Earth and Planetary Science Letters, 76 (1985/86) 135 150 135 Elsevier Science Publishers B.V., Amsterdam - Printed in The Netherlands

[6]

Comparison of Atlantic and Pacific paleochemical records for the last 215,000 years: changes in deep ocean circulation and chemical inventories

E . A . B o y l e 1 a n d L . D . K e i g w i n 2

J Department of Earth, Atmospheric, and Planetary Science, Massachusetts Institute of Technology, E34- 258, Cambridge, MA 02139 (U.S.A.)

2 Woods Hole Oceanographic Institution, Woods Hole, MA 02543 (U.S.A.)

Received March 20, 1985; revised version received August 13, 1985

Detailed Cd/Ca and (~13C data have been obtained for benthic foraminifera from western North Atlantic and Equatorial Pacific sediment cores. In the modern ocean, both tracers are closely linked to nutrient distributions. The sedimentary records for both tracers indicate that bottom waters overlying the Atlantic site have been nutrient-depleted relative to those at the Pacific site over the last 215,000 years. From this evidence it is reasonable to infer that there has been a continuous net flux of nutrient-depleted water from the western North Atlantic into the Pacific. This exchange has undergone significant fluctuations, with the export of nutrient-depleted Atlantic water diminishing by about a factor of two relative to the inflow from Southern Ocean sources. Over the last 215,000 years, carbon isotope fluctuations in both regions are dominated by variable storage of isotopically light carbon on continents with a lesser contribution from these deep ocean circulation changes. The cadmium signal in the North Atlantic is dominated by changes in deep ocean circulation patterns; cadmium shows less variability in the Pacific which may reflect changes in the global average cadmium content or minor changes in deep Pacific circulation patterns. Using these records to estimate global averages, it appears that glacial ocean water had 22% more Cd and 0.46%0 less 13C than the modern ocean. These numbers are estimates which may be revised as more data become available, although they are not likely to be as much as 20% in error for Cd or 0.29;~ for 13C. Relative North Atlantic Deep Water (NADW) formation rates are modulated with a significant 41 kyr periodicity linked to obliquity-induced variations in high latitude insolation; NADW lags 8 _+ 2 kyr behind insolation, however.

1. In troduct ion

Late Q u a t e r n a r y g l a c i a l / i n t e r g l a c i a l c l ima t i c

f l uc tua t ions h a v e a l te red deep ocean c i r cu la t ion

pa t t e rn s and g loba l ocean ic g e o c h e m i c a l i nven to -

r ies [ 1 - 7 a n d re fe rences therein]. T h e r eco rd of

these changes is p r e se rved in the i so top ic and

chemica l c o m p o s i t i o n o f shells of b e n t h i c for-

amin i fe ra . T h e record at any single site is de-

t e r m i n e d by changes in b o t h g loba l i nven to r i e s

a n d c i r c u l a t i o n - i n d u c e d ocean i c he te rogene i ty . T h e

r e l a t i v e i m p o r t a n c e o f t he se p r o c e s s e s to

p a l e o c h e m i c a l r ecords has b e e n con t rovers ia l . F o r

example , i t has b e e n sugges ted that the f lux of

N o r t h A t l a n t i c D e e p W a t e r ( N A D W ) re la t ive to

A n t a r c t i c d e e p wa te r sources has: (1) nea r ly c o m -

0012-821x/85/$03.30 © 1985 Elsevier Science Publishers B.V.

p le te ly ceased [2,17], (2) d i m i n i s h e d by a b o u t a

f ac to r of two [1], and (3) c h a n g e d l i t t le if at all [5].

R e s o l u t i o n of this c o n t r o v e r s y requ i res g loba l da t a

on ocean ic chemica l d i s t r ibu t ions , so tha t var ia-

t ions due to c h a n g i n g inven to r i e s can be dis-

t i ngu i shed f r o m those due to c h a n g i n g ocean circu-

l a t ion pa t te rns . A first pass at this p r o b l e m re-

qu i res da t a f r o m the A t l a n t i c and Pacif ic , s ince

these oceans r ep resen t the ex t r emes of ocean ic

chemica l f r ac t iona t ion . The re fo r e , new d a t a for

deep wa te r n u t r i e n t - l i n k e d t racers ( C d / C a and

813C in b e n t h i c fo ramin i f e r a ) f r o m cores in the

wes t e rn N o r t h A t l a n t i c and E q u a t o r i a l Paci f ic were

o b t a i n e d to assess m e c h a n i s m s c o n t r i b u t i n g to

p a l e o c e a n o g r a p h i c c h e m i c a l var iabi l i ty .

136

2. Core selection

The Atlantic core (Chain 82, Station 24, Core 4PC; 41°43'N, 32°51'W; 3427 m) has been the subject of previous studies including a few 14C dates, 23°Th dating, and foraminiferal micro- paleontology [8-10]. The core was raised on August 6, 1968; there was a void interval between 347 and 380 cm (using the depth markings on the core liner in which the core has been stored) apparently resulting from mechanical fracturing of the sedi- ment during the coring process due to incomplete penetration. At the time of splitting this void was described as a "missing" interval. Previous pub- lications do not mention this gap; time-strati- graphic correlations of their data with our 6180 record indicate that previously reported depths appear to be estimates of contiguous core sample rather than those indicated on the core liner. The oxygen isotope record (Fig. 1) shows no evidence for a stratigraphic gap in this section. The data reported here is given relative to the notation on

the original core liner; this is done to minimize offsets during future resampling. To compare our data with that of the previous literature, it is necessary to subtract 32 cm from depths greater than 347 cm. Correction for shrinkage during dry- ing (perhaps about 3% relative to the core liner depths) may also be necessary. Earlier publications also refer to this core by station number rather than by core number (i.e., CHN82-24). This study utilizes Uvigerina spp. for the C d / C a data, except in the upper portion of the core where C. kullen- bergi is used. Uvigerina are scarce in oxygen iso- tope stage 5e (the last interglacial period) in this core; one should therefore regard the stage 5e cadmium data with some reservation due to bio- turbation.

A shorter section covering the last 30,000 years from another Atlantic core (Chain 82, Station 50, Core 20; 43°29.9'N, 29°52.0'W; 3070 m) with a higher sedimentation rate was also studied to de- termine if bioturbational smoothing of the above core was a significant problem, and if different

CHAIN 82 STA 24 CORE 4PC 4 2 ° N 33°W 3427m

b l a O , PDB ~ ) l s C , PDB Cd/Ca, p m o l / m o l

4 3 1 0 0 .04 0 .08 0 .12

i _ _ __ . . . . 3

C. w u e l l e r s t o r f i C. wuellerstorfi ~ e r i n a , C. kullenbergi_

Fig. 1. 8]SO, 813C, and C d / C a data from benthic foraminifera from a core in the western North Atlantic ocean. Isotope data with respect to PDB. Species analyzed are C. wuellerstorfi (black diamonds), C. kullenbergi (open triangles), and Uvigerina spp. (black squares). Note coring void at 347-380 cm. Oxygen isotope stage numbers indicated between oxygen and carbon isotope records.

species of benthic foraminifera provide consistent signals. The top of the piston core (PC) was lost in the coring process, so the pilot gravity core (PG) was spliced with the piston core at a point where oxygen and carbon isotope values coincided. This composite core has a sedimentation rate of about 5 cm/kyr . 613C was analyzed both in samples of C. wuellerstorfi and Uvigerina spp., while C d / C a was analyzed in samples of these species as well as C. kullenbergi.

The Pacific core (Knorr 73, Station 4, Core 3PC; 0°22'S, 106°11 'W/ 3606 m) was collected for heat flow measurements and no previous strati- graphic work had been undertaken. The core was chosen on the basis of appearance and preliminary calcium carbonate data. Oxygen isotope data indi- cates good recovery of Holocene and late Pleisto- cene sections. The oxygen isotope record is similar to other cores for the most of the older portion, but there is an interval between early 6aSO stage 2 and late stage 5d which is not recognizable in terms of the standard global oxygen isotope stratigraphy [11]. Chemical and isotopic data for this section of the core are reported but excluded from interpretation. The disappearance of pink Globigerinoides ruber [12] between 300 and 320 cm in this core establishes that the interglacial peak below this confused zone is stage 5e. Because benthic foraminifera are scarce in many parts of this core, the cadmium data is reported as averages of all analyses within a depth interval comprising several individual sediment samples. The cadmium data from this core is based exclusively on Uvigerina spp., although analyses of three other species in the upper portion give results that are compatible with the data from Uvigerina spp.

Both cores were sampled in contiguous 3 cm intervals. The outer few millimeters adjacent to the core liner were removed and discarded. Samples were disaggregated in distilled water on a shaker table and wet-sieved at 150/lm. The samples were picked for the specific genera "semiquantitatively"; i.e. all of the individuals seen in a single pass through a sample were removed and counted. Re- picks of some samples indicated that this proce- dure usually yielded most of the desired species; those individuals missed in the first pass are usu- ally small and would not contribute much to the analytical results. Dried fine and coarse fractions were weighed so that absolute abundances per

137

gram could be estimated. Isotope data in both cores is based on C. wuellerstorfi with a few sup- plementary analyses of C. kullenbergi and Uvigerina spp.

3. Analytical methods

Oxygen and carbon isotope analyses were car- ried out with slight modifications of methods re- ported previously [1]. Analytical precisions are better than 0.1%o for 81SO and 813C. Data relative to the reference gas was converted to PDB assum- ing NBS 20 61~O = -4.18%o and 813C = -1.06%o [13]. For this data set the full corrections recom- mended by Craig [14] are applied; this correction was not applied in a previous study [1] so there are slight offsets between these data sets. The correc- tion is largest at periods of most positive 8180 (glacial intervals); the corrected results are -0.10%o lower in 813C than the uncorrected re- sults at the glacial maximum.

Samples (0.2-0.5 mg before cleaning) are analyzed for C d / C a by a modification of the method first reported by Boyle [15]. The method has been altered considerably so a complete de- scription is given in Appendix 1. Precision has been assessed by multiple measurements of stan- dard solutions and by replicate analyses of sam- ples. The relative standard deviation is a function of the concentration; small samples are imprecise with the error decreasing to a limiting precision at the opt imum sample size. 53 analyses of a solution containing 3.2 p m o l / m l cadmium and 17.3 /~mol/ml calcium had standard deviations of 3.4% for Cd and 1.5% for Ca. Precision of replicate graphite furnace injections is usually better than 2% for Cd. Independent pickings of foraminifera from the same core sample show poorer reproduci- bility than replicates of standard solutions. This difference is partially due to bioturbation, since real samples contain a mixture of individuals which lived at different times [16]. Lack of perfect rep- licability also may be due to imperfect cleaning. In cores with high sedimentation rates, minimal tem- poral variability, and little manganese carbonate overgrowth, precision approaches that of standard solutions, implying that imperfections in the clean- ing procedure are not generally a problem. Esti- mated precision of a single measurement on real

138

samples of opt imum sample size (about 0.4 mg) is 5%. On average, 3 -4 replicates were run for each reported data point, thus the precision of the data should be about 2.5-3%. Accuracy should be as good as precision, but for various reasons C d / C a ratios may have slight systematic biases. To com- pensate for potential bias, 1 kg of standard solu- tion (henceforth referred to as ARS-1) was pre- pared so that adjustments could be made if higher accuracy is required in the future. 64 analyses of this reference standard during the period when these samples were analyzed gave C d / C a = 0.185 + 0 .006/~mol/mol (lo std. dev. of population).

Finally, no trace metal analysis is free from the possibility of occasional contamination. During the period these samples were analyzed about 5% of the analyses fell high of trends in the data and

of replicate samples from the same core sample. These analyses are excluded from means, but such exclusion is noted in the data tables. Samples which would be erroneous for known causes are not reported: e,g. (1) the sample is too small; samples containing less than 3.3 /~mol/ml of Ca are always rejected while samples between 3.3 and 6.6 ~tmol/ml are retained only if consistent with replicates or trends in the data, (2) sloppy han- dling, or (3) excessive manganese carbonate over- growth.

4. Core data

Chemical and isotopic data for these cores are displayed in Figs. 1 and 2. Data tables are availa- ble upon request to the authors. The records show

KNR 73 S t a 4 C o r e 3PC 0*22's 106°11'w 3606m

~ ) 1 ' 0 PDB ~ ) ,~C PDB Cd/Ca, g m o l / m o l

4 3 0 .5 0 .0 - 0 . 5 0 .16 0 . 2 0 0 . 2 4

3 . . . . . . . . . . . . . . . _~ . . . . . . . . . . . . . . . . ~ . . . . . . . . . . . . ~ _ _ . . . .

Depth, rn ~ ~ C ~"

to o8%, +~0 05%= t2 5% C. w u e l l e r s t o r f i U v i g e r i n a spp.

Fig. 2. 6t80, ¢~13C, and C d / C a data from benthic foraminifera from a core in the equatorial Pacific ocean. Isotope data with respect to PDB. Species analyzed are C. wuellerstorfi (black diamonds) and Uvigerina spp. (black squares). Note disturbed oxygen isotope record from 140 to 215 cm. Oxygen isotope stage numbers indicated between oxygen and carbon isotope records. Approximate lo analytical errors indicated on bottom.

B E N T H I C A B U N D A N C E S PER GRAM, CHN 8 2 S T A 24 C O R E 4PC

E

111

U v i g e r i n a s p p . C. k u l l e n b e r ~

0 100 0 3 0 0 ~ 1

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

C. w u e l l e r s t o r f i

i~--_' . . . . . . ! . . . .

~ 7Z-7_22~2~

:illiii:i Fig. 3. Number /weight abundance of benthic foraminifera in sediment core from the western North Atlantic.

139

some loss of amplitude due to bioturbation. The long Atlantic record has about 90%, and the Pacific record has about 85%, of the maximum 81SO range found in other cores [17,18]. The expected reduc- tion of the major signals for carbon isotopes and cadmium approaches analytical precision, how- ever. Some sections of these cores have high sedi- mentation rates. The stage 5 section of core CHN82 Sta 24 Core 4PC has a sedimentation rate of 3.8 cm/kyr, one of the more rapidly accumulating stage 5 records reported to date. Conversely, many of the usable sections of core KNR73 Sta 4 Core

3PC have sedimentation rates less than 2 cm/kyr. Since benthic foraminifera undergo large abun-

dance changes, problems due to biological sedi- ment stirring are magnified [19]. In regions of rapid abundance changes, data from different species can appear to have spurious phase shifts relative to one another. In zones where abundance is low, reworking of individuals from elsewhere could result in loss of amplitude (at best) and spurious peaks (at worst). In order to enable evaluation of these effects, abundance data for benthic foraminifera are presented in Figs. 3 and

BENTHIC ABUNDANCES PER GRAM, KNR 73 STA 4 CORE 3PC

Uvi_gerina spp. G. w_uellerstorfi 8 16 0 4 8

2

Depth, m ~ - . . . . . . . . . . . . -6- - i ~ . . . . . .

Fig, 4. Number/weight abundance of benthic foraminifera in sediment core from the equatorial Pacific.

140

13-

U (.9 ,<

100

200

300

400

b~3C PDB

1.0 0.0 - ' .o o.,o,

2

3

- - - 4 ~

5

6

7

8

1 0

Cd/Ca , lamol /mol

0.1 0.2 T ~

ATLANTIC

PACIFIC

__/

. . . ~

0.3 0

100

200

300

4.00

Fig. 5. Carbon isotope and cadmium records from western North Atlantic and equatorial Pacific sediment cores plotted vs. estimated age.

4. Two zones of low abundance in particular should be noted: Uvigerina spp. in stage 5e of the Atlantic core (the cadmium record) and C. wuellerstorfi in late stage 6 (isotope record). These sections of the cores are likely to miss the true extrema.

The chronology in these cores is based on the oxygen isotope records as correlated to radiometri- cally-dated events. The depth-age assignments are shown and described in detail elsewhere [20]. Plots of the Atlantic and Pacific records vs. estimated age are shown in Fig. 5.

The discussion relies on the observation that Uvigerina spp. and C. kullenbergi incorporate cadmium into their shells in proportion to the amount in deep waters overlying a site. This rela- tion has been established by studies of core tops, which indicated a proportionality constant D = 2.0 +_ 0.4, lo [21]. The uncertainty more likely reflects geological error in finding modern benthic foraminifera rather than true variability in D;

nonetheless, it should be kept in mind that there is some uncertainty over the degree to which benthic foraminifera are precise recorders of bottom water chemistry. The core top values for CHN82 Sta 24 Core 4PC and KNR73 Sta 4 Core 3PC are 0.074 and 0.172 ~mol/mol , respectively. Using the P concentrations at these depths for the nearest GE- OSECS stations (Atlantic station 27, P at 3441 m = 1.19 ~mol /kg; Pacific station 339, P at 3597 m = 2 . 4 4 /~mol/kg) and the slightly different Pacific and Atlantic Cd-P relations summarized by Bruland [22], these core top values correspond to D = 2.55 and D = 2.21 respectively. The data fall within the envelope of the data of Hester and Boyle [21]. Since these cores have documented stratigraphies where geological problems are minimal, it is reassuring that D estimates for these cores are within 7% of their mean, comparable to the error of the analyses.

Uncertainty in benthic foraminiferal response

to bot tom water chemistry also applies to carbon isotope data, perhaps with more force: there is particular concern over the cause of the 0.7-1.1%o offset between C. wuellerstorfi and Uvigerina spp. [23]. In order to assess potential problems associ- ated with these offsets, a number of coexisting pairs of these species have been analyzed. In par- ticular, in the North Atlantic core CHN82 Sta 50 Core 20, the two species were analyzed for both C d / C a and 613C wherever they coexisted during the latest Pleistocene section of the core (Fig. 6). C. kullenbergi, C. wuellerstorfi, and Uvigerina spp. do not appear to have differential fractionation for Cd even though they have significant offsets in oxygen and carbon isotopic ratios. Offsets could result either from metabolic differences between species or from the influence of sedimentary pore water chemistry on shell chemistry [26]. If sedi- mentary pore water influence were the sole reason for inter-specific differences, one would expect similar differences for carbon isotopes and cadmium, since both of these tracers resemble nutrient distributions in sedimentary pore waters [24 26]. The data shown in Fig. 6 clearly rules out a pore-water-only explanation for species offsets in this mid-ocean core; the larger portion of the

141

6'-~C offset between these species must be due to metabolic effects. It is becoming apparent that C. wuellerstorfi and Uvigerina spp. show inconsistent downcore 613C offsets [6,7]; this difference also has been attributed to near-surface productivity variations [27]. N.J. Shackleton provided glacial stage 2 samples from eastern Atlantic core M12392 for Cd analyses of C. wuellerstorfi (120 cm; 613C = +0.23%o, C d / C a = 0.067 /~mol/mol) and Uvigerina spp. (200 cm; 6t3C = -1.20%o, C d / C a = 0.094 ~mol /mol) . 613C in Uvigerina spp. from this core does not differ significantly from 813C in contemporaneous specimens from Pacif ic core V19-30 [2], but C d / C a in both species from this core are significantly different from Pacific values, as are the carbon isotopic values in C. wuellerstorfi. C d / C a data from M12392 are comparable to that observed in Uvigerina spp. and C. wuellerstorfi from North Atlantic samples of the same age used in the present study. A limited amount of work on C. wuellerstorfi and N. umbonifera from Pacific core KNR73 Sta 4 Core 3PC also indicates that these species also do not show large variations in

Dcd. This evidence suggests that there are spatially

and temporally variable metabolic influences on

C H N 8 2 S t a 5 0 C o r e 2 0 43 o N, 30 o W, 3020 rn

6180, %oPDB ~13C, %oPDB C d / C a , / Jmo l /mo l U v i / g m k u l / g m w u e / g m

0 1 2 0 2 4 6 g G 5 4 3 2 10 05 o o 005 0.10 i i J i . • . , . . . . ~ ~ + . , i . . . . i i ~ ~ + , i J , , 0 5 1015

(3. 50- a

100- ~ e-

r-----2.0 ~,6o---~

Fig. 6. Benthic isotopic and cadn'uum records f rom a sediment core in the No r th At lant ic . Note separate depth scales for p i lo t grav i ty

(PG) and piston core (PC). Data plotted as circles is for Uvigerina spp.; diamonds, 62 wuellerstorfi; and squares, C. kullenbergi. Lines are drawn between adjacent points for each species. Since chemical and isotopic measurements were made on both species, few replicates were run; a few outlier points have been ignored in drawing the lines. Note that 62 wuellerstorfi abundance shows a local maximum during glacial extremum, so that bioturbation cannot drastically alter the chemical and isotopic observations.

142

81-~C in Uvigerina which do not appear to affect C d / C a . This effect appears to be particularly pro- nounced in continental margin cores under areas of high productivity. Pore water induced artifacts appear to be small, if not necessary negligible, since there is little evidence for parallel species offsets between 613C and C d / C a . In the following discussion, it will be assumed that C. wuellerstorfi accurately records (~13C of the bot tom water and that all three foraminifera, C. wuellerstorfi, C. kul- lenbergi and Uvigerina spp., provide accurate re- cords of bot tom water Cd. Nonetheless, despite the evidence cited above in support of this as- sumption, further work on mechanisms creating species offsets is necessary to retain confidence in the use of benthic foraminifera as tracers of deep water chemistry.

5. Global chemical inventory shifts

Variability in cadmium concentrations and 8~)C at a site can be caused both by changes in the global oceanic averages and by changes in deep ocean circulation patterns: ~13C also may be al- tered by variations in the mean oceanic phos- phorus concentration [28]. In principle, evaluation of contributions from global intentory change re- quires data from throughout the ocean. In prac- tice, a reasonable approximation can be obtained from representative cores from the western North Atlantic and central Pacific ocean. At present, the Pacific shows only slight chemical gradients be- cause of effective lateral advection and mixing [29]. Carbon isotope data from the eastern Equa- torial Pacific [2] through the open Equatorial Pacific (this work) to the western Equatorial Pacific [30,31] show similar 613C changes between core top and glacial maximum, indicating that the Pacific was also relatively chemically uniform dur- ing the most recent glacial period as well. Al- though the Atlantic is heterogeneous at present [32], and appears to have been more so in the most recent glacial period [3], the glacial western North Atlantic always represents nutrient-depleted deep water [1]. Since the Pacific dominates the average in any event, an analysis based on core data from the western North Atlantic and Equatorial Pacific is adequate for preliminary assessment of chemical inventory shifts.

Average deep water values for benthic for-

aminiferal C d / C a and 613C have been computed for several time intervals, including two adjacent pairs of full glacial/interglacial transitions (Table 1; Fig. 7) (stage 5e was not included because of scarcity of adequate benthic foraminifera for C d / C a analyses and the potential for errors due to bioturbation). The global average is calculated using the volume-weighted relation:

Oceanic average =

3 × (Pacific value) + 1 x (Atlantic value) (1)

(3+1t Applying this calculation to the present-day phos- phorus concentration of bot tom waters above these sites gives an average oceanic phosphorus con- centration of 2.1 /~mol/kg, which compares favorably with the value of 2.25 ~amol/kg based on averaging of extensive GEOSECS data [33]. Also, mean oceanic glacial/interglacial 613C con- trasts obtained from this formula are similar to estimates for the isotope stage 5 / 6 contrast based on more extensive data [4]. The data for these periods have been examined in different ways to test the robustness of the estimates:

(1) 8~SO extrema were chosen and correspond- ing 613C and C d / C a data for this depth were used. This method has the advantage of objectivity but the disadvantage of susceptibility to analytical and other random error sources. It also does not allow for an error estimate apart from purely analytical considerations.

(2) Individual analyses within a narrow range of depth interval were averaged. Since any single data point contains some analytical and geological error, this method provides an improved mean and a standard deviation for assessment of statistical significance. Although choice of depth interval is subjective, there is likely to be little controversy over these choices.

(3) The depth range of method 2 was adopted but the averaged Cd numbers reported in Appen- dix 1 (i.e., averages of all replicates within smaller depth intervals) were weighted by the length of the analyzed section. This method corrects a short- coming of method 2 caused by variable numbers of replicates for samples within the chosen depth range. Method 3 may provide a better time- weighted average for the depth interval when sig- nificant property changes are occurring over that time. This method has the disadvantage of giving

TA

BL

E 1

Gla

cial

/in

terg

laci

al o

cean

ic c

hem

ical

in

ven

tori

es

A.

Ex

trem

um

met

ho

d

Sta

ge

Atl

anti

c C

HN

82

-24

-4P

C

dep

th

Pac

ific

KN

R7

3-4

-39

C

318

O

313

C

Cd

/Ca

d

epth

Glo

bal

av

erag

e

81

80

31

3C

Cd

/Ca

31

3C

Cd

/Ca

Lat

e 1

10

13

2.59

1.

29

0.06

5 13

.5

17

2.31

0.

28

0.17

2 +

0.5

3

0.14

5 2

62

-65

4.

41

0.53

0.

113

92.5

95

4.

01

0.19

0.

188

- 0.

04

0.16

9

Lat

e 6

50

3-5

07

4.

12

- 0.

13

0.11

4 38

0 -3

83

.5

3.75

-

0.27

0.

201

-0.2

4

0.17

9 7a

6

40

-64

3.5

2.

72

0.79

0.

070

510

-51

3.5

2.

57

0.09

0.

168

+ 0

.27

0.14

3

B.

Ind

ivid

ual

an

aly

sis

aver

agin

g m

eth

od

Sta

ge

Atl

anti

c C

HN

82

-24

-4P

C

dep

th

~13

C-+

2o

/~n

n

ran

ge

Cd

/Ca-

+ 2

o/~

/n

Pac

ific

KN

R7

3-4

-39

C

dep

th

313

C -

+ 2

o/~

/n

ran

ge

Cd

/Ca

_+ 2

o/~

fn

Glo

bal

Av

erag

e

313C

_+ 2

o ~

C

d/C

a +

2o

~

Lat

e1

0-2

2

1.0

6+

0.1

0 (

n =

9)

2 5

6-7

1

0.48

-+0.

06 (

n =

3)

6 5

00

-52

1

0.13

_+0.

32 (

n -

3)

7a

62

7-6

65

0.

91 _

+ 0.1

0 ( n

= 1

3)

0.06

6 ±

0.0

05 (

n-1

6)

0-2

7

0.10

9-+

0.0

05 (

n =

10)

57

95

0.

118

-+ 0

.00

7(n

= 1

7)

33

7-3

83

0.07

2 ±

0.0

04 (

n =

29)

47

7 52

3.5

C.

Len

gth

-wei

gh

ted

av

erag

ing

met

ho

d (

Cd

on

ly)

0.21

+0

.12

(n

= 5

)

-0.2

1 _

+0.

06 (

n =

12)

-0.3

1 +

0.0

4 (

n -

13

) 0.

06_+

0.03

(n

=1

4)

0.17

1 +

_0.0

06 (

n =

5)

+0

.42

_+

0.0

9

0.14

5 _+

0.00

5

0.20

0_+

0.01

0 (n

= 2

0)

0.04

_+0.

05

0.17

7+_0

.008

0.

206

+ 0

.008

(n

= 1

8)

-0.2

0_

+0

.09

0.

184-

t-0.

006

0.18

5_+

0.00

8 (n

-1

9)

+0

.30

_+

0.0

3

0.15

7_+

0.00

6

Sta

ge

Atl

anti

c C

HN

82

-24

-4P

C

Pac

ific

KN

R7

3-4

-3P

C

Glo

bal

av

erag

e

dep

th

Cd

/Ca

d

epth

ran

ge

ran

ge

Cd

/Ca

C

d/C

a

Lat

e 1

0 -2

2

0.06

5 0

-27

0.

171

0.14

5 2

56

-71

0.

106

57

-95

0.

198

0.17

5

Lat

e 6

500.

5-52

1 0.

119

337

387

0.20

7 0.

185

7a

627

665

0.07

3 4

77

-52

7

0.17

9 0.

153

2o

he

re r

efer

s to

the

pro

bab

le e

rro

r o

f th

e w

eig

hte

d m

ean

, V

'(3

/4 o

r)z+

(1~4

0A)'.

144

E o E d o 3

0.2-

0.1

0.0

0.0"

0-

~o 0.5

1.0

ATLANTIC Cd

1

~ d L a t e ~

ATLANTIC 613C

INVl

CIRC

I L", 'e 2 L%,e Z a

PACIFIC Cd

T • INV CIRC

Late Late 1 2 6 7a

PACIFIC 813C

I N V . . . .

I I

t . t e Lalte 2 % 7a

Fig. 7. Chemical variability estimated for the whole ocean (dashed lines) and Atlantic and Pacific averages (solid bar plots) for two full glacial/full interglacial periods. The vertical scales indicate property values; the horizontal dimension indi- cates different times. Numbers near the bottom of the bars indicate the oxygen isotope stages being considered. For the stage 1 /2 comparison, the contributions due to changes in global inventories (1N V), deep ocean circulation (C1RC), and phosphorus inventory ( ,SP) are indicated. In principle, it is possible to compute continuous estimates through time; in practice, sedimentary distortions and time scale imprecision create difficulties which could lead to misleading results. These problems are minimized here by choosing time intervals with favorable foraminiferal abundances.

equal weight to numbers of varying reliability. Estimates from the three methods (Table 1)

tend to agree within standard deviations estimated by method 2. Subjectivity and the method of aver- aging therefore are not critical, and method 2 probably gives a reasonable estimate for the uncer- tainties due to analytical and geological errors.

These averages and equation (1) are used to estimate the average carbon isotope and benthic

foram C d / C a for deep water during glacial and interglacial periods. Data from CHN82 Sta 50 Core 20 PG, PC were not used in these calcu- lations, although it is consistent with averages from core 4PC. Average oceanic deep water during the last glacial maximum was about -0 .46 _+ 0.10 (2o formal errors) lower in 313C than during the latest Holocene. Average deep water 313C also appears to have been -0 .50 _+ 0.10 lower in glacial late stage 6 relative to interglacial stage 7a. Aver- age deep water benthic foram C d / C a during the last glacial maximum is calculated to be + 0.032 + 0.009 ( + 22%) higher relative to the latest Holo- cene; isotope stage 6 is +0.027 + 0.008 ( + 17%) higher than interglacial stage 7a. Estimates are not made for stage 5e because of the scarcity of ben- thic foraminifera used in the Cd analyses and the likelihood that mixing has obscured the true value.

A decrease in glacial 313C has been observed previously and attributed to variations in the amount of isotopically light carbon on the conti- nents, either as trees and soils or as organic-rich marine continental shelf deposits [6,7,20]. Ap- parent change in the cadmium content of the deep ocean is observed for the first time in this work. Broecker [7,34] proposed that if the carbon isotope changes are due to transfers of marine organic carbon on the continental shelf, then changes in the average deep water phosphorus and cadmium contents would be observed, although the glacial to interglacial inventory change estimated here (+22%) is less than Broecker hypothesized ( + 42%). The changes in global Cd inventory could be due to the mechanism envisioned by Broecker (with reduced amplitude), but the data do not provide insight into the mechanism of inventory fluctuation, but rather only provides an estimate of the magnitude.

It would be inadvisable to regard the exact value of cadmium change too seriously, since the proportional change is small and the estimate will be revised as more data accumulate. Nonetheless, the fact that a shift of similar magnitude is ob- served in two glacial/interglacial transitions sug- gests that small but significant changes in the oceanic inventory of cadmium (and therefore per- haps also phosphorus) are possible. It would be difficult to reconcile these data with the magnitude of change proposed by Broecker, however, since this scenario would require the Pacific estimate to

be 20% low relative to the hypothesized value. It is unlikely that a signal of this magnitude would be missed; preliminary work on benthic foraminifera from 2 other cores in the Pacific do not indicate higher glacial C d / C a values than observed here.

6. Changes in Atlantic deep ocean circulation pat- terns

Modern oceanic distributions of 6~3C and Cd reflect the interaction of large-scale deep water circulation with biological cycling. Low-cadmium, ~3C-enriched deep water is formed by cooling waters of subtropical origin at high latitudes in the North Atlantic. This water mass spreads south within a broad depth range centered at about 3000 m. In the North Atlantic, waters low in Cd and more positive 813C are indicative of southward- spreading North Atlantic Deep Water (NADW), and waters high in Cd and of more negative 313C are indicative of northward-spreading Antarctic Bottom Water (AABW) and shallower Antarctic Intermediate Water (AAIW). Detailed analysis of hydrographic properties allows further subdivision of modern water masses, but we are unlikely to be able to make fine distinctions from paleoceano- graphic data. In the Pacific, as remnants of N A D W (and less nutrient-depleted Antartic Bottom Water) age along the flow path, slightly higher Cd and more negative 813C are observed [22,29].

Various lines of evidence suggest that major changes have taken place in deep ocean circulation patterns over the late Quaternary, particularly for the NADW. The extent of N A D W diminishment during glacial periods has been controversial, with estimates ranging from nearly complete cessation [2,18], about a factor of two diminishment [1], to little or no change [5]. The data presented here shows that the western North Atlantic has re- mained nutrient-depleted relative to the Equatorial Pacific at all times over the last 250,000 years. This conclusion differs from that of Shackleton et al. [2] who presented carbon isotope data indicating no carbon isotope fractionation between the oceans during glaciation. Shackleton et al. utilized Uvigerina spp. in their analyses, which shows a different response than C. wuellerstorfi which is used for 313C in this study. It seems likely that the different conclusion they reached is a consequence of using a less-reliable species of benthic for-

145

aminifera (for 813C) and their continental-margin core location (which appears to enhance variable metabolic offsets in Uuigerina 813C). C d / C a in C. wuellerstorfi and Uuigerina spp. and 313C in C. wuellerstorfi all indicate that the central western North Atlantic has always been nutrient-im- poverished relative to the eastern Pacific during at least the last 215,000 years. Some confidence can be placed in this conclusion since it is based on concordant evidence from different tracers.

The extent of chemical fractionation between the oceans has varied significantly, however. Fig. 7 illustrates chemical changes for two glacial/ interglacial transitions. Variability at a core site through time is due both to changes in the global inventory and circulation-driven redistribution be- tween basins. A dimensionless circulation parame- ter ~b is defined as:

~b = C d / C d (2)

where Cd is the cadmium content of water at a site and C--c] is the mean oceanic cadmium concentra- tion at the time. ~b depends on ocean circulation patterns but not on changes in global chemical inventories. It can be shown (using an extension of an analysis developed fully elsewhere [28]) that the fraction of the total cadmium change at a site due to circulation is:

fc~r¢ = A• (3) aCd _ _ + + a ~ k Cd

where A+ and ACd are the observed differences in ~b and Cd, at a site, and ~b and Cd are the initial values. A relationship apportioning 813C variabil- ity can be derived by similar reasoning. Changes in C d / C a and 813C between latest Holocene and last glacial maximum are indicated in Fig. 7 and apportioned as per the above formula. Carbon isotope variability in both oceans is dominated by global average 813C. In the Atlantic, changes in 8a3C due to ocean circulation are countered by an opposing contribution due to variation of global mean phosphorus (assuming that variations in global mean phosphorus concentration are linked to those of cadmium). Atlantic C d / C a variability is dominated by changes in deep ocean circulation patterns, with a significant contribution from changes in the oceanic mean. For both tracers,

146

Pacific variability is dominated by the global mean, with small contributions due to ocean circulation.

Boyle and Keigwin [1] devised a simple model for the chemical composition of NADW. From C d / C a data, and assuming that the cadmium in- ventory remains constant, it was estimated that glacial NADW flux (FD) was 50% lower than at present relative to the flux of nutrient-enriched southern-source waters (Fss) as calculated by the formula

F s s _ - Fss Css = ~-~D Css (4)

CD Fs + Fs s

where CI) and Css are the concentrations of cadmium in N A D W and southern-source waters. If this same model is applied in view of the present estimate for changes in oceanic mean Cd, net N A D W reduction appears to be about 31%. This percentage reduction applies to the flux of com- posite N A D W (the sum of the flow of the primary sinking nutrient-depleted water plus the entrained southern source water). If the relative change in the sinking of nutrient-depleted water ( F s) is com- puted instead, the new inventory data suggests a reduction of about 48% in the flux of nutrient-de- pleted water into the deep North Atlantic (again, relative to the flux of southern source waters).

Mix and Fairbanks [5] argued against this inter- pretation of reduced glacial N A D W and volumet- ric replacement of this N A D W by southern source waters. This argument was based on the large 813C difference signal between glacial Uvigerina spp. in Pacific core V19-30 and North Atlantic core V30- 97 compared with a much smaller difference be- tween (offset-corrected) Uvigerina in V30-97 and C. wuellerstorfi in Equatorial Atlantic core V25-59. The magnitude of the difference they observed is not supported by the isotopic records of C. wuel- lerstorfi in CHN82 Sta 24 Core 4PC and CHN82 Sta 50 Core 20 PC, PG, nor is it supported by offset-corrected data on Uvigerina spp. in core C H N 82 Sta 50 Core 20 PC, PG; it is also incon- sistent with the C d / C a data. The discrepancy can be attributed to the unreliability of Uvigerina spp. as a recorder of bottom-water 6~3C, as discussed previously; this problem is highlighted by the large discrepancy between the signals f rom this species in two nearby cores (V30-97 and CHN82 Sta 50 Core 20 PC, PG). If the 613C data from this species is discounted, their objections to a reduced

N A D W flux are eliminated. Mix and Fairbanks also argued that the nutrient contrast between the oceans might not require a net export of water from the Atlantic to the Pacific since " the resi- dence time of deep water within the small Atlantic could be less than in the large Pacific" and there- fore the nutrient content of the southern source water might thereby be augmented more in the Pacific than in the Atlantic by the accumulated decomposition of biological debris. This scenario begs the question of where the nutrients come from; i.e., since all the biological debris is formed by upwelling deeper waters to the surface, biologi- cal recycling is not a net source of nutrients to the deep water. It would be possible to postulate a complex three-dimensional circulation in which the Atlantic only received an input of southern source water and then exported intermediate water to the southern ocean (thereby losing nutrients, since most biological decomposition occurs in the upper waters of the ocean). In part, this is just a semantic distinction, since the definitions dis- tinguishing intermediate water formation from deep water formation are arbitrary. Furthermore, there is no evidence for such a process, and it seems more reasonable to presume that the inter- mediate water circulation was similar to that of the modern ocean until and unless evidence to the contrary is found.

Mix and Fairbanks [5] also proposed that in- creases in the nutrient content of deep North Atlantic waters could be caused by an increase in the preformed-nutrient content of sinking NADW; if so, the flux of N A D W could have remained constant. They admitted that this scenario was weakened by their data on the carbon isotope composition of planktonic foraminifera in the Greenland Sea which is similar to the record of variation in the global average carbon isotope composition, suggesting that there have not been major changes in pre-formed nutrients. In fact, their data demonstrating little difference between global and pre-formed 6~3C is a more subtle con- straint. Their contention that only the forest /soi l portion of global 613C change should be sub- tracted from the high-latitude planktonic 6~3C sig- nal is incorrect. Extending the formal relations between oceanic phosphorus and carbon isotopes discussed elsewhere [28], it can be shown that:

Ppr = [8'3Cpt - 8 ' 3 C ] / R + P (5)

where R is the A6.13C/AP ratio in deep water, P is the mean oceanic phosphorus content, 613C is the mean oceanic 613C, and the subscript pf signifies the preformed concentration. If there is no signifi- cant change in the preformed carbon isotopic com- position, then preformed phosphorus could only change in parallel with mean oceanic phosphorus; mean phosphorus variations of the magnitude re- quired to explain all of the deep North Atlantic paleonutrient signal (nearly a factor of two as determined by Cd/Ca) seem unlikely, as it would have drastic (and unobserved) consequences, for atmospheric carbon dioxide [7,34]. On the other hand, since the inventory estimates discussed above suggest that oceanic cadmium (and therefore per- haps also phosphorus) increased by 22% during glacials, and if in fact there was no change in Atlantic preformed 613C relative to the global mean, then preformed phosphate could have in-

creased by as much as 0.4/~mol/kg.

7. Time-series analysis

It is apparent from the data that Cd and 613C variations are more complex than a glacial/ interglacial dichotomy. Some C d / C a and 613C values in isotope stage 3 are comparable to those of the Holocene, even though oxygen isotope val- ues are always closer to glacial than interglacial conditions. There is also evidence for short-term events of a few thousand years duration. A mid- Holocene minimum in C d / C a appears in both of the North Atlantic records presented here; the simplest interpretation of this observation would be that deep water of the North Atlantic is already moving towards its glacial mode. This trend is not seen in the carbon isotope data, although this difference could result from changes in continental biomass which mask the circulation signal. There also is evidence for shorter-term events within the Pleistocene as well. In early stage 3 there is a well-replicated spike of low C d / C a (near a similar spike of low 8180). Numerous other short-term events can be seen in the records. Because spurious events could be induced by worm tubes or other sedimentary distortions, serious consideration of these short-term events must await confirmation in other sediment cores, preferably with higher sedi- mentation rates. One generalizable feature of the Pacific Cd data is the occurrence of peak Cd

$

0 el

, x e'~

0 E

"5 E c

1 0 -

5 -

4?7

4 3

(213)

( 1 9 )

. . . . I I

147

1 , 0 - . . . . . . . . . . . . . . . . . . . 9 5 % C l

C l . . . . . . . . . . . . . . . . . . . . . . . . . 9 0 %

~" 0 . 9 - e~

o • ~ 0 . 8 -

0 . 7 -

05i . c 0 . 3 O

0 1

. . . . t . . . . 1 0 . 0 0 0 . 0 5 0 . 1 0

Frequency, cp kyr

Fig. 8. Spectral analysis of C d / C a data from the western North Atlantic. At top is the periodogram (2 d.f.). At bottom is the smoothed coherence estimates (4 d.f.). Spectral analysis of the carbon isotope data is presented elsewhere [20].

values on or slightly after glacial terminations. Such events are seen at the stage 9/10, 7/8, 5/6, and 1 /2 boundaries. Although this appears to be a consistent feature of the data, there is no obvious explanation for this feature.

Spectral analysis has been applied to the data in order to examine the frequency response, coherence and phase of the records with respect to variations in the earth's orbital parameters (Fig. 8). The spectral analysis technique is described elsewhere [20] and is based on the FFT method. The pres- ence of a strong 23 kyr precession-coherent signal in the carbon isotope records has been attributed to variations in tropical forest biomass linked to

148

precession-induced changes in continental humid- ity [20]. There is no 23 kyr spectral peak evident in the Cd data; if a spectral peak comparable to that observed in the carbon isotope record [20] were present, then it would be observable in this time series. The absence of a 23 kyr C d / C a peak sup- ports the contention that continental biomass is the mechanism leading to the peak in carbon isotope spectra, since other mechanisms that could affect 6~3C (deep water circulation and continen- tal shelf carbon storage) would be expected to affect Cd as well. 613C in both oceans, and C d / C a in the Atlantic, show high concentration of spec- tral power at the 100 kyr period which appears in almost all paleoclimatic spectra. The Pacific Cd record (for the interval from 410 kyr B.P. to 115 kyr B.P.) shows no significant spectral peaks.

The Atlantic C d / C a periodogram shows a strong peak near the 41 kyr obliquity period (Fig. 8). Although the record is too short for the statis- tics of this peak to be improved by band averag- ing, C d / C a is significantly coherent with obliq- uity in this frequency band (coherence= 0.979 where coherence for 90% confidence = 0.948). In this record, the average phase for C d / C a lags 112 _+ 25 ° behind that of obliquity; i.e. obliquity leads low C d / C a by 8 _+ 2 kyr (90% confidence limits). The 6~3C record is 180 ° out of phase with C d / C a in this frequency band; i.e., higher C d / C a and more negative 6~3C are precisely in phase, both lagging obliquity by 8 kyr. Since both #13C and C d / C a show periodogram peaks in this frequency band, it is probable that this variability in both tracers is driven by the same mechanism (either shelf carbon storage or deep water circu- lation patterns). In view of the previous discussion, the latter mechanism is more likely. 41 kyr peaks also have been reported previously in C d / C a data from the Rio Grande Rise [35] and from 6a3C on the Mid-Atlantic Ridge at 41°N [5,36]. The data from CHN 82 Sta 24 Core 4PC establishes that there is an 8 kyr lag of low C d / C a and more negative 613C relative to peaks in high-latitude summer insolation. This can be compared with a zero phase-relationship between benthic 8~3C and high-latitude foraminiferal SST and CaCO 3 re- ported by Mix et al. [36] on their core from the same latitude. Other types of evidence and model- ling will be needed to establish mechanisms link- ing obliquity to the deep ocean; further evidence

from deep ocean time series will help to establish the spatial and temporal variability of this linkage.

8. Conclusions

The western North Atlantic has been nutrient depleted relative to the equatorial Pacific at all times over the last 215,000 years. From this evi- dence it is reasonable to infer that there has been a net flow of nutrient-depleted deep water from the Atlantic into the Pacific. There have been signifi- cant fluctuations in the intensity of this exchange, however, with evidence that NADW is modulated in part with a 41 kyr periodicity which lags high latitude insolation maxima by 8 kyr. The data presented here can be used to estimate that the glacial ocean had about 22% more Cd than at present and was depleted in ~3C by about 0.47%0; this estimate will be revised as further data is obtained but is unlikely to be in error by more than 20% for Cd or 0.2%0 for 813C.

Acknowledgements

C. Eben Franks did the isotopic analyses (with great skill and humor). E.B. did the Cd analyses (with some skill but not much humor). We thank Glenn Jones, Wally Broecker, and another re- viewer for comments on the manuscript. Isotopic analyses were supported by NSF grant OCE 8217644 (to L.D.K.); sediment preparation, foram picking, and C d / C a analyses were supported by NSF grant OCE 8411141 (to E.A.B.). WHOI con- tribution No. 6025.

Appendix 1--Cd/Ca analysis of foraminifera

0.2 to 0.5 mg samples are gently crushed between glass plates (only sufficiently to open chambers to the cleaning process) and transferred to 0.5 ml polypropylene centrifuge tubes. Samples are cleaned ultrasonically in distilled water three times for 2 minutes each, siphoning off the supernatant to remove clays and fine-grained carbonates; the process is re- peated twice with methanol. Samples are cleaned of surface organics by a solution of hot (80-90°C) 2 × 1 0 5 M H202 in 0.1 M NaOH (10 minutes). Supernatant is siphoned off and fragments are then rinsed with distilled water. Samples are then cleaned by heating in 100/~1 of reducing cleaning solution (0.25 M citric acid in 16 M ammonia, made up to 1 M in hydrazine (N H 2 NH 2 ) Just before the cleaning step ( caution: hydrazine is volatile, toxic and explodes on contact with oxidants [37]). This reductive cleaning is continued for 30 minutes with a few seconds of ultrasonic agitation every two minutes. Fragments

149

are rinsed with 3 portions of distilled water, and then further cleaned by heating in two hot distilled water rinses (5 minutes each) followed by a final rinse in room temperature distilled

water. At this point samples are transferred to clean polypropylene

centrifuge tubes (because the inner surfaces of the original vials become contaminated by adsorption during the cleaning pro- cess). Samples are ultrasonically rinsed 1-5 times for 30 sec- onds each time in 0.001N HNO 3 (the number of rinses de- pends on the initial sample size and the amount lost in the early cleaning stages) to remove any contaminants adsorbed on their surfaces. Occasionally, if high levels of manganese carbonate were expected [38], the outer surfaces of samples were dissolved with 10 /~1 of 0.06N HNO 3. Samples are then rinsed with distilled water and dissolved in 50/~1 or more of 0.06N I-INO 3 using an ultrasonic bath to speed dissolution. Samples are tested for complete dissolution by a pH check (a few /d of solution placed on indicator paper should indicate pH < 4). Solutions are centrifuged for 5 minutes at 10,000g to remove residual particles, and analyzed for Cd by graphite furnace flameless atomic absorption spectrophotometry. 20 ~tl splits of these solutions are diluted to 5 ml (using a 10-3 M lanthanum solution in 0.05 M H C I + 2 X l 0 -3 M HNO 3 to minimize interferences from PH 3 in the acetylene [391) and then analyzed for calcium by flame AAS.

All containers were rigorously cleaned by procedures that have been described for trace metal analysis in seawater (e.g. see papers reviewed in [221). A clean air bench was used for all steps beginning with the fragment transfer to the final polypro- pylene vial; other standard trace element precautions were used throughout. Graphite furnace analysis was performed on a Perkin-Elmer models 5000/400 combination using normal non-pyrrolized graphite tubes and an atomization step with a one second ramp to 2400°C and peak height measurement. This choice of conditions was chosen after experimentation with alternate conditions (pyrrolitic tubes of several types, L'vov platforms, maximum power heating, and peak area mode) since this gave the best sensitivity and most freedom from interferences due to variations in the calcium concentration. There was still a small residual interference (<10%) on cadmium sensitivity depending on the calcium concentration of the final solution. This interference does not depend on the age of the graphite tube. To minimize the effects of this inter- ference, the range of calcium concentrations encountered is limited by using dilute acid dissolution and gradually adding acid until the sample is dissolved. The standard was made up in a matrix with a calcium concentration in the middle of the range encountered. Finally, a small correction was applied for the variation of the sensitivity due to differences in the calcium concentration of the sample solution relative to the standards as determined for the analytical conditions described above.

Since manganese carbonate overgrowths many produce spurious C d / C a data, samples are checked for their manganese concentration. The determinations are carried out on 10 /~1 aliquots by GFAAS using a non-pyrrolized tube with 2400°C maximum power heating and peak height measurement.

References

1 E.A. Boyle and L.D. Keigwin, Deep circulation of the North Atlantic over the last 200,000 years: geochemical evidence, Science 218, 784-787, 1982.

2 N.J. Shackleton, J. Imbrie, and M.A. Hall, Oxygen and carbon isotope record of East Pacific core V19-30: impli- cations for the formation of deep water in the late Pleisto- cene North Atlantic, Earth Planet. Sci. Left. 65, 233-244,

1983. 3 W. Curry and G.P. Lohmann, Reduced advection into

Atlantic Ocean deep eastern basins during last glaciation maximum, Nature 306, 577-580, 1983.

4 J.C. Duplessy and N.J. Shackleton, Carbon-13 in the world ocean during the last interglaciation and the penultimate glacial maximum, Progr. Biometor. 3, 48-54, 1984.

5 A. Mix and R.G. Fairbanks, North Atlantic surface-ocean control of Pleistocene deep-ocean circulation, Earth. Planet. Sci. Lett. 73, 231-243, 1985.

6 N.J. Shackleton, Carbon-13 in Uvigerina, tropical rainforest history and the equatorial Pacific carbonate dissolution cycles, in: The Fate of Fossil Fuel CO 2, N. Andersen and A. Malahoff, eds., pp. 401-427, Plenum Press, 1977.

7 W.S. Broecker, Ocean chemistry during glacial time, Geo- chim. Cosmochim. Acta 46, 1689-1705, 1982.

8 T.L. Ku, 1. Bischoff and A. Boersma, Age studies of mid- Atlantic Ridge sediments near 42°N and 20°N, Deep-Sea Res. 19, 233-247, 1972.

9 J. Bischoff, and T.L. Ku, Pore fluids of recent marine sediments, II. Anoxic sediments of 35 ° to 45°N Gibraltar to Mid-Atlantic Ridge, J. Sediment. Petrol. 41, 1008-1017, 1971.

10 T.J. Crowley, Temperature and circulation changes in the eastern North Atlantic during the last 150,000 years: evi- dence from the planktonic foram record, Mar. Micro- paleontol. 6, 97-129, 1981.

11 N.J. Shackleton, The oxygen isotope stratigraphic record of the Late Pleistocene, Philos. Trans. R. Soc. London, Ser. B 280, 169-182, 1977.

12 P.R. Thompson, A.W,H. Be, J.C. Duplessy and N.J. Shackleton, Disappearance of pink-pigmented Gloi- gerinoides ruber at 120,000 yr b.p. in the Indian and Pacific Oceans, Nature 280, 554-558, 1979.

13 P. Blattner and J.R. Hulston, Proportional variations of geochemical O-18 sca les - -an interlaboratory comparison, Geochim. Cosmochim. Acta 42, 59-62, 1978.

14 I-1. Craig, Isotopic standards for carbon and oxygen and correction factors for mass-spectrometric analysis of carbon dioxide, Geochim. Cosmochim. Acta 12, 133-149, 1957.

15 E.A. Boyle, Cadmium, zinc, copper, and barium in for- aminifera tests, Earth Planet. Sci. Lett. 53, 11-35, 1981.

16 E.A. Boyle, Sampling statistic limitations on benthic foram chemical and isotopic data, Mar. Geol. 58, 213-224, 1984.

17 D. Ninkovich and N.J. Shackleton, Distribution, strati- graphic position and age of ash layer "L" , in the Panama Basin region, Earth Planet. Sci. Lett. 27, 20-34, 1975.

18 S.S. Streeter and N.J. Shackleton, Paleocirculation of the deep North Atlantic: 150,000-year record of benthic for- aminifera and oxygen-18, Science 203, 168-171, 1979.

150

19 W. Hutson, Bioturbation of deep-sea sediments: oxygen isotopes and stratigraphic uncertainty~ Geology 8, 127-130, 1980.

20 L.D. Keigwin and E.A. Boyle, Carbon isotopes in deep-sea benthic foraminifera: precession and changes in low-lati- tude biomass, in: The Carbon Cycle and Atmospheric CO2: Natural Variations Archean to Present, E. Sundquist and W. Broecker, eds., pp. 319-328, American Geophysical Union, 1985.

21 K. Hester and E. Boyle, Water chemistry control of the Cd content of benthic foraminifera, Nature 298, 260-261, 1982.

22 K.W. Bruland, Trace elements in seawater, in: Chemical Oceanography, Vol. 8, I.P. Riley and R. Chester, eds., pp. 157-220.

23 J.-C. Duplessy, N.J. Shackleton, R.K. Matthews, W. Prell, W.F. Ruddiman, M. Caralp and C.H. Hendy, C-13 record of benthic foraminifera in the last interglacial ocean: impli- cations for the carbon cycle and the global deep water circulation, Quat. Res. 21,225-243, 1984.

24 G. Klinkhammer, Early diagenesis in sediments from the eastern equatorial Pacific, II. Pore water metal results, Earth Planet. Sci. Lett. 49, 81-101, 1980.

25 G. Klinkhammer, D.T. Heggie and D.W. Graham, Metal diagenesis in oxic marine sediments, Earth Planet. Sci. Lett. 61,211-219, 1982.

26 D.C. McCorkle, S.R. Emerson and P.D. Quay, Stable carbon isotopes in marine porewaters, Earth Planet. Sci. Lett. 74, 13-26, 1985.

27 M. Sarnthein, K. Winn and R. Zahn, Effect of paleoproduc- tivity on atmospheric CO 2 during glacial and interglacial times; the deep-sea sediment record, Trans. Am. Geophys. Union 66, 291, 1985.

28 E.A. Boyle, Paired carbon isotope and cadmium data from benthic foraminifera: implications for changes in deep oce- an P and atmospheric carbon dioxide, Geochim. Cosmo- chim. Acta, in press, 1985.

29 H. Craig, W.S. Broecker and D.W. Spencer, GEOSECS Pacific Ocean Expedition, Vol. 4. Sections and Profiles,

U.S. Government Printing Office, Washington, D.C., 1982; R.F. Weiss, W.S. Broecker, H. Craig, and D.W. Spencer, GEOSECS Indian Ocean Expedition, Vol. 6. Sections and Profiles, U.S. Government Printing Office, Washington D.C., 1982.

30 E. Vincent, J.S. Killingley and W.H. Berger, Stable isotope composition of benthic foraminifera from the equatorial Pacific, Nature 289, 639-643, 1981.

31 E. Vincent, J.S. Killingley and W.H. Berger, Stable Isotopes in benthic foraminifera from Ontong-Java Plateau, Box cores ERDC 112 and 123, Paleogeogr., Paleoclimatol., Paleoecol. 33, 221-230, 1981.

32 A.E. Bainbridge, GEOSECS Atlantic Ocean Expedition, Vol. 2. Sections and Profiles, U.S. Government Printing Office, Washington, D.C., 1980.

33 T. Takahashi, W.S. Broecker and A.E. Bainbridge, Supple- ment to the alkalinity and total carbon dioxide concentra- tion in the world oceans, in: Carbon Cycle Modelling, Scope 16, B. Bolin, ed., pp. 159-200, John Wiley, New York, N.Y., 1981.

34 W.S. Broecker, Glacial to interglacial changes in ocean chemistry, Progr. Oceanogr. 11. 151-197, 1982.

35 E.A. Boyle, Benthic hydrography and the 41 kyr obliquity cycle, in: Climate Processes and Climate Sensitivity, J. Hansen and T. Takahasi, eds., Am. Geophys. Union, Geo- phys. Monogr. 29, 360-368, 1984.

36 A. Mix, R. Fairbanks, A. McIntyre and W. Ruddiman, O-18, C-13, and CaCO 3 in late Quaternary North Atlantic sediments related to abyssal circulation, Geol. Soc. Am. Abstr. Prog. 14, 567, 1983.

37 J.W. Mitchell, T.D. Harris and L.D. Blitzer, Safe handling and purification of aqueous hydrazine, Anal. Chem. 52, 774-776, 1980.

38 E.A. Boyle, Manganese carbonate overgrowths on for- aminifera, Geochim. Cosmochim. Acta 47, 1815-1819, 1983.

39 G.L. Long and C.B. Boss, Phosphine depression of calcium in flame atomic emission spectrometry with a uniform droplet generator, Anal. Chem. 54, 2496 2502, 1982.