \, .. hi ch we ha \"e complete chemi ca l a nd iso topic d a ta . Examples a re sh o\\'n in \\'oods (1994) a nd summ a ri zed in Figure 3 .

C

" ;g a; 8 c 0 "@ ~

" 0

Age (yr)

0 500 1000 1500 2000 2500 1

Some deposll ional Correlations primarily with grain growth effects on

correlations

Stable isotopes

0

t,-

" .6-- _ ______ _ __ __

Sum of impurities

: , to.

·1 0 tOO 200 300 400 500 600

Depth (m)

Fig. 3. Correlation coifficients oJ grain-growth. ra te 10 weighl oJ soLuble imjJurilies, and 10 slabLe-isotojJic comjJosition, jJLolled against depth Jar those sam/JLes Fom Figure 2 Jar which we have comjJLele chemical and isolo/Jic data. Shallow sam/Jles include some information from de/Josition as well as grain growth; deejJer samples have eA/JerielZced slifficient grain growlh 10 Lose 1110St oJ the depositional inJormation . T he /lositive correlatioll to isotopes in the shallow samjJles reflects the jormatioll of coarse grains in isoto/licalLy heaV)' summer snow during its jirst )lan; the relation to imjJurities is noi5)' at this age. T I"ith increasing age, the correlation to is%pic ratio disajJjJears but all inverse conelaLioll to im/Jurity loading develops, suggesting imjJU rif)' conlrol oJ growth rate .

Our m ai n obsen 'a tions a re:

( I ) Th e ra nge of aver age gra in-sizes, A, or layers millim eters to centim eters thi ck within a 2 yea r (50 cm ) thin sec ti on increases with increasing d epth , showing th a t g ra in-g rowth ra tes d iffer substanti a ll y be tween adj acent layc rs in th t" co re (d iffere nt J( des pite th e same in situ tempera ture) (Fig. 2) .

(2) l\'ega ti\"C correla ti ons betll'een grain-growth ra te a nd concentra ti on of a so luble impurity a re obtained more frequentl y tha n \\"Q uld be expec ted by cha nce, a lthough the res ul ts a re q uite " noisy" (Fig. 3; W ood s, 1994). Fo r example, to have a ll [our of the deepe r samples ex hibit a nega ti\'e correl a tio n between growth rate a nd impurity loading as observed wo uld occur on ly once in 16 tests if the

258

di stributions a re random, a nd o ther such nega ti ve corre la­ti ons with oth er measures of im p urity loading increase the p robability tha t non-random e ffects are present.

(3) Consistent co rrela ti ons b e tween grain-g ro wth rates a nd isotopic com positions ar e n o t obtained .

(4) " Fores t-fire layers" those centimeter-scale layers in which ammonium cOll centra ti o n is high, elec tri cal con­ducti vity is lo w, the ice appears cloud y or fa in t ly g reenish with sma ll bub bles , and in which ca rbon bl ac k h as been fo und when soug ht, sho" 'ing sig nifi cant fa ll out from forest­fire sm oke plum es - have ve ry slow g r a in g row th (Legra nd and others, 1992; Whitlow and o th ers, 1994; Chylek a nd o th ers, 1995; T aylo r a nd others, 1996; see Fig. 4). vVe have obseryed frequentl y the co rrela ti o n between elec tri cal-conducti\'ity lows a nd sma ll grain-sizes (see Woods, 1994, fi gs 5.3 and 5 .4, fo r example ), with suffi cient confirm a ti o n o f the conduc ti\'ity-l oll"(amm o nium-high rela ti on (T a ylo r a nd others, 1996) to be com ·in cing.

4.5 r-----r-----,----,----r------, 160

4 140

120

1 3.5 100

80 " . ~

.~ " 2.5

60

40 2

20

0 583 .6 583.7 583.8 583 .9 584

Depth (m)

Fig . 4. i\!Jeau grain in terce/)t agaillsL ammonium concentraLion across a 'forest·fire" spike il1 a lhill section jiom 583.5- 584 III dejJth ( ajJproximate(y 2600 )lairs age) . The correlation oJ Jine grains to high ammonium near 583.68 m dejJth is clear. These la)lers showing Lhe chemical signal oj fo rest-jireJallout are readiD' identijied b)l elee/ricaL condllctiviLy ( Tajlor and others, 1996), and we have See/I

this correlalion with fine grain-sizes il1 other layers.

DISCUSSION

~ :g ~

The a \'e rage g rain-grOll"th r a te obtained by regress ion throug h th e pl ot in Fig ure 2, for the GTSP2 site tem pera ture o f - 31°C, fa lls cl ose to the ac ti\"a ti on-energy plot for o th er si tes in Grre nl a nd a nd Anta rc ti ca when the da ta are reduced in a consistent way (Gow, 1969; Woods, 1994). This ind ica tes tha t o ur da ta fro m \"enica l thin sec tions yield th e sa me basic rela tions as data from other sites . as expec ted .

The d a ta in Figure 2 are fro m thin sec tio ns 40- 50 cm in leng th , a nd typica ll y covering 2 years each . Th e shall oll'es t d a ta incl ud e di sc re te samples coll ec ted over 2 m dep t h (3 yea rs) in two snow pits, with sampling designed to o bta in the mos t extreme gra in-sizes present. All of th e ice wi thin a sampl e has ex perienced the same tempera ture history after its first yea r or two, a nd the sa me stra in history (cumul ati\'e \'erti cal stra in in the upper 600 m of the ice shee t of only a bout 20 % ), yet the

grain-growth rates differ as shown in Figure 2. 'vV e a rgue th at this demo nstrates that something in th e ice a ffects grain growth (1< in Equation (1) depends on some fac tor or factors other than in situ temperature ).

Because all of th e ice in regim e I has experienced essen tially the same firnification tempera ture, our data cannot directly address the Petit and o thers (1987 , 1988) hypothes is of a diagenetic-temperature memory effect on 1<; our data would allow both impurities and diagenetic temperature to affect grain-growth rates. (We still believe that physical argum ents presented elsewhere do not a llow thi s, but th e reader must assess those arguments ind ep­endently.) We note that the lack of significant correlation of grain-growth ra te to iso topic composition allovvs us to exc lud e depositional temperature or true isotopic effects as strong control s on gra in-growth rate.

The negative correlations between gra in-growt h rates a nd impurity concentrations are quite noi sy a nd var iab le. This ma y be due partl y to sampling constraints; even the high-resolution chemical and isotopic sampling used here did not resolve all of the layering in the ice core . This is exacerbated by the difficulty uf fJrecisely registering thin sec tions to chemi cal samples . \\re do not belie\'e that thi s is sufIicient to explain the noisiness. however. The impur­ity-drag hyputhesis is \'e ry poorly quantified. As outlined by All ey and others (1986a, b), th e impurity-drag efIect changes by orders of magnitude for pla usible changes in impurity distributions among liquid or solid second phases and dissoh 'ed species (h olff and others, 1988). The state of impuriti es in th e ice is poorl y known , probabl y \ 'aries between layers and may \'ary over time (Wolff, 1996).

The fin e-grain ed " forest-fire" layers sugges t that \\'e can learn more abo ut the contro ls of grain g rowth. \Ve cannot tell , of course. what chemica l or particul ate species in th e forest-fire layers con trols the grain grQ\l'th , as several propertics seem to vary toge th er. The strong a mmonium sig na l, and the obsen'atio n that ammon ium has a sign ifi cant effect on ice in a variety of ways (e.g . Gross and others, 1978 ), lead us to suspect ammonium as an aCli\'e playe r. Kote that if thi s is true, th ere ma y be threshold effects or other spec ies invoked ; ammo nium \'a riations do not explain a ll gra in-size va riations, as shown in Figure 4 a nd in other data .

Mu ch furth e r work is need ed. The six data points III

Figure 4 , representing detailed iso topic and ch em ica l studi es a nd thin sec tions covering J 2 years and a cou nting exercise by Woods ( 1994) il1\ 'oh-ing approx im ate ly 50000 intersec tions between tes t lines and grains, are sti ll not sufficient to reveal a ll of the contro ls on grain growth. Other sites with g reater total g rain growth ill th e "soap­bubble" or normal regim e might give more-dcfiniti\'e rcsu I ts. Nonetheless, we concl ud e tha t somethi ng in the ice does a[)lxt g rain-growth rates, so that gra in-size is not a useful paJeo th e rmometer. Impurity drag seem s most likely, and is supported by correlation analysis, even at the very low concentrations of impurities in Holocenc ice from central Greenland.

ACKNOWLEDGEMENTS

W e thank K. Cu[Iey, B. Eld er, J. Fitzpa tri ck, A .J . GOII' , G. Jablunovsky, \IV. K apsner, D. M eese and G. Zielinski

A lLey and W oods: Impuril)l influence 01/ grain growtlt

for assistance in sampling, P. Mayewski and co-workers for unpublished chem ica l data , J. V/hite, B. Vaughn and co-workers for unpublished deuterium-iso topic data, P. Grootes and co-workers for unpubli sh ed oxygen-isotopic data, A .J . Gow, D. lV[eese and two a non ymous reviewers for helpful suggestions, the GISP2 Science 'Management Office, the Polar I ce Coring Office, th e 109th New York Air N a tional Gua rd , the National I ce Core Lab a nd assorted GISP2 coll eagues for assistance and logistical support, and th e National Science Foundation Office of Polar Programs for fi nancial support; some funding a lso was provided by the D. and L. Packarcl Foundation and NASA-EOS.

REFERENCES

Alley, R. B. 1987a. T ex lLlre of polar Grn for remo tc sensing . .. 11111 . Glacial., 9, 1 +.

All ey , R . B. 1987b. Transfo rmations in po lar firn. (Ph.D. th es is. U nive rsit\· of Wisconsin ',,[ad ison. )

AIIc\·. R . B. and B. R . K oei. 1990. R ece nt wa rming in ce nrra l C recn la ncJ:' . filII. Glacial., 14. 6 8.

All e)', R. B. , J. H. Percpcz ko and C. R. Bentley. 1986a. Grain growth in po la r ice I Theory.]' (;Iao ol .. 32 (112 ), 41 5-A·24.

Alley, R . IL J 1-1. Percpez ko and C. R. Bentley. 1986b. Grain gro\\·th in polar ice: Il. Appli C<ll ion.]' ell/ciol .. 32 ( 11 2) .425-433.

Alley, R. B. ,JI-I. Perepez ko and C. R. Bentle y. 1988. Long-term climate c h" ngcs from crvsta l g row th . . \fIlII re. 332 ( 6 165 ). 592- 593.

All cy, R. B. alld 10 olllen . 1993. Abrupt increase in Greenland sno\\' a cc umul a tion a t the end of the Yo unge r Dryas e\-en t . . \ 'all/re , 362(6420). 527- 529.

Alley. R. B. , A.J. Go\\' a nd D. A. \[eese. 1995. \Iapping c-"xis bbrics lO

stud y pllysiral processes in ice.]. Glacial .. 41 ( 137 ), 197- 203 . . \n anc};)krishnan. S .. .J .J. Fitzpat ri ek, R. B. Alle \ ', A.J Go\\' and D.I\.

l\[ eese . 199+. Shea r-wan' d etec ti on ofas\'llllllctri c c-axis fabrics in th e G ISP2 ice ro re. Greenland.]. (;Iociol .. 40 ( 136 ). 491 - 496.

Bu cltl. \\ ' . F. and 1'.1-I . .J acka. 1989. A IT\·iew a fi re rheo logy for ice sheet l11odC'iling. Cold Reg. ,)'ci. Teellllol., 1612 1. 107 I H.

Ch5' lek. 1' .. B. J ohnson. P. :\. Damiano. K. C. Ta ylor and P. Clement. 1995. Bioma" burning reco rd and black carbon in the G ISP2 ice core. Geo/l/ll'J. Res. Lell ., 22 (21,89 92 .

cum·)" K. \1. , G. D. C law, R. B. Alley, i\1. Slui\'(: r , E. D. \\ 'adding to n and R . \\ '. SallUs. 1995. Large Arct ic tcm pcralurc cha nge a t the \\ ' isconsin Holorenc g lac ial transiti on. Science, 270 (5235 ). 4-55-458.

Du \'a l, P. a nd C. Lorius. 1980. Crystal size and climatic record d O\\'Il to the last ice age from Allla rctic ice. El/rlh Plallel. Sci. Lell .. 48 ( I ) , 59-6+.

GO\\·. A. J . 1969. On the rat es of growth of gra ins and cr\'s tals in South Pob r rirn.]' (;Iaoo!., 8 (53 ). 2-l 1 252 .

Gow, A . .J. and T. \\' illi a m son. 1976. Rh eo logica l implica tions or the interna l structure and c rysta l fab ri cs of th e \ Vest Antarcti c ice sheet as rC\Talcd bl' deep co re drilling al Byrd S tatio n. Geo!. Soc. Alii. B lIll., 87 ( I 2 ), 1665- 1677.

Gow, :\.J. , D. A. i\Ieese, R . B. Alley a nd J.J . FitzPatrick. 1993. Crysta llinc structure a nd c-axis fabrics o f' th e GISP2 core fro m surface to bed roc k. EOS, 74 (43 ), Supplement, 89.

GroOles, P. \1. . i\l. Stui\·er. .1 . \\ ' . C. \\'hite, S. Johnsen a nd J . J ouzel. 1993. Comparison of' oxygen iso tope reco rd s fi-olll the G I S 1'2 a nd GRI I' G rCl'n lanci ice cores .. \ ill",.,'. 366 (6+55 ), 552- 55+.

G ras." G. \\ ' .. l. c:. Hayslip a nd R. i\. 1-1 0). 1978 . Elec tri ca l conducti\·it\· a nd relaxa tion in ice crysta ls with kn C)\\"Il impurity Cont e nt. J. Glaciol., 21 (85 ), 14 3 160.

J ouze l. J . alld 6 olhen . 1ge7. \ '"stok ice co re: a continu ous iso tope temperature record o \ 'C r the las t climatic cycle (160.000 years ). " \ (lllm , 329 (6 138 ). 'f03 - +08.

La ngway, C. C., Jr, H . S hoji and ,\I, ,\ zuma. 1988. Crysta l size a nd oric lllation pallerns in th e \\ 'isconsin-agc ice from Dye 3, Gree nl a nd . . "fllll. Glacial., 10. 109- 11 5.

Legrand, i\l., i\I. De Ange li s, 1'. SlafTelbac h , A. :\Jeft cl and B. SwufTer. 1992. Large perturbations of ammonium a nd o rga ni c acids conte nt in the SUlllmit Gree nl a nd ice co re: fin ge rprint from forest fires' Gro/JI,)ls . Rrs. l,ell .. 19 (5 ) .473--475.

259

J ournal oj Glaciology

\faye\\'s ki , P. A. alld 7olhers. 1993. The atmosphere during the Younger Dryas . Scifllce, 261 51 18 . 195 197.

t--I ee,e, D. A. and 8 olhers. 199+. The acc umulation record from the G ISP2 core as an indicato r of climate change throughout the Holocene. Sciena, 266 ' 519 1), 1680 1682.

Paterson. \\ ' . S. B. 1991. \\,h y ice-age ice is sometimes ' ·soft". Cold Reg. Sri . Tahllol .. 20( I ), 75 98.

Pct it , .1. R .. P. Du"al and C. Lo rius. 1987. Long-term climatic changes indicated by trysta l growth in po la r ice. ,\ 'aiure, 326 (6108). 62- 64·.

Pctit, J. R .. P. Duya l and C. Lorius. 1988, Scientific correspondence. Long­term cli ma te changes from crys ta l growth. ,\ 'alllre, 332 (6165 ). 593.

T ay lor, K. C., P. A . \fa\T\\'ski, '\1. S. T\\'icklcr and S. 1. \\' hillow. 1996. Biomass burning recorded in the G IS P2 ice core: a record from eas tern Canada" Holocelle, 6 ( I ) , I 6.

Thorsteinsson. Th .. J. KipfslUhl , 1-1. Eicken. .J. J o hnsen and K. Fuhrer. 1995. Crystal , iz( \ 'a ri a ti o ns in Eemian-agc ice from the

G RIP ice core, ce ntra l Greenland. Earlh Plallfl. Sri. Lell .. 131 (3 +1, 381 - 39+.

U nderwood . E. E. 1970. QJlalllilalh'e slfreolog)'. R eading, t--IA , Addison­\ \ ' eslel' Publishing Co.

Whi tlow, S .. P. r.laye\\·ski. J. Dibb, G. H oldswort h and r.L T wicklcr. 1994. An ice-core-based record of biomass burning in the Arc ti c ancl subarClic, 1750- 1980. Tellus, 468 (3), 234- 242.

\\ 'olff, E. W. 1996 . L ocation, mo\'CmCI1[ a nd react ions of' impuriti es in so li d ice. III \\' olff, E. V" . and R. Bales, eds. Chemical "challge bellt'eelllhe almosphere alld polar '<110/1.'. Berlin , etc., Springer- \ 'crlag. (l\ATO AS I Series. I +3 ),541 -565.

\\'olff. E. \\' .. R . r. luh'a ney and K , Oates. 1988. The location of impurities in Antarctic ice. ,l llll. Glaciol., 11. 194-197.

\\'ood s, G. A. 199+. Graill groldh beha";or of t/ie GfSP2 ice [ore frolll celllral Greelllalld. Cni\'ersity Pa rk , PA , PcnllSyh'ania Sta te Cni\ 'C rsity. Eanh System Science Center. (T echnica l Report 94-002.

MS received 11 Seplember /995 and accepled in revised JOI111 28 D ecember /995

260