Embed Size (px)
Citation preview
226Ra^230Th^238U disequilibria of historical Kilauea lavas(1790^1982) and the dynamics of mantle melting within the
Hawaiian plume
Aaron J. Pietruszka *, Kenneth H. Rubin, Michael O. GarciaHawaii Center for Volcanology, Department of Geology and Geophysics, University of Hawaii, Honolulu, HI 96822, USA
Received 13 January 2000; accepted 3 January 2001
Abstract
The geochemical variations of Kilauea's historical summit lavas (1790^1982) document a rapid fluctuation in themantle source and melting history of this volcano. These lavas span nearly the entire known range of sourcecomposition for Kilauea in only 200 yr and record a factor of V2 change in the degree of partial melting. In this study,we use high-precision measurements of the U-series isotope abundances of Kilauea's historical summit lavas and two`ingrowth' models (dynamic and equilibrium percolation melting) to focus on the process of melt generation at thisvolcano. Our results show that the 226Ra^230Th^238U disequilibria of these lavas have remained relatively small andconstant with V12 þ 4% excess 226Ra and V2.5 þ 1.6% excess 230Th (both are þ 2c). Model calculations based mostlyon subtle variations in the 230Th^238U disequilibria suggest that lavas from the 19th to early 20th centuries formed atsignificantly higher rates of mantle melting and upwelling (up to a factor of V10) compared to lavas from 1790 and thelate 20th century. The shift to higher values for these parameters correlates with a short-term decrease in the size of themelting region sampled by the volcano, which is consistent with fluid dynamical models that predict an exponentialincrease in the upwelling rate (and, thus, the melting rate) towards the core of the Hawaiian plume. The Pb, Sr, and Ndisotope ratios of lavas derived from the smallest source volumes correspond to the `Kilauea' end member of Hawaiianvolcanoes, whereas lavas derived from the largest source volumes overlap isotopically with recent Loihi tholeiiticbasalts. This behavior probably arises from the more effective blending of small-scale source heterogeneities as themelting region sampled by Kilauea increases in size. The source that was preferentially tapped during the early 20thcentury (when the melt fractions were lowest) is more chemically and isotopically depleted than the source of the early19th and late 20th century lavas (which formed by the highest melt fractions). This inverse relationship between themagnitude of source depletion and melt fraction suggests that source fertility (i.e. lithology) controls the degree ofpartial melting at Kilauea. Thus, rapid changes in the size of the melting region sampled by the volcano (in the presenceof these small-scale heterogeneities) may regulate most of the source- and melting-related geochemical variationsobserved at Kilauea over time scales of decades to centuries. ß 2001 Elsevier Science B.V. All rights reserved.
Keywords: Kilauea; volcanoes; Hawaii; lavas; geochemistry; uranium; isotopes; mantle plumes; melting; models
0012-821X / 01 / $ ^ see front matter ß 2001 Elsevier Science B.V. All rights reserved.PII: S 0 0 1 2 - 8 2 1 X ( 0 1 ) 0 0 2 3 0 - 8
* Corresponding author. Present address: Carnegie Institution of Washington, Department of Terrestrial Magnetism, 5241 BroadBranch Road, N.W., Washington, DC 20015, USA. Tel. : +1-202-478-8476; E-mail : [email protected]
EPSL 5747 22-2-01
Earth and Planetary Science Letters 186 (2001) 15^31
www.elsevier.com/locate/epsl
1. Introduction
Historical lavas from active Hawaiian shieldvolcanoes such as Kilauea and Mauna Loa areideal for investigating mantle melting becausethey display systematic melting-related variationsof incompatible trace element ratios (e.g. La/Ybor Nb/Y) over short periods of time (years todecades [1^4]). Models based on these £uctuationsin lava chemistry suggest that the degree of partialmelting has varied by a factor of V2 for Kilaueasince 1790 [3] and V40% relative for Mauna Loasince 1843 [2], which are signi¢cant portions ofthe overall range inferred for Hawaiian shield vol-canoes (a factor of V2.5 [5]). The origin of theseshort-term changes in the degree of partial melt-ing is poorly understood, but may be related tothe interplay between mantle melting and small-scale source heterogeneities [2,3]. A similar pat-tern of rapid, source- and melting-related geo-chemical variations has also been observed instratigraphic sections of prehistorical shield lavasfrom Hawaiian volcanoes (e.g. Mauna Loa [6]and Mauna Kea [7,8]). Thus, a detailed geochem-ical sampling of lavas over time scales of decadesto centuries may provide the best clues to under-standing the dynamics of mantle melting withinthe Hawaiian plume.
The U-series isotope abundances of lavas (e.g.230Th and 226Ra) are excellent tracers of mantlemelting because the half-lives of 230Th and 226Ra(V75 kyr and 1600 yr, respectively) are thoughtto be similar to the time scale of magmatic pro-cesses (e.g. [9]). This unique feature of U-seriesisotopes o¡ers the potential to study both the na-ture and timing of the processes that fractionateincompatible trace elements during magma gene-sis. This possibility derives from two importantproperties of the 238U decay chain. First, priorto melting, the source regions of basaltic lavaswill generally be in a state of radioactive, or sec-ular, equilibrium in which the 238U, 230Th, and226Ra activities (decay rates) are equal. When sec-ular equilibrium prevails, the (230Th/238U) and(226Ra/230Th) ratios (parentheses indicate activ-ities) of the source are known (and equal tounity), in contrast to most other geochemical trac-ers of the melting process (e.g. stable, incompat-
ible trace elements). Second, any process thatfractionates these elements, such as partial melt-ing of the mantle, induces a state of radioactivedisequilibrium in which the activities of 238U,
Fig. 1. Summary of the melting, source, and eruptive historyof Kilauea Volcano. Kilauea's historical lavas record a rapid£uctuation in the degree of partial melting (inferred from in-compatible trace elements) that correlates with the volcano'seruption rate (and, presumably, its magma supply rate). Thecomposition of the mantle source region tapped by Kilauea(or proportions of the source components) has also changedover time (shown as a % of the early 20th century sourcecomposition, which corresponds to the `Kilauea' isotopic endmember for Hawaiian volcanoes). The vertical shaded linesmark the explosive summit eruptions of 1790 and 1924. See[3] for the methods and/or references used to calculate eachof these parameters. The average melt fraction and sourcecomposition for the Mauna Ulu rift zone lavas are estimatedusing data from [1] and unpublished Pb isotope analyses (A.Hofmann, personal communication, 1999). The sample sym-bols are grouped according to eruption date: up triangles(1790), circles (1820^1921), diamonds (1929^1954), squares(1961^1982), left triangles (Mauna Ulu), and right triangles(Puu Oo).
EPSL 5747 22-2-01
A.J. Pietruszka et al. / Earth and Planetary Science Letters 186 (2001) 15^3116
Tab
le1
Th,
U,
Ba,
Sran
d22
6R
aab
unda
nces
,U
-ser
ies
isot
ope
acti
vity
rati
os,
and
Sris
otop
era
tios
for
the
hist
oric
alsu
mm
itla
vas
ofK
ilaue
aV
olca
no
Sam
ple
Eru
ptio
nda
te[T
h][U
][B
a][S
r][22
6R
a](23
0T
h/23
2T
h)(23
0T
h/23
8U
)(22
6R
a/23
0T
h)0
(234U
/238U
)87
Sr/86
Sr(W
gg3
1)
(Wg
g31)
(Wg
g31)
(Wg
g31)
(fg
g31)
1790
-117
900.
6960
0.23
4479
.67
278.
692
.0þ
0.7
1.04
9þ
0.01
11.
022
1.14
71.
003
þ0.
004
0.70
3612
91.4
þ0.
61.
040
þ0.
011
1820
-118
20^1
823
0.78
120.
2641
98.3
431
0.6
98.6
þ0.
61.
041
þ0.
015
1.01
41.
103
1.00
3þ
0.00
30.
7036
3899
.4þ
0.8
1.03
9þ
0.01
118
6618
661.
156
0.39
6914
1.7
382.
40.
7035
4818
94-2
(gro
undm
ass)
1894
1.13
30.
3931
144.
238
5.1
162.
7þ
1.0
1.06
7þ
0.01
61.
020
1.21
21.
004
þ0.
004
0.70
3547
1.08
2þ
0.00
918
94-2
(gla
ss)
1894
1.20
10.
4135
146.
038
2.6
0.70
3548
95-T
AJ-
319
121.
254
0.43
9914
4.3
398.
20.
7035
3619
1719
171.
206
0.42
4313
2.5
397.
716
2.8
þ1.
81.
093
þ0.
012
1.02
41.
114
1.01
3þ
0.00
30.
7034
91K
il191
919
191.
225
0.42
7913
5.0
401.
516
7.4
þ1.
31.
085
þ0.
011
1.02
31.
129
1.00
3þ
0.00
20.
7034
7116
5.0
þ0.
819
2919
291.
306
0.42
8914
6.5
392.
517
1.3
þ2.
51.
035
þ0.
010
1.04
21.
131
1.00
5þ
0.00
30.
7035
5116
8.7
þ3.
31.
042
þ0.
013
1931
HM
1931
^193
21.
336
0.43
9315
0.1
400.
00.
7035
4619
54-2
1954
1.28
70.
4284
150.
440
0.9
0.70
3607
K61
-22
1961
1.23
20.
4141
145.
838
8.7
155.
3þ
1.1
1.04
8þ
0.00
91.
029
1.08
11.
006
þ0.
003
0.70
3584
1.05
1þ
0.01
519
71S
Sep
1971
1.08
70.
3633
132.
038
6.4
0.70
3580
1982
A-2
0A
pr19
820.
9832
0.32
9712
1.3
360.
112
6.2
þ0.
71.
052
þ0.
014
1.02
91.
103
0.70
3566
1.04
2þ
0.00
9
All
unce
rtai
ntie
sar
eþ
2c.
The
repr
oduc
ibili
tyof
the
Th,
U,
Ba,
and
Srco
ncen
trat
ions
(and
rati
osth
ereo
f)is
þ0.
3%or
bett
er.
The
unce
rtai
ntie
sfo
rth
e22
6R
aco
ncen
trat
ions
and
the
(230T
h/23
2T
h)an
d(23
4U
/238U
)ra
tios
liste
din
the
tabl
ear
eth
ew
ithi
n-ru
nm
easu
rem
ent
erro
rs.
The
(230T
h/23
8U
)an
dth
eag
e-co
rrec
ted
(226R
a/23
0T
h)ra
tios
are
calc
ulat
edus
ing
the
aver
age
226R
aab
unda
nces
and
(230T
h/23
2T
h)ra
tios
for
each
sam
ple.
The
accu
racy
and
repr
oduc
ibili
tyof
our
Th
isot
ope
rati
om
easu
rem
ents
wer
eve
ri¢e
dw
ith
the
Uni
vers
ity
ofC
alif
orni
aSa
nta
Cru
z(U
CSC
)T
hst
anda
rd`A
'.O
urav
erag
efo
rU
CSC
Th
`A'
was
(230T
h/23
2T
h)=
1.08
6þ
0.01
2(n
=39
),w
hich
is0.
3%lo
wer
than
the
pred
icte
dva
lue
(1.0
89).
Rep
licat
ean
alys
esof
the
Nat
iona
lIn
stit
ute
ofSc
ienc
ean
dT
echn
olog
y(N
IST
)U
stan
dard
SRM
U01
0ga
ve(23
4U
/238U
)=0.
9993
þ0.
0054
(n=
10),
whi
chis
0.04
%hi
gher
than
the
cert
i¢ca
teva
lue
(0.9
989)
.R
eplic
ate
anal
yses
ofth
eN
IST
Srst
an-
dard
SRM
987
duri
ngtw
ope
riod
sof
tim
ese
para
ted
bya
maj
orm
achi
neup
grad
ega
ve87
Sr/86
Sr=
0.71
0259
þ0.
0000
12(n
=10
)an
d87
Sr/86
Sr=
0.71
0268
þ0.
0000
14(n
=14
).A
llda
taar
eex
pres
sed
rela
tive
to87
Sr/86
Sr=
0.71
0259
for
SRM
987.
The
wit
hin-
run
unce
rtai
ntie
son
indi
vidu
al87
Sr/86
Srm
easu
rem
ents
wer
eal
way
seq
ual
toor
less
than
the
exte
rnal
repr
oduc
ibili
tyof
SRM
987.
The
tota
lpr
oced
ural
blan
ksfo
rab
unda
nce
mea
sure
men
tsw
ere:
Th
(1^6
pg),
U(2
^7pg
),B
a(0
.2^1
ng),
Sr(2
0^60
pg),
and
226R
a(6
0.01
fg).
Bla
nkco
rrec
tion
sof
0.00
3^0.
04%
for
Th,
0.02
^0.1
%fo
rU
,an
d0.
005^
0.07
%fo
rB
aw
ere
mad
eto
deri
veth
e¢n
alco
ncen
tra-
tion
s(S
ran
d22
6R
abl
anks
wer
ene
glig
ible
).T
heto
tal
proc
edur
albl
anks
for
isot
ope
rati
om
easu
rem
ents
wer
ene
glig
ible
:T
h(1
0^20
pg)
and
U(3
^4pg
).T
hede
cay
cons
tant
sus
edar
e:V23
8U
(1.5
51U
103
10yr
31),V23
4U
(2.8
35U
103
6yr
31),V23
2T
h(4
.948U
103
11yr
31),V23
0T
h(9
.195U
103
6yr
31),
andV22
6R
a(4
.332U
103
4yr
31).
EPSL 5747 22-2-01
A.J. Pietruszka et al. / Earth and Planetary Science Letters 186 (2001) 15^31 17
230Th, and 226Ra are unequal. The magnitude ofthe 226Ra^230Th^238U disequilibria in basalticlavas is believed to depend, in part, upon the ratesof mantle melting and upwelling [10^12].
Early U-series studies of Hawaiian shield lavasused radioactive decay counting techniques toidentify small, but signi¢cant, 230Th and 226Raexcesses relative to their respective parents, 238Uand 230Th, and proposed a magmatic (mantlemelting) origin for these disequilibria [13^15].More recent, high-precision decay counting andmass spectrometric U-series investigations of his-torical Kilauea and Mauna Loa lavas observedsmall 230Th excesses and moderate 226Ra excessesup to 28% [16^20]. Modeling of the small 226Ra^230Th^238U disequilibria of these lavas (comparedto mid-ocean ridge basalts [21,22] and lavas fromother ocean island volcanoes [23]) has led to theidea that the rates of mantle melting and upwell-ing for tholeiitic basalt production within the Ha-waiian plume are relatively high [16^19,23].
This paper presents new U-series isotope, Srisotope, and trace element abundance data forthe historical summit lavas of Kilauea Volcano(1790^1982). These samples span nearly the entireknown range of source composition for this vol-cano and record a factor of V2 change in thedegree of partial melting [3]. In this study, wefocus on mantle melting at Kilauea using the U-series isotope systematics of its historical lavasand two `ingrowth' models for producing 226Ra^230Th^238U disequilibria (dynamic melting [10]and equilibrium percolation melting [12]). Ourgoal is to ¢nd an internally consistent modelthat accounts for the 226Ra^230Th^238U disequilib-ria of Kilauea lavas in the context of other geo-chemical evidence (from incompatible trace ele-ments and Pb, Sr, and Nd isotope ratios [3]) fora short-term change in the mantle source andmelting history of this volcano (Fig. 1).
2. Results
High-precision measurements of the Ba, Th, U,Sr, and 226Ra concentrations and (230Th/232Th),(230Th/238U), (226Ra/230Th), (234U/238U), and 87Sr/86Sr ratios of Kilauea's historical summit lavas by
thermal ionization mass spectrometry (TIMS) arereported in Table 1 (for additional sample infor-mation and analytical methods see the EPSL On-line Background Dataset1). In this section, we de-scribe the temporal geochemical variations of thelavas (Fig. 2). All uncertainties are þ 2c.
2.1. 230Th^238U disequilibria
The 230Th^238U disequilibria of Kilauea lavashave remained nearly constant since 1790(V2.5 þ 1.6%, 2c) with a narrow 1.4^4.2% rangein the amount of excess 230Th (Fig. 2). The 230Th^238U disequilibria from other recent studies of Ki-lauea lavas (except [19,20]) are signi¢cantly morevariable than our results and include some sam-ples with (230Th/238U)6 1 [16,17]. Despite thesmall (230Th/238U) range observed in this study,the samples display a systematic temporal varia-tion. The (230Th/238U) ratios increase subtly fromthe 19th to early 20th centuries (V1%), increaseabruptly after the explosive summit eruption of1924 (V2%), decrease slightly thereafter (V1%),and remain essentially constant during the late20th century. These short-term changes in the230Th^238U disequilibria (V1^2%) are close toour average analytical uncertainty ( þ 1.1%) basedon repeated Th isotope standard measurements( þ 1.1%) and our þ 0.3% reproducibility of Th/U ratios. However, we believe these subtle di¡er-ences are signi¢cant because (1) they were veri¢edby duplicate Th isotope ratio measurements formost samples, (2) lavas that erupted during re-stricted periods of time have extremely constantTh/U, (230Th/232Th), and (230Th/238U) ratios (the1917 vs. 1919 lavas or the late 20th century lavas;Fig. 2), and (3) duplicate (230Th/232Th) analyses(and recalculated (230Th/238U) ratios) of severalsamples (1820-1, Kil1919, and six Mauna Uluand Puu Oo lavas; A. Pietruszka, unpublisheddata) using new, high-precision plasma ionizationmulti-collector mass spectrometric (PIMMS) tech-niques [24] agree within V0.3% of our TIMS data(Fig. 2). A systematic £uctuation in excess 230Th
1 http://www.elsevier.nl/locate/epsl; mirrorsite: http://www.elsevier.com/locate/epsl
EPSL 5747 22-2-01
A.J. Pietruszka et al. / Earth and Planetary Science Letters 186 (2001) 15^3118
has not been previously observed in lavas from aHawaiian shield volcano.
2.2. 226Ra^230Th disequilibria
The 226Ra^230Th disequilibria of Kilauea lavas
have also remained relatively constant over thelast 200 yr with an 8^15% variation in the amountof excess 226Ra (excluding the 1894 groundmass,which is anomalous; Fig. 2). Compared to otherrecent studies of Kilauea lavas [16,17], this is asmaller average 226Ra excess and (226Ra/230Th)
Fig. 2. Temporal variations of U-series isotope and incompatible trace element ratios in Kilauea's historical lavas. The samplesymbols are grouped according to eruption date as described in Fig. 1. For comparison with our TIMS data (larger black sym-bols), duplicate (230Th/232Th) analyses (and recalculated (230Th/238U) ratios) for several samples using PIMMS (A. Pietruszka, un-published data) are also shown (smaller black symbols). The black Mauna Ulu and Puu Oo symbols show average values forthese rift zone eruptions of Kilauea, which display a relatively narrow range in composition (V0.3 and 0.8% variation in theamount of excess 230Th, respectively; A. Pietruszka, unpublished data). The other symbols are recent Kilauea literature data col-lected by TIMS (the gray squares and left triangle are from [16] and the gray, right triangles are from [19,20]) or a combinationof TIMS and radioactive decay counting techniques (white symbols; [17]). Our average TIMS þ 2c error bars are shown on eachplot (unless they are smaller than the size of the larger black symbols). The PIMMS (230Th/232Th) and (230Th/238U) ratios haveþ 2c analytical uncertainties of V0.2% and 0.3%, respectively (less than or equal to the size of the smaller black symbols).
EPSL 5747 22-2-01
A.J. Pietruszka et al. / Earth and Planetary Science Letters 186 (2001) 15^31 19
range (the data of [19] agree with our results ofV12 þ 4%, 2c). Although the V7% range in226Ra^230Th disequilibria observed in this studyis signi¢cantly greater than our estimate of ana-lytical uncertainty ( þ 1.4%), the samples do notdisplay any systematic temporal trends.
2.3. Th/U, Ba/Th, and Ba/U ratios
Kilauea's historical lavas display signi¢cant andsystematic temporal variations in their highly in-compatible trace element ratios (Fig. 2). The Ba/Th (not shown) and Ba/U ratios decrease fromthe 19th to early 20th centuries, shift to highervalues after the 1924 eruption, and increase dur-ing the rest of the 20th century. The Th/U ratiosdecrease from the 19th to the early 20th centuries,shift suddenly (V6%) after the 1924 eruptionfrom the lowest to the highest values observedin the present study, decrease until the mid-20thcentury, and remain essentially constant there-after. The Th/U ratios of historical Kilauea lavaswere previously thought to be nearly constant[25].
2.4. (230Th/232Th) and (226Ra)/Ba ratios
Kilauea's historical lavas show small, but sig-ni¢cant temporal changes in their Th isotope and(226Ra)/Ba ratios (Fig. 2), which have not beenpreviously recognized for lavas from a Hawaiianshield volcano. The (230Th/232Th) ratios increasefrom the 19th to early 20th centuries (V5%), shiftabruptly to lower values after the 1924 eruption(V5%), increase slightly towards the mid-20thcentury (V1%), and remain constant during thelate 20th century. The overall (230Th/232Th) and(226Ra)/Ba variations mirror (inversely) the tem-poral changes in the Th/U and Ba/U (or Ba/Th)ratios of the lavas, respectively.
2.5. (234U/238U) ratios
All of the samples have (234U/238U) ratios thatare slightly greater than unity (1.005 on average;Table 1). We do not regard this small deviationfrom secular equilibrium (not expected for Ki-lauea's fresh, historical lavas) as signi¢cant be-
cause the (234U/238U) variation is only slightlygreater than our analytical uncertainty ( þ 0.7vs. þ 0.5%, respectively) and there is no correla-tion between the 230Th and apparent 234U ex-cesses. Furthermore, we note that the (234U/238U) ratios would be V0.3% lower if we usethe new half-life proposed for 234U [26] insteadof the commonly used value in Table 1.
3. Evaluation of post-melting e¡ects on the226Ra^230Th^238U disequilibria
The main purpose of this study is to use the226Ra^230Th^238U disequilibria of Kilauea's his-torical lavas to investigate mantle melting withinthe Hawaiian plume. However, processes such asradioactive decay or crustal assimilation may havemodi¢ed the lavas' mantle-derived (226Ra/230Th)and (230Th/238U) ratios. In this section, we brie£yevaluate these possibilities.
3.1. The time scale of magma transport andstorage
The compositions of Kilauea's historical sum-mit lavas are controlled mainly by the addition orremoval of olivine [27]. However, some of thelate 19th^early 20th century lavas have slightlyevolved compositions (down to 6.5 wt% MgOfor the samples analyzed in this study; M. Garcia,unpublished data) and may also have fractionatedsmall amounts of clinopyroxene and plagioclase.Neither olivine control nor minor clinopyroxeneor plagioclase fractionation will directly a¡ect the226Ra^230Th^238U disequilibria of the lavas (be-cause Ra, Th, and U are thought to be highlyincompatible in these minerals [28^34]), but radio-active decay during crystallization may have beenimportant.
The 226Ra excesses of the lavas indicate that the(230Th/238U) ratios have not decreased signi¢-cantly during magma transport and storage. Sincethe residence time of magma in Kilauea's summitmagma reservoir is believed to be much shorterthan the half-life of 226Ra (6 80^200 yr during theearly 19th to early 20th centuries and 30^40 yrduring the late 20th century [35]), the time scale
EPSL 5747 22-2-01
A.J. Pietruszka et al. / Earth and Planetary Science Letters 186 (2001) 15^3120
of magma storage was probably too short tocause a measurable decrease in the (226Ra/230Th)ratios. In contrast, the total duration of magmatransport is poorly constrained. Although thetransit time from the source to the crust mustbe signi¢cantly less than V8000 yr in order topreserve the observed, mantle-derived 226Ra^230Th disequilibria [21], the (226Ra/230Th) ratiosof the lavas must strictly represent minimum pa-rental magma values.
3.2. Crustal assimilation
One sample, 1894-2, is geochemically anoma-lous because its microcrystalline groundmass hasan elevated (226Ra/230Th) ratio compared to theother historical Kilauea lavas (21% excess 226Ravs. 12% on average), and much higher Ba/Th(4.7%) and Ba/U (3.9%) and lower Th/Sr (6.3%)and U/Sr (5.6%) than its coexisting glass (Fig. 2and Table 1). Assuming DTh = DU = 0 andDSr = 1.8 for plagioclase (calculated from [36]with a 1150³C magmatic temperature and a pla-gioclase composition of An75), V5.3 wt% plagio-clase accumulation accounts for the di¡erences inthe Th, U, and Sr concentrations between thegroundmass and glass (the lava contains V5.5vol% plagioclase-dominated glomerocrysts). How-ever, an unreasonably high plagioclase DBa s 1.2(assuming DTh = DU = DBa = DSr = 0 in olivine,DTh = DU = DBa = 0 and DSr = 0.1^0.2 in clinopyr-oxene [29,30]) would be required to explain the Basignature of the 1894 groundmass by plagioclaseaccumulation (cf. DBa = 0.2 estimated for plagio-clase from [36] as described above). Instead, therelatively high (226Ra/230Th), Ba/Th, and Ba/U ra-tios of the 1894 groundmass (beyond the minore¡ect of plagioclase accumulation) probably resultfrom crustal contamination.
The exact mechanism of this contamination atKilauea is unknown, but assimilation of hydro-thermally altered country rock containing youngbarite [17] or zeolites [37] are two possibilities.Although either scenario might elevate the(226Ra/230Th), Ba/Th, and Ba/U ratios in the mag-ma, secondary zeolites are signi¢cantly more com-mon than barite in lavas drilled from Kilauea'sactive hydrothermal system (D. Thomas, personal
communication, 1999). Since the 226Ra excess ofthe 1894 groundmass may be anomalous due tocrustal contamination, we do not use its (226Ra/230Th) ratio in our discussion of mantle melting.The overall 226Ra^230Th disequilibria of the otherKilauea lavas erupted since 1790 are believed tore£ect mantle processes, although we cannot ruleout the possibility that the small variations in theamount of excess 226Ra (Fig. 2) result from crus-tal assimilation. However, this will not signi¢-cantly a¡ect our model results because the meltingrates calculated from the (226Ra/230Th) and(230Th/238U) ratios of the lavas are insensitive tosubtle di¡erences in the amount of excess 226Ra(Fig. 3).
4. Modeling of mantle melting within the Hawaiianplume
`Instantaneous', or time-independent, meltingmodels (e.g. batch melting [38]) have been in-voked previously to explain the incompatibletrace element variations of Hawaiian shield lavas[1^4]. Although such models may account for thesubtle 230Th excesses of historical Kilauea andMauna Loa lavas (at melt fractions of V3^7%[20]), they fail to reproduce the combined 226Ra^230Th^238U disequilibria [17]. Instead, a class oftime-dependent models that allow the `ingrowth'of the short-lived daughter nuclides of 238U dur-ing melting are frequently employed (see [39] for areview). A number of `ingrowth' models havebeen proposed [10^12,40^42], but we use onlythe two simplest end members: dynamic melting[10] and equilibrium percolation melting [12]. Inthis section, we apply these models to the 226Ra^230Th^238U disequilibria of Kilauea's historicallavas and discuss the implications of our modelresults for mantle melting within the Hawaiianplume.
4.1. `Ingrowth' melting models and assumptions forKilauea
Both the dynamic and equilibrium percolationmelting models assume that melting occurs at aconstant rate in a one-dimensional, steady-state
EPSL 5747 22-2-01
A.J. Pietruszka et al. / Earth and Planetary Science Letters 186 (2001) 15^31 21
column of upwelling mantle, and that the meltand the residue remain in chemical equilibriumuntil the melt is extracted. During dynamic melt-ing, all melt in excess of a constant porosity isinstantaneously removed from the residue (at alldepths of the melting column, simultaneously)and mixed together [10]. In contrast, the meltand residue continuously interact as the meltmoves up the melting column during equilibrium
percolation melting [12]. For both models, the `in-growth' of U-series isotopes (and the creation of226Ra^230Th^238U disequilibria) is a direct conse-quence of the greater incompatibility of the short-er-lived daughter nuclide compared to the parent,which leads to di¡erent residence times on themelting column for 226Ra, 230Th, and 238U [39].Excesses of 226Ra and 230Th during dynamic melt-ing arise from a slow rate of melting relative tothe half-lives of 226Ra and 230Th, which allow theparent nuclides in the matrix to `grow' additionaldaughters and preferentially transfer them to themelt [10]. For equilibrium percolation melting,226Ra and 230Th excesses arise both from a chro-matographic e¡ect in which the daughter nuclidesmove faster up the melting column than their pa-rents, and from the continuous decay of slowerparents to augment their faster daughters [12].
We make the following assumptions for our`ingrowth' model calculations: (1) the source isin secular equilibrium prior to melting, (2) theheight of the melting column is 55 km (V80^135 km deep [43,44]), and (3) melting occurs inthe presence of residual garnet [1] with a constantsource mineralogy of 60% olivine, 15% clinopyr-oxene, 15% orthopyroxene, and 10% garnet. Inaddition, we use estimates for the degree of partialmelting at Kilauea (V5^10% [3]; Fig. 1) to ac-
Fig. 3. Melting `grids' calculated from the (a) dynamic and(b) equilibrium percolation melting models (using the `Ki-lauea' D values from Table 2 and a constant melt fraction of10%) and contoured for di¡erent values of melting rate (M :kg m33 yr31) and melt-zone porosity (P : vol. fraction). Forthe modeling, similar melting grids were computed to deter-mine the M and P values required to match the 226Ra^230Th^238U disequilibria of each lava (at its particular meltfraction from Fig. 1). This calculation was repeated (for eachlava) using the range of partition coe¤cients from Table 2.For comparison, melting trends for the stable trace elementequivalents of the dynamic [11] and equilibrium percolation[12] melting models are also shown. These heavy dashed andsolid curves represent the possible 226Ra^230Th^238U disequi-libria that can be obtained without the `ingrowth' of radioac-tive daughter nuclides (plotted at a range of melt fractions).Our average þ 2c error bars are shown on each plot. Allmodel calculations assume a 2800 kg m33 liquid density, a3300 kg m33 solid density, and (for equilibrium percolationmelting) that melt is distributed in tubes (permeability expo-nent = 2). The sample symbols are the same as in Fig. 1.6
EPSL 5747 22-2-01
A.J. Pietruszka et al. / Earth and Planetary Science Letters 186 (2001) 15^3122
count for any melt fraction e¡ect on the 226Ra^230Th^238U disequilibria. Since the partition coef-¢cients for Ra, Th, and U during mantle meltingmay vary as a function of pressure, temperature,and composition [31], we also evaluate a range ofexperimental D values (each combination of DRa
and DTh;U from [28^31,45,46]) and a set of opti-mal `Kilauea' partition coe¤cients chosen fromthese experiments (Table 2). Given these assump-tions (and a few others listed in Fig. 3), we deter-mined the unique combination of melting rate andmelt-zone porosity required to match the (226Ra/230Th) and (230Th/238U) ratios of each lava (at itsparticular melt fraction from Fig. 1) using theequations for dynamic melting from [11] and anumerical solution to the equations for equilibri-um percolation melting [47]. This method is illus-trated in Fig. 3.
4.2. Model results
The 226Ra^230Th^238U disequilibria of Kilaueaand Mauna Loa lavas have been used previouslyto quantify the melting parameters for tholeiiticbasalt production within the Hawaiian plume [16^19,23]. These studies calculated melting rates ofs 0.0005^0.01 kg m33 yr31 and melt-zone poros-ities of 6 0.1^3% using both the dynamic andequilibrium percolation melting models. We dis-covered that it was relatively di¤cult to reproducethe combined 226Ra and 230Th excesses of Ki-lauea's historical lavas (even though these areamong the smallest 226Ra^230Th^238U disequilib-
ria observed in lavas from mid-ocean ridges[21,22] or other ocean island volcanoes [23]) be-cause some combinations of DRa and DTh;U pro-duced (226Ra/230Th) and (230Th/238U) ratios thatwere simply too small or too large. Furthermore,we found that the melting parameters dependedstrongly (by an order of magnitude or more) onthe values of the partition coe¤cients (Fig. 4).Both of these problems result from the narrowshape of the melting `grid' for small 226Ra^230Th^238U disequilibria (Fig. 3). Thus, it is di¤-cult to infer a melting rate or porosity solely fromthe (230Th/238U) and (226Ra/230Th) ratios of Ha-waiian shield lavas. Instead, geological constraintson the melting process are required.
4.3. Melting rate and melt-zone porosity estimatesusing a geological constraint
The geochemical di¡erences between the adja-cent, active Hawaiian volcanoes, Kilauea, MaunaLoa and Loihi (e.g. Pb, Sr, and Nd isotope [4] orTh/U [48] ratios) require that the mantle sourcesand melting regions of these volcanoes remainphysically distinct on the 25^50 km length scaleof the distance between them [49]. This idea canpotentially be used to constrain the maximum vol-ume of source mantle tapped by Kilauea. Sincethe rate of magma supply (Q) must be balancedby melting a source volume (V) at a rate (M) suchthat Q = VUM, the minimum M sampled by thisvolcano can then be calculated with estimates ofQ (from volcanological observations) and this
Table 2Summary of mineral/liquid partition coe¤cients
Clinopyroxene Garnet
Ra Th U Ra Th U
B 0.0005 0.0013 0.0009 0.00001 0.0015 0.0096LaT ^ 0.01 0.0045 ^ 0.0017 0.015H 0.0058 0.014 0.013 0.00070 0.0014 0.0059S and L ^ 0.005 0.005 ^ 0.009 0.028Kil 0.0005 0.005 0.005 0.00001 0.012 0.024
B: Beattie [29,45]. LaT: LaTourrette and Burnett [28]; LaTourrette et al. [46]. H: Hauri et al. [30]. S and L: Salters and Longhi[31]. Kil: `Kilauea' partition coe¤cients from [29,45] for DRa in clinopyroxene and garnet, [31] for DTh and DU in clinopyroxene,and a single experiment of [31] for DTh and DU in garnet. Partition coe¤cients for Ra have never been measured, but DRa isprobably less than DBa in these minerals due to the larger ionic radius of Ra [34]. We assume DRa = DBa for simplicity, but eval-uate a range of values. The D values for Ra ( = Ba), Th, and U in olivine and orthopyroxene are expected to be extremely low[29] and are assumed to be zero.
EPSL 5747 22-2-01
A.J. Pietruszka et al. / Earth and Planetary Science Letters 186 (2001) 15^31 23
maximum estimate of V (from the intershield geo-chemical di¡erences).
To apply this geological constraint, we assumethat Kilauea taps a cylindrical volume of sourcemantle with a height of 55 km (V135^80 kmdeep; Fig. 5), which is thought to represent themaximum vertical extent of the melting region onthe axis of the Hawaiian plume [43,44]. For thisgeometry, the radius of Kilauea's melting column(a relatively small portion of the much wider up-
Fig. 4. Melting rates and melt-zone porosities calculatedfrom the (a) dynamic and (b) equilibrium percolation meltingmodels using a range of bulk partition coe¤cients for Ra,Th, and U. The values plotted are those that best match theaverage (226Ra/230Th) and (230Th/238U) ratios of our data atmelt fractions of 5, 10, and 20%. The literature sources ofthe partition coe¤cients used for each calculation are indi-cated on the ¢gure (symbols for DRa and letter codes forDTh;U keyed to Table 2). The melt fractions (small numbers)are shown only for those parameterizations that display asigni¢cant change of melting rate or porosity as a functionof melt fraction. Some combinations of DRa and DTh;U pro-duced 226Ra^230Th^238U disequilibria that were simply toosmall (e.g. the `H' values of DTh;U from [30] for both `in-growth' melting models) or too large (e.g. the `S and L' val-ues of DTh;U from [31] for equilibrium percolation melting atmelt fractions of 5%) to account for the average 226Ra^230Th^238U disequilibria of the lavas. In these cases, approxi-mate values are shown (if possible).
Fig. 5. Maximum source volumes for historical Kilauea lavasestimated from the dynamic and equilibrium percolationmelting models using a range of partition coe¤cients for Ra,Th, and U. Ideally, the radii of these hypothetical sourcevolumes (plotted on each map assuming a cylindrical melt-column geometry with a maximum height of 55 km [43,44])should be less than half the distance between Kilauea,Mauna Loa and Loihi in order to preserve their geochemicaldi¡erences. The letter codes in each plot refer to the litera-ture source of the values used for DRa and DTh;U, respec-tively (Table 2). The results of this calculation are shownonly for those sets of partition coe¤cients that successfullyreproduce the observed range of 226Ra^230Th^238U disequilib-ria. See the text for more details.
EPSL 5747 22-2-01
A.J. Pietruszka et al. / Earth and Planetary Science Letters 186 (2001) 15^3124
welling plume) should ideally be less than half thedistance between the three active volcanoes in or-der to preserve the distinct geochemical signatureof Kilauea lavas. If we use a more conservativeradial limit of 25 km (V75% of the distance be-tween the summits of Kilauea and Mauna Loa),the maximum volume of source mantle sampledby Kilauea would be V1.1U105 km3.
Residence time analysis of the geochemical £uc-tuations observed in historical Kilauea lavas sug-gests that the volume of magma within the volca-no's summit magma storage reservoir hasremained essentially constant since 1790 [35].For such a steady-state volcanic system, the erup-tion rate may be used as a proxy for the magmasupply rate. Combining our best estimates of Q(from the eruption rates in Fig. 1) with the max-imum V of V1.1U105 km3 results in minimum Mfor Kilauea (assuming a 2800 kg m33 melt den-sity) that ranges from s 0.0001 kg m33 yr31 dur-ing periods of low magma supply (e.g. after the1924 eruption) to s 0.003 kg m33 yr31 duringperiods of high magma supply (e.g. 1790, theearly 19th century, and the Puu Oo eruption).
To be successful, our `ingrowth' model calcula-tions must be consistent with these minimummelting rates. In other words, the combinationof Q (from Fig. 1) and M (inferred from a givenparameterization of either `ingrowth' model) foreach lava must result in a V that is small enoughto satisfy the geological constraint from the inter-shield geochemical di¡erences. Thus, we deter-mined a V for each lava (for each combinationof DRa and DTh;U from Table 2) using the meltingrate inferred from its (226Ra/230Th) and (230Th/238U) ratios (at its particular melt fraction; Fig.1) and our best estimate of the volcano's magmasupply rate at the time of its eruption (Fig. 1). Foreach set of partition coe¤cients, this calculationresults in a range of V values that re£ect the dif-ferent values of Q and M used for each lava.Radii were calculated from these source volumesas described above. For comparison, the largestmelt-column radius for each set of partition coef-¢cients that reproduced the observed range of226Ra^230Th^238U disequilibria is plotted on amap of the Island of Hawaii (Fig. 5).
In all cases, the maximum melt-column radii
are similar to or signi¢cantly larger than halfthe distance between the three active volcanoes.Even if we use the more conservative radial limitof 25 km, only two sets of Ra, Th, and U parti-tion coe¤cients (the values of [29,45], and theoptimal `Kilauea' D values from [29,45] for DRa
Fig. 6. Temporal variations of (a) melting rate, (b) upwellingrate, and (c) source volume inferred from Kilauea's historicalsummit lavas. The melting rates are calculated to match the(226Ra/230Th) and (230Th/238U) ratios of each sample usingthe equilibrium percolation model, the `Kilauea' D valuesfrom Table 2, and melt fractions from Fig. 1. The upwellingrates and source volumes are estimated from these meltingrates as described in the text. The range in these parametersfor the 1894 groundmass was calculated using our total ob-served variation of 226Ra^230Th disequilibria for Kilauea,rather than the (226Ra/230Th) ratio of this sample (which maybe elevated due to crustal contamination). The previousranges for these melting parameters based on Puu Oo lavasare from [19]. The symbols are the same as in Fig. 1.
EPSL 5747 22-2-01
A.J. Pietruszka et al. / Earth and Planetary Science Letters 186 (2001) 15^31 25
and [31] for DTh;U) with the equilibrium percola-tion melting model satisfy the geological con-straint. Since the Th and U partition coe¤cientsof [31] are thought to be more appropriate for themelting of garnet peridotite, we use the `Kilauea'D values (and the equilibrium percolation meltingmodel) for the remainder of the discussion.
Thus, our best estimates of the melting param-eters for tholeiitic basalt production within theHawaiian plume (as tapped by Kilauea) are amelt-zone porosity of V2% and melting rates ofV0.003^0.03 kg m33 yr31 (excluding a relativelylow melting rate of V0.001 kg m33 yr31 inferredfrom the 1929 lava, which is geochemically anom-alous [3] ; Fig. 2). This porosity estimate is similarto the highest values previously inferred from Ki-lauea and Mauna Loa lavas (V2^3% [19]),although it must strictly represent a maximumvalue (because of 226Ra decay during magmatransport and storage). In contrast, our calculatedmelting rates range to signi¢cantly higher valuesthan previous estimates for Hawaiian shield vol-canoes (s 0.0005^0.01 kg m33 yr31 [16^19,23]).
Interestingly, our `ingrowth' model results alsosuggest that Kilauea lavas from the 19th to early20th centuries were produced at systematicallyhigher rates of mantle melting than lavas from1790 and the late 20th century (V0.02^0.03 vs.0.003^0.005 kg m33 yr31, respectively; Fig. 6a).This shift derives mostly from the smaller 230Th^238U disequilibria of the 19th^early 20th centurylavas (at a given melt fraction; Fig. 7). The abso-lute range in the melting rate at Kilauea is di¤cultto quantify independently, but our calculationsusing the geological constraint from the inter-shield geochemical di¡erences suggest that rela-tively high Mv0.003 kg m33 yr31 (regardless ofany `ingrowth' model assumptions or choice ofpartition coe¤cients) are required during periodsof high magma supply (e.g. 1790 and the Puu Ooeruption). The 19th^early 20th century lavas mayhave formed at much higher melting rates (up to afactor of V10) even though their (230Th/238U)ratios are only slightly lower because both `in-growth' models predict that large changes in therate of mantle melting (at the high M appropriatefor Kilauea) are required to cause even a subtleshift in the amount of excess 230Th (Fig. 7).
Although this wide range is model-dependent,we believe that the relative di¡erences in M be-tween these periods are signi¢cant because (1)similar results (of di¡erent magnitude) are ob-tained using both `ingrowth' models with all com-binations of DRa and DTh;U that reproduce theobserved range of 226Ra^230Th^238U disequilibria(the parameterizations shown on Fig. 5) and (2)they correlate with the changes in the trajectory ofthe temporal geochemical variations that followthe two major explosive eruptions at the summitof the volcano in 1790 and 1924 (Fig. 2; [3]).However, we also evaluate the possibility thatthe apparent di¡erences in M result from an in-correct model assumption.
Fig. 7. The e¡ect of melt fraction on 226Ra^230Th^238U dis-equilibria calculated from the equilibrium percolation meltingmodel (using the `Kilauea' D values from Table 2) and con-toured for di¡erent values of melting rate (M) or melt-zoneporosity (P). The constant M and P values indicated in theupper left corners of the plots are the melting parametersthat best match the average (226Ra/230Th) and (230Th/238U)ratios of our data. For comparison, the melting trend for thestable trace element equivalent of the equilibrium percolationmelting model [12] is also shown. The melt fraction estimatesare from Fig. 1. The larger black symbols (TIMS data) arethe same as in Fig. 1. The average TIMS þ 2c error barsare shown on each plot. The smaller black symbols are recal-culated (230Th/238U) ratios using high-precision PIMMS(230Th/232Th) analyses (A. Pietruszka, unpublished data). ThePIMMS (230Th/238U) ratios have þ 2c analytical uncertaintiesof V0.3% (similar to the size of the smaller symbols).
EPSL 5747 22-2-01
A.J. Pietruszka et al. / Earth and Planetary Science Letters 186 (2001) 15^3126
4.4. The e¡ect of melt fraction
Kilauea's historical lavas display a systematictemporal £uctuation of incompatible trace ele-ment ratios (e.g. La/Yb or Nb/Y) that is thoughtto result from a rapid change in the degree ofpartial melting at this volcano over the last 200yr [3]. Our previous `instantaneous' (batch) melt-ing model calculations based on these geochemi-cal variations (Fig. 1) suggest that the melt frac-tion at Kilauea decreased by a factor of V2(V10^5%) from the 19th to mid-20th century,and subsequently increased towards the present(again by a factor of V2). These melt fractionestimates are compatible with the `ingrowth' melt-ing models because the geochemical e¡ects of dy-namic melting (at melt fractions above V5% andrelatively high porosities) and equilibrium perco-lation melting for stable trace elements are similar[11] or identical [12] to batch melting, respectively.Although the relative (factor of V2) di¡erence inthe degree of partial melting that we use for our`ingrowth' model calculations is probably accu-rate, the absolute values are model-dependent[3]. However, incorrect, assumed melt fractionscannot account for the range in melting rates be-cause neither `ingrowth' model predicts a strongrelationship between 226Ra^230Th^238U disequilib-ria and melt fraction (over the V5^20% rangeexpected for Hawaiian shield lavas [50] ; Fig. 7).At most, the decrease in the degree of partialmelting from the 19th to mid-20th century mayexplain the subtle (V1%) increase in the (230Th/238U) ratios of the lavas erupted during this peri-od, which plot along a line of nearly constant M(Fig. 7). Even if we assume a constant melt frac-tion (despite the trace element evidence to thecontrary [3]), the M inferred from the 19th^early20th century lavas would still be relatively highdue to their subtly lower (230Th/238U) ratios (e.g.melting `grid' at a constant 10% melt fraction;Fig. 3).
4.5. The role of source heterogeneity
The short-term £uctuation in melt fraction atKilauea since 1790 correlates with an overallchange in the mantle source composition (or pro-
portion of source components; Fig. 1). Thesource that was preferentially tapped during theearly 20th century (when melt fractions were low-est) corresponds to the `Kilauea' isotopic endmember for Hawaiian volcanoes (characterizedby relatively low 87Sr/86Sr, and high ONd and206Pb/204Pb), which is thought to represent along-term depleted component within the Hawai-ian plume [3]. The relatively high (230Th/232Th)and (226Ra)/Ba, and low Ba/U, Ba/Th, and Th/Uratios of the early 20th century lavas compared toother Kilauea (Fig. 1 and Table 1) and historicalMauna Loa lavas [16^20] suggest that this `Ki-lauea' source component is also more depletedin highly incompatible trace elements at thepresent time (compared to the source of the early19th and late 20th century Kilauea or historicalMauna Loa lavas). A simple explanation for thisinverse relationship between the magnitude ofsource depletion and melt fraction (Fig. 1) isthat source fertility (i.e. lithology) controls thedegree of partial melting at Kilauea. For example,the (more depleted) source of the early 20th cen-tury lavas might have a relatively high solidustemperature, which would prevent it from meltingas much as the (less depleted) source of the early19th and late 20th century lavas.
The results of the `ingrowth' melting models for226Ra^230Th^238U disequilibria are independent ofthe mantle source composition (e.g. Th/U ratio)as long as the source is in secular equilibriumprior to melting [10^12]. However, melting alithologically heterogeneous source in variableproportions might cause the (230Th/238U) ratiosof lavas to vary if the bulk Th and U partitioncoe¤cients are di¡erent for each component[51,52]. The magnitude of this e¡ect dependsmostly on the relative abundances of garnet sincethis is the mineral typically expected to createsigni¢cant 230Th excesses during mantle melting[28^31,45,46]. Indeed, a mixed source lithologyhas been proposed for Hawaiian lavas (garnet-poor peridotite vs. garnet-rich pyroxenite or eclo-gite) based on a correlation between their 230Th^238U disequilibria and Th/U ratios [52] and othergeochemical evidence [53,54].
However, the overall 230Th excesses of Ki-lauea's historical lavas do not correlate with either
EPSL 5747 22-2-01
A.J. Pietruszka et al. / Earth and Planetary Science Letters 186 (2001) 15^31 27
their (230Th/232Th) or Th/U ratios (compare plotsin Fig. 2), or any other tracer of mantle sourcecomposition (cf. [3]). Even the well de¢ned trendtowards the `Kilauea' end member from the 19thto early 20th centuries is characterized by a subtleincrease in (230Th/238U), which is opposite to therelationship expected from melting a more de-pleted, garnet-poor source over time. If sourcefertility controls the degree of partial melting atKilauea (as we propose), the lithological hetero-geneity must be dominated by di¡erences in theabundance of clinopyroxene (rather than garnetand clinopyroxene). This could provide the neces-sary di¡erence in solidus temperature with littlee¡ect on the 230Th^238U disequilibria (due to thegeneral similarity of Th and U partition coe¤-cients for this mineral [28^31,45,46]). The onlypossible exception to this scenario comes fromthe geochemically anomalous lavas erupted justafter 1924, which have relatively high (230Th/238U) and Th/U and low (230Th/232Th) ratios(Fig. 2) and may have tapped a source that wasless depleted and richer in garnet than the sourceof the early 20th century lavas. In any case,source heterogeneity cannot account for the rangein melting rates.
4.6. The importance of the melt-column height
Our `ingrowth' model calculations assume aconstant melt-column height (d) for Kilauea. Al-ternately, it is possible that d is variable. How-ever, the melting rate inferred from the (226Ra/230Th) and (230Th/238U) ratios of a lava (for aconstant source mineralogy) is independent ofthe melt-column height for both dynamic melting(because d is not a parameter in the equations[11]) and equilibrium percolation melting (becaused cancels out of the equations [12] for melts leav-ing the top of the melting column). Thus, varia-tions in d cannot explain the di¡erences in M.
4.7. The dynamics of mantle melting at Kilauea
Based on our evaluation of the `ingrowth' mod-el parameters, an incorrect assumption is unlikelyto account for the range in melting rates inferredfrom the 226Ra^230Th^238U disequilibria of Ki-
lauea's historical lavas. To explore the implica-tions of these di¡erences in M, we converted ourbest estimates of the rate of mantle melting at thisvolcano since 1790 (from Fig. 6a) to upwellingrates (W) using the relationship W = MUd/bsUFfrom [14], where d is the melt-column height (55km), bs is the solid density (3300 kg m33) and F isthe melt fraction from Fig. 1. This calculation(Fig. 6b) suggests that Kilauea lavas from the19th to early 20th centuries were derived frommantle that was upwelling signi¢cantly fasterthan the mantle that melted to form the lavasfrom 1790 and the late 20th century (V300^1000 vs. 50^90 cm yr31, respectively). These up-welling rates range to signi¢cantly higher valuesthan previous estimates for Hawaiian shield vol-canoes (40^100 cm yr31 [19]). Although the widevariation (up to a factor of V10) is model-depen-dent, we believe that the relative di¡erences in Wbetween these periods are signi¢cant for the samereasons as the range in melting rates. Even if weassume that (1) the melt fraction is constant (de-spite the trace element evidence to the contrary[3]) or (2) the melt-column height varies in pro-portion to the degree of partial melting (i.e. aconstant melt productivity, F/d), the W inferredfrom the 19th^early 20th century lavas would stillbe relatively high due to their higher M.
Fluid dynamical models predict an exponentialincrease in the rate of mantle upwelling (and,thus, the rate of melting) towards the core ofthe Hawaiian plume (up to 1000 cm yr31
[43,44,55]). This basic pattern of W and M hasbeen veri¢ed using measurements of U-series iso-tope disequilibria in zero-age, spatially dispersedHawaiian lavas [19]. The range in the melting pa-rameters (both M and W) at Kilauea since 1790can also be explained in the context of this modelif (1) this volcano taps the center of the plume (incontrast to previous suggestions [8,53]) where Wand M are expected to be the highest and themost variable and (2) the volume (and radius) ofthe melting region sampled by the volcanochanges on a time scale of years to decades. Forexample, the 19th^early 20th century lavas mayhave been derived by melting a narrow regionnear the core of the plume (at the highest Mand W), whereas the lavas from 1790 and the
EPSL 5747 22-2-01
A.J. Pietruszka et al. / Earth and Planetary Science Letters 186 (2001) 15^3128
late 20th century may have been derived by melt-ing a wider area of the plume (at lower M and Won average).
Indeed, our `ingrowth' model results suggestthat the volume of the melting region sampledby Kilauea has changed signi¢cantly since 1790(Fig. 6c) with a short-term decrease during the19th to early 20th centuries (calculated from therelationship Q = VUM using our best estimates ofQ from Fig. 1 and M from Fig. 6a). Assuming acylindrical melting region with a constant height,these di¡erences in V translate directly to di¡er-ences in the melt-column radius. Smaller sourcevolumes (and radii) for the 19th^early 20th cen-tury lavas are obtained even if we assume a small-er range of M and/or a variable melt-columnheight (i.e. a constant melt productivity, F/d) be-cause the volcano's eruption rate (and, presum-ably, its magma supply rate) was signi¢cantlylower during this period (Fig. 1). In general, lavasthat formed at relatively high M and W derivefrom smaller volumes (and radii) of source mantleand vice versa (compare plots in Fig. 6), which isconsistent with the £uid dynamical models of theHawaiian plume [43,44,55].
The smallest source volumes were tapped byKilauea during the early 20th century when thevolcano erupted lavas with Pb, Sr, and Nd iso-tope ratios that correspond to the `Kilauea' endmember for Hawaiian volcanoes [3], whereaslavas derived from the largest source volumes(such as those from the Puu Oo eruption) overlapisotopically with recent Loihi tholeiitic basalts [4].This behavior probably results from the more ef-fective blending of small-scale heterogeneitieswithin the Hawaiian plume as the melting regionsampled by Kilauea increases in size (e.g. [5]).Thus, rapid changes in the size of the meltingregion tapped by the volcano (in the presence ofthese heterogeneities) may regulate most of thesource- and melting-related geochemical varia-tions at Kilauea over time scales of decades tocenturies.
5. Conclusions
In this study, we have used high-precision mea-
surements of the U-series isotope abundances ofKilauea's historical lavas (in conjunction with Pb,Sr, and Nd isotope ratios and incompatible traceelements) to examine the process of melt genera-tion at this volcano. Our results show that the226Ra^230Th^238U disequilibria of these lavashave remained relatively small and constant since1790 with V12 þ 4% excess 226Ra andV2.5 þ 1.6% excess 230Th (both are þ 2c). Modelcalculations based mostly on subtle variations inthe (230Th/238U) ratios suggest that Kilauea lavasfrom the 19th to early 20th centuries formed atsigni¢cantly higher rates of mantle melting andupwelling (up to a factor of V10) compared tolavas from 1790 and the late 20th century. Thesedi¡erences in M and W cannot be explained byanalytical error or by an incorrect model assump-tion. The shift to higher values for these parame-ters correlates with a short-term decrease in thesize of the melting region sampled by the volcano,which is consistent with £uid dynamical modelsthat predict an exponential increase in the upwell-ing rate (and, thus, the melting rate) towards thecore of the Hawaiian plume. The Pb, Sr, and Ndisotope ratios of lavas derived from the smallestsource volumes correspond to the `Kilauea' endmember of Hawaiian volcanoes, whereas lavasthat tapped the largest source volumes overlapisotopically with recent Loihi tholeiitic basalts.This behavior probably arises from the more ef-fective blending of small-scale source heterogene-ities as the melting region sampled by Kilaueaincreases in size. The source that was preferen-tially tapped during the early 20th century(when the melt fractions were lowest) is morechemically and isotopically depleted than thesource of the early 19th and late 20th centurylavas (which formed by the highest melt frac-tions). This inverse relationship between the mag-nitude of source depletion and melt fraction sug-gests that source fertility (i.e. lithology) controlsthe degree of partial melting at Kilauea. Thus,rapid changes in the size of melting region tappedby the volcano (in the presence of these small-scale heterogeneities) may regulate most of thesource- and melting-related geochemical varia-tions at Kilauea over time scales of decades tocenturies.
EPSL 5747 22-2-01
A.J. Pietruszka et al. / Earth and Planetary Science Letters 186 (2001) 15^31 29
Acknowledgements
We thank M. Spiegelman for stimulating dis-cussions about equilibrium percolation meltingand help with his `Usercalc' program, A. Hof-mann for sharing his unpublished Mauna UluPb isotope data, K. Spencer for his skilled main-tenance of the Hawaii isotope lab, L. Sacks forhelp with the `unspiked' analysis of Kil1919, andK. Sims for a helpful review. We greatly appreci-ated the assistance of R. Muse, K. Kolysko-Rose,and M. Ito with the preparation of the samplesfor chemical analysis. This work was supportedby NSF Grants EAR 9628288 (to K.R.) and9614247 and 0001123 (to M.G.). This is SOESTContribution #5301.[FA]
References
[1] A.W. Hofmann, M.D. Feigenson, I. Raczek, Case studieson the origin of basalt, III, Petrogenesis of the MaunaUlu eruption, Kilauea, 1969^1971, Contrib. Mineral. Pet-rol. 88 (1984) 24^35.
[2] J.M. Rhodes, S.R. Hart, Episodic trace element and iso-topic variations in historical Mauna Loa lavas: implica-tions for magma and plume dynamics, in: J.M. Rhodes,J.P. Lockwood (Eds.), Mauna Loa Revealed: Structure,Composition, History, and Hazards, AGU, Geophys.Monogr. 92, 1995, pp. 263^288.
[3] A.J. Pietruszka, M.O. Garcia, A rapid £uctuation in themantle source and melting history of Kilauea Volcanoinferred from the geochemistry of its historical summitlavas (1790^1982), J. Petrol. 40 (1999) 1321^1342.
[4] M.O. Garcia, A.J. Pietruszka, J.M. Rhodes, K. Swanson,Magmatic processes during the prolonged Puu Oo erup-tion of Kilauea Volcano, Hawaii, J. Petrol. 41 (2000) 967^990.
[5] M.D. Norman, M.O. Garcia, Primitive magmas andsource characteristics of the Hawaiian plume: petrologyand geochemistry of shield picrites, Earth Planet. Sci.Lett. 168 (1999) 27^44.
[6] M.D. Kurz, T.C. Kenna, D.P. Kammer, J.M. Rhodes,M.O. Garcia, Isotopic evolution of Mauna Loa Volcano:a view from the submarine southwest rift zone, in: J.M.Rhodes, J.P. Lockwood (Eds.), Mauna Loa Revealed:Structure, Composition, History, and Hazards, AGU,Geophys. Monogr. 92, 1995, pp. 289^306.
[7] F. Albare©de, High-resolution geochemical stratigraphy ofMauna Kea £ows from the Hawaii Scienti¢c DrillingProject core, J. Geophys. Res. 101 (1996) 11841^11853.
[8] J.C. Lassiter, D.J. DePaolo, M. Tatsumoto, Isotopic evo-lution of Mauna Kea Volcano: results from the initial
phase of the Hawaii Scienti¢c Drilling Project, J. Geo-phys. Res. 101 (1996) 11769^11780.
[9] V.M. Oversby, P.W. Gast, Lead isotope compositions anduranium decay series disequilibrium in recent volcanicrocks, Earth Planet. Sci. Lett. 5 (1968) 199^206.
[10] D. McKenzie, 230Th^238U disequilibrium and the meltingprocesses beneath ridge axes, Earth Planet. Sci. Lett. 72(1985) 149^157.
[11] R.W. Williams, J.B. Gill, E¡ects of partial melting on theuranium decay series, Geochim. Cosmochim. Acta 53(1989) 1607^1619.
[12] M. Spiegelman, T. Elliott, Consequences of melt transportfor uranium series disequilibrium in young lavas, EarthPlanet. Sci. Lett. 118 (1993) 1^20.
[13] S. Krishnaswami, K.K. Turekian, J.T. Bennett, The be-havior of 232Th and the 238U decay chain nuclides duringmagma formation and volcanism, Geochim. Cosmochim.Acta 48 (1984) 505^511.
[14] S. Newman, R.C. Finkel, J.D. Macdougall, Comparisonof 230Th^238U disequilibrium systematics in lavas fromthree hot spot regions: Hawaii, Prince Edward and Sa-moa, Geochim. Cosmochim. Acta 48 (1984) 315^324.
[15] I.M. Reinitz, K.K. Turekian, The behavior of the ura-nium decay chain nuclides and thorium during the £ankeruptions of Kilauea (Hawaii) between 1983 and 1985,Geochim. Cosmochim. Acta 55 (1991) 3735^3740.
[16] A.S. Cohen, R.K. O'Nions, Melting rates beneath Ha-waii: evidence from uranium series isotopes in recentlavas, Earth Planet. Sci. Lett. 120 (1993) 169^175.
[17] C. Hemond, A.W. Hofmann, G. Heusser, M. Condo-mines, I. Raczek, J.M. Rhodes, U^Th^Ra systematics inKilauea and Mauna Loa basalts, Hawaii, Chem. Geol.116 (1994) 163^180.
[18] A.S. Cohen, R.K. O'Nions, M.D. Kurz, Chemical andisotopic variations in Mauna Loa tholeiites, Earth Planet.Sci. Lett. 143 (1996) 111^124.
[19] K.W.W. Sims, D.J. DePaolo, M.T. Murrell, W.S. Bal-dridge, S. Goldstein, D. Clague, M. Jull, Porosity of themelting zone and variations in the solid mantle upwellingrate beneath Hawaii: inferences from the 238U^230Th^226Ra and 235U^231Pa disequilibria, Geochim. Cosmo-chim. Acta 63 (1999) 4119^4138.
[20] K.W.W. Sims, D.J. DePaolo, M.T. Murrell, W.S. Bal-dridge, S.J. Goldstein, D.A. Clague, Mechanisms of mag-ma generation beneath Hawaii and mid-ocean ridges: ura-nium/thorium and samarium/neodymium isotopicevidence, Science 267 (1995) 508^512.
[21] K.H. Rubin, J.D. Macdougall, 226Ra excesses in mid-ocean-ridge basalts and mantle melting, Nature 335(1988) 158^161.
[22] A.M. Volpe, S.J. Goldstein, 226Ra^230Th disequilibrium inaxial and o¡-axis mid-ocean ridge basalts, Geochim. Cos-mochim. Acta 57 (1993) 1233^1241.
[23] F. Chabaux, C.J. Alle©gre, 238U^230Th^226Ra disequilibriain volcanics: a new insight into melting conditions, EarthPlanet. Sci. Lett. 126 (1994) 61^74.
[24] A.J. Pietruszka, R.W. Carlson, E.H. Hauri, Accurate and
EPSL 5747 22-2-01
A.J. Pietruszka et al. / Earth and Planetary Science Letters 186 (2001) 15^3130
precise measurement of U-series disequilibria in volcanicrocks by plasma ionization multicollector mass spectrom-etry (PIMMS), EOS Trans. AGU, 2000, p. 422.
[25] K.P. Jochum, A.W. Hofmann, Contrasting Th/U in his-torical Mauna Loa and Kilauea lavas, in: J.M. Rhodes,J.P. Lockwood (Eds.), Mauna Loa Revealed: Structure,Composition, History, and Hazards, AGU, Geophys.Monogr. 92, 1995, pp. 307^314.
[26] H. Cheng, R.L. Edwards, J. Ho¡, C.D. Gallup, D.A.Richards, Y. Asmerom, The half-lives of uranium-234and thorium-230, Chem. Geol. 169 (2000) 17^33.
[27] T.L. Wright, Chemistry of Kilauea and Mauna Loa inspace and time, U.S. Geol. Surv. Prof. Pap. 735 (1971)1^40.
[28] T.Z. LaTourrette, D.S. Burnett, Experimental determina-tion of U and Th partitioning between clinopyroxene andnatural and synthetic basaltic liquid, Earth Planet. Sci.Lett. 110 (1992) 227^244.
[29] P. Beattie, The generation of uranium series disequilibriaby partial melting of spinel peridotite: constraints frompartitioning studies, Earth Planet. Sci. Lett. 117 (1993)379^391.
[30] E.H. Hauri, T.P. Wagner, T.L. Grove, Experimental andnatural partitioning of Th, U, Pb and other trace elementsbetween garnet, clinopyroxene and basaltic melts, Chem.Geol. 117 (1994) 149^166.
[31] V.J.M. Salters, J. Longhi, Trace element partitioning dur-ing the initial stages of melting beneath mid-ocean ridges,Earth Planet. Sci. Lett. 166 (1999) 15^30.
[32] W.C. Phinney, D.A. Morrison, Partition coe¤cients forcalcic plagioclase: implications for Archean anorthosites,Geochim. Cosmochim. Acta 54 (1990) 1639^1654.
[33] P. Henderson, Inorganic Geochemistry, Pergamon Press,New York, 1982, 353 pp.
[34] J. Blundy, B. Wood, Prediction of crystal-melt partitioncoe¤cients from elastic moduli, Nature 372 (1994) 452^454.
[35] A.J. Pietruszka, M.O. Garcia, The size and shape of Ki-lauea Volcano's summit magma storage reservoir: a geo-chemical probe, Earth Planet. Sci. Lett. 167 (1999) 311^320.
[36] J.D. Blundy, B.J. Wood, Crystal-chemical controls on thepartitioning of Sr and Ba between plagioclase feldspar,silicate melts, and hydrothermal solutions, Geochim. Cos-mochim. Acta 55 (1991) 193^209.
[37] N.C. Sturchio, J.K. Bohlke, C.M. Binz, Radium^thoriumdisequilibrium and zeolite^water ion exchange in a Yel-lowstone hydrothermal environment, Geochim. Cosmo-chim. Acta 53 (1989) 1025^1034.
[38] D.M. Shaw, Trace element fractionation during anatexis,Geochim. Cosmochim. Acta 34 (1970) 237^243.
[39] T. Elliott, Fractionation of U and Th during mantle melt-ing: a reprise, Chem. Geol. 139 (1997) 165^183.
[40] Z. Qin, Disequilibrium partial melting model and its im-
plications for trace element fractionations during mantlemelting, Earth Planet. Sci. Lett. 112 (1992) 75^90.
[41] H. Iwamori, 238U^230Th^226Ra and 235U^231Pa disequilib-ria produced by mantle melting with porous and channel£ows, Earth Planet. Sci. Lett. 125 (1994) 1^16.
[42] C. Richardson, D. McKenzie, Radioactive disequilibriafrom 2D models of melt generation by plumes and ridges,Earth Planet. Sci. Lett. 128 (1994) 425^437.
[43] S. Watson, D. McKenzie, Melt generation by plumes: astudy of Hawaiian volcanism, J. Petrol. 32 (1991) 501^537.
[44] N.M. Ribe, U.R. Christensen, The dynamical origin ofHawaiian volcanism, Earth Planet. Sci. Lett. 171 (1999)517^531.
[45] P. Beattie, Uranium^thorium disequilibria and partition-ing on melting of garnet peridotite, Nature 363 (1993) 63^65.
[46] T.Z. LaTourrette, A.K. Kennedy, G.J. Wasserburg, Tho-rium^uranium fractionation by garnet evidence for a deepsource and rapid rise of oceanic basalts, Science 261(1993) 739^742.
[47] M. Spiegelman, UserCalc: a web-based uranium seriescalculator for magma migration problems, Geochem.Geophys. Geosyst. 1, 2000, 1999GC000030.
[48] M.O. Garcia, K.H. Rubin, M.D. Norman, J.M. Rhodes,D.W. Graham, D.W. Muenow, K. Spencer, Petrologyand geochronology of basalt breccia from the 1996 earth-quake swarm of Loihi seamount, Hawaii: magmatic his-tory of its 1996 eruption, Bull. Volcanol. 59 (1998) 577^592.
[49] F.A. Frey, J.M. Rhodes, Intershield geochemical di¡er-ences among Hawaiian volcanoes: implications for sourcecompositions, melting process and magma ascent paths,Phil. Trans. R. Soc. London 342 (1993) 121^136.
[50] S. Watson, Rare earth element inversions and percolationmodels for Hawaii, J. Petrol. 34 (1993) 763^783.
[51] C.C. Lundstrom, Q. Williams, J.B. Gill, Investigating sol-id mantle upwelling rates beneath mid-ocean ridges usingU-series disequilibria, 1: a global approach, Earth Planet.Sci. Lett. 157 (1998) 151^165.
[52] O. Sigmarsson, S. Carn, J.C. Carracedo, Systematics ofU-series nuclides in primitive lavas from the 1730-36 erup-tion on Lanzarote, Canary Islands, and implications forthe role of garnet pyroxenites during oceanic basalt for-mations, Earth Planet. Sci. Lett. 162 (1998) 137^151.
[53] E.H. Hauri, Major-element variability in the Hawaiianmantle plume, Nature 382 (1996) 415^419.
[54] J.C. Lassiter, E.H. Hauri, Osmium-isotope variations inHawaiian lavas: evidence for recycled oceanic lithospherein the Hawaiian plume, Earth Planet. Sci. Lett. 164 (1998)483^496.
[55] E.H. Hauri, J.A. Whitehead, S.R. Hart, Fluid dynamicsand geochemical aspects of entrainment in mantle plumes,J. Geophys. Res. 99 (1994) 24275^24300.
EPSL 5747 22-2-01
A.J. Pietruszka et al. / Earth and Planetary Science Letters 186 (2001) 15^31 31
