Paleoceanography of the South Atlantic Ocean from the Middle Eocene through the Oligocene

MISSING IMAGE

Material Information

Title:
Paleoceanography of the South Atlantic Ocean from the Middle Eocene through the Oligocene
Physical Description:
xiv, 315 leaves : ill. ; 29 cm.
Language:
English
Creator:
Mead, Gregory A
Publication Date:

Subjects

Genre:
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1995.
Bibliography:
Includes bibliographical references (leaves 294-314).
Statement of Responsibility:
by Gregory A. Mead.
General Note:
Typescript.
General Note:
Vita.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 002046815
notis - AKN4753
oclc - 33430257
sobekcm - AA00004716_00001
System ID:
AA00004716:00001

Full Text














PALEOCEANOGRAPHY OF THE SOUTH ATLANTIC OCEAN
FROM THE MIDDLE EOCENE THROUGH THE OLIGOCENE










By

GREGORY A. MEAD


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


1995


























This Dissertation is dedicated to the memories of my

Grandmother Minnie Morris and my Mother Shirley Lopez,

Father Daniel Weiss, and Step-father Bruce Mead.









ACKNOWLEDGMENTS


I would like to thank first my advisor and friend, Dr.

David A. Hodell, and the rest of my committee, Drs. Paul F.

Ciesielski, Douglas S. Jones, Neil D. Opdyke, and Claire L.

Schelske, for (at various times) advice, encouragement, and

financial support. Drs. G. W. Brass, J. P. Kennett, D. V.

Kent, K. G. Miller, J. S. Oslick, E. Thomas, J. D. Wright,

and J. C. Zachos contributed valuable comments on the

chapters. The idea for the Braarudosphaera study was

initiated in a class led by Dr. E. L. (Jerry) Winterer. Dr.

Kennett also supplied the washed samples from ODP Hole 689B,

and Dr. Zachos allowed me to use unpublished data from ODP

Hole 748B. It is impossible to thank Ray Thomas enough for

keeping the equipment running. Rich Cooke, Jason Curtis,

Dr. Ann Heatherington, Vicky Mejia, and Dr. Paul Mueller

also helped materially in the laboratory. Richard Crockett

was the Scanning Electron Microscope technician. Hallie

Smith produced the drafted figures.

My fellow Micropaleo lab students, Alex Amigo, Jennifer

Callahan, Jason C., Jose Garrido, Rick Greenfield, Anon

Vayavananda, and Kathy Venz, were invaluable sources of

advice, commentary, brain-storming, and entertainment. I

thank them, along with all those who worked in the lab,

Soraya Balmes, Gail Beals, Belinda Seidel Breidenbach, Gil

Butler, Rich C., Charlotte Davison, Sabine Eppler, Paul

Gremillion, Binhe Gu, Alexandra Isern, Vicky M., Danny

Muller, Mark Mulligan, Tamara Mullis, Sarah Palmer, Mark










Schult, Laura Stanley, Loni Terry, Jang Vayavananda, Ana

Vrba, and Doug Wilder, for their friendship and for making

the lab a pleasant (if at times distracting) place to work

and study.

In addition, I especially thank Gail B., Kim D'Arcy and

Paul Meyers, Vic Divenere (also for teaching me recorder),

Ken Gilland, Lori Guthrie, Jack Holt, George Houston, Paul

(the Mac Master) Kirk, Nancy Mullins (also for the loan of

her printer), Mark S., Ana V., and Mick and Julie Whitelaw

for their companionship, understanding, and friendship over

the years I've been here.

I want also to thank all my family, and my extended

families, the Abrams and the Beals, for their support and

love during this (at times) seemingly endless endeavour.

I also thank the Gainesville Cycling Club, the

Gainesville Area Fiddle Jam and Fast Molasses, and the J. J.

Finley Ultimate players for helping keep me healthy (more or

less) and sane (more or less).

Finally, I thank my cat Trash, and my other 4-footed

friends, Harold, Rhett, Yoda, and Barclay, for their

unquestioned and unending affection.

This research was supported by National Science

Foundation grants OCE-8858012 and DPP-8717854 and by the

Donors of the Petroleum Research Fund administered by the

American Chemical Society.









PREFACE


The late Paleogene (middle Eocene through Oligocene)

was an important time in the Earth's biogeographic,

paleoclimatic, paleogeographic, and paleoceanographic

history. It marked the transition between the warm

Cretaceous and early Paleogene, and the cooler Neogene

[Miller et al., 1991a,b]. The continents of the Earth

were approaching the present positions, as Australia and

India moved northward, the Atlantic Ocean continued to

widen, and the Tethys progressively constricted [Dercourt

et al., 1986]. It was during this time that major seaways

opened around Antarctica [Kennett et al., 1972], allowing

the Antarctic to cool to the point where glacial ice could

eventually build up.

A band of siliceous deposition widened as a result,

and the planetary temperature gradient increased, causing

more restricted biogeographic zonation. Deep thermohaline

circulation in the oceans also developed at this time.

In the Oligocene, a series of deposits in the

Atlantic and Indian Oceans (Braarudosphaera oozes) suggest

that unusual conditions repeatedly occurred, for reasons

that are poorly understood. Biostratigraphic markers are

relatively rare in the Oligocene, leading to poor

resolution in stratigraphy.

Isotopic studies of geochemical tracers have been

important parts of research into this time period in the









oceans. Progressive oxygen isotopic enrichment of

planktic and benthic microfossils has documented oceanic

cooling, mostly of high latitudes, with a major shift in

the earliest Oligocene suggesting probable glaciation

(confirmed by the finding of ice rafted debris in Indian

Ocean drill holes [Zachos et al., 1992b]) and cooling,

especially of bottom waters. Additional glaciations have

been postulated during the Oligocene as a result of other

6180 enrichments. An inversion in the usual 6180 gradient

in two sites just off the Antarctic coastline has been

interpreted to result from the production of warm saline

bottom water in the Eocene and warm saline deep water with

an underlying psychrosphere [Kennett and Stott, 1990].

The oceanic record of carbon isotopes has shown that

changes have occurred more uniformly, with less geographic

variability. The changes have been inferred to be overall

reservoir changes rather than basin-to-basin

fractionation. The low paleogeographic variation has been

suggested to be due to low productivity, especially during

the Oligocene [Miller, 1992; Wright and Miller, 1993].

The balance between continental weathering of 87Sr-

rich rocks and hydrothermal weathering of 86Sr-rich

oceanic crustal rocks controls the marine 87Sr/86Sr ratio

preserved in foraminiferal shells. The middle to late

Eocene 87Sr/86Sr ratio was held low at a near constant

level, indicating a stable continental weathering to

hydrothermal weathering ratio, with domination by the










hydrothermal component. A transition to steadily

increasing conditions in the late Eocene has been used to

infer increased continental weathering rates, either

through increased mountain building [Richter et al., 1992]

or increasing glaciation [Oslick et al., 1994]. Not

incidentally, the steadily increasing 87Sr/86Sr values

also allow improvements in Oligocene stratigraphic

resolution, as biostratigraphic events can be dated more

precisely even in the absence of magnetostratigraphy, and

as the ratio itself can provide a unique date once the

curve is correlated to the Geomagnetic Polarity Time Scale

(GPTS).

This dissertation is composed of four chapters, which

serve to examine aspects of late Paleogene

paleoceanography. It is concentrated in the South

Atlantic, but many of its conclusions may be extrapolated

world-wide.

Since the research in this dissertation refers back

to several older data sets, most dated magnetostratigraph-

ically, the first chapter presents a compilation and

correlation of the 12 comprehensive GPTSs developed since

1968. The correlations enable conversion of dates

generated under older time scales to newer ones. The

correlation table has been placed on-line at the National

Geophysical Data Center to enable other researchers to

have easy access to it.


,,









Strontium isotope stratigraphy from the middle Eocene

through the Oligocene is the topic of Chapter 2. A well

dated, nearly continuously deposited and recovered section

from Ocean Drilling Program (ODP) Hole 689B, a key high

latitude site, was used to generate a fairly high

resolution (166 ky/sample) 87Sr/86Sr record. This record

was used in two ways. First, the 87Sr/86Sr record was

used to form a refined strontium isotope stratigraphy from

the mid Eocene to the end of the Oligocene. This will

allow further researchers to correlate biostratigraphic

markers to the GPTS. Second, the record forms the basis

for modelling experiments to determine the tectonic

changes controlling the marine 87Sr/86Sr ratio. Since

both hydrothermal activity [Edmond et al., 1979; Owen and

Rea, 1985; Lyle et al., 1987] and mountain building [Raymo

and Ruddiman, 1992] can have implications for affecting

the Earth's climatic history, it is important to

understand what these changes are.

The third chapter uses 6180 and 613C to examine deep

circulation changes in the Antarctic and South Atlantic

Oceans, especially with respect to Warm Saline Deep Water.

It uses pre-existing data from ODP Sites 689 and 690

[Kennett and Stott, 1990] with new data from ODP Sites 699

and 703 to examine the vertical 6180 and 613C distribution

in this region during the late Paleogene, and examines the

changes in circulation which may have taken place during


__- J A









those times, the loci of deep- and bottom-water formation,

and the influence of glaciation.

The final chapter examines the genesis of

Braarudosphaera oozes in the South Atlantic by examining

613C and 6180 records from mid-South Atlantic Deep Sea

Drilling Project sections. These data are analyzed in

terms of their possible relationships to vital effects,

productivity, shallow- and intermediate-depth circulation,

and how the circulation may have been affected by

Oligocene glaciation.

Collectively, these studies have two purposes.

First, they act to improve stratigraphic correlation and

resolution in marine sections of the late Paleogene.

Second, these studies are intended to help delineate

circulation in shallow, intermediate, and deep parts of

the South Atlantic, and to help us understand how this

circulation affected and was affected by changes in the

Earth's paleoclimatic, paleogeographic, paleoceanographic,

and paleotectonic regimes.










TABLE OF CONTENTS
page
ACKNOWLEDGMENTS .. ................................... iii
PREFACE ... .... .............. ....... ........ .. .... .v
ABSTRACT ...............................................xii
CHAPTERS
1 THE CORRELATION OF GEOMAGNETIC
POLARITY TIME SCALES .............. .......1

Introduction....................................
Calibration of the Time Scales...............5
Correlation Between Time Scales............... 13
Discussion................... ..... ...... ..... 21
Conclusion ................ .................. 26

2 CONTROLS ON THE 87Sr/86Sr COMPOSITION OF
SEAWATER FROM THE MIDDLE EOCENE TO OLIGOCENE:
HOLE 689B, MAUD RISE, ANTARCTICA .............. 28

Introduction......... ...................... 28
Methods And Materials........................ .33
Results....o................... ................ 43
Discussion........... .......... ..... ... .... 50
Conclusion.............. ..................... 92

3 LATE EOCENE TO OLIGOCENE VERTICAL OXYGEN
ISOTOPIC GRADIENTS IN THE SOUTH ATLANTIC:
IMPLICATIONS FOR WARM SALINE DEEP WATER.......95

Introduction................................95
Methods ....... ...... .......................... 99
Results..................................... 113
Discussion.................. ................ 121
Conclusion................................. 151

4 OLIGOCENE BRAARUDOSPHAERA OOZES AND CHALKS:
GENESIS AND PALEOCEANOGRAPHIC IMPLICATIONS
FROM ISOTOPIC ANALYSES OF FINE FRACTION AND
FORAMINIFERA .............................. 154

Introduction... ... .......................... 154
Methods ............................ ......... 166
Results .... .................................. 171
Discussion....... -o.................. ........ .. 204
Conclusion........... .... ................... 220
APPENDICES
A LARGE DATA TABLES... ........................ 223

1 Correlated Geomagnetic Polarity
Time Scales from 1968 to 1994.............224









2 Sr, 0, C Isotope and Sr/Ca Data,
ODP Hole 689B ........................... 239
3 Oxygen and carbon isotope analyses of
Cibicidoides spp. for ODP Hole 703A...... 253
4 Oxygen and carbon isotope analyses of
Cibicidoides spp. for ODP Hole 699A...... 262
5 Oxygen and carbon isotope data from
DSDP Site 363, cores 5, 6, and 7.........269
B DETAILED METHODS AND DERIVATIONS............. 276

1 Instructions for obtaining and using
on-line data tables...................... 277
2 Strontium extraction method .............. 280
3 Operating instructions for thermal
ionization mass spectrometer ............ .282
4 Preparation of foraminifera for
oxygen and carbon isotopic analysis......287
5 Derivation of equations ................. 291
REFERENCES ................... ......................... 294
BIOGRAPHICAL SKETCH ..............................o...... 315









Abstract of Dissertation Presented to the Graduate School of
the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

PALEOCEANOGRAPHY OF THE SOUTH ATLANTIC OCEAN
FROM THE MIDDLE EOCENE THROUGH THE OLIGOCENE

By

GREGORY A. MEAD

May, 1995

Chairman: Dr. David A. Hodell
Major Department: Geology

In middle Eocene through Oligocene time, the Earth

experienced major changes, including major plate tectonic

events and the transition from relatively warm conditions to

polar ice and high planetary thermal gradients. These

paleoclimatic and tectonic events led to changes in the

isotope geochemistry of the oceans, the basis for this

dissertation.

Marine microfossil 6180 evidence suggests that the

warmest temperatures of the Cenozoic were during the early

Eocene. Planktic and deep-sea benthic foraminiferal 6180

data suggest that temperatures began to decrease from that

time in a series of rapid changes separated by periods of

relative stasis. Despite these Eocene climatic changes, the

marine 87Sr/86Sr ratio stayed relatively constant, implying

that the balance between continental erosion and

hydrothermal input to the oceans was unchanging.


* *









During the Eocene, 6180 data from Ocean Drilling

Program (ODP) Sites 689 and 690 show a reversal in the

normal temperature gradient, interpreted as mid-water warm,

salty deep waters underlying cooler, fresher, shallow

waters. Additional data from ODP Sites 699 and 703 show

that at least by the late middle Eocene, this warm, saline

deep water mass extended to abyssal depths.

Strontium isotopic data from ODP Site 689 show that in

the late Eocene (about 35.5 Ma), a profound, sudden change

in the crust's geochemical balance occurred as marine
87Sr/86Sr ratios began to rise rapidly, probably as

hydrothermal alteration decreased relative to continental

weathering. The timing of this event suggests that it was

not due to increased continental weathering due to either

glaciation or uplift. In the earliest Oligocene (about 33.5

Ma), marine 6180 values increased by about 10/oo at all

paleodepths in the deep South Atlantic, indicating an

increase in continental ice volume and/or lower temperature.

This change, estimated to be due 35% to increased Antarctic

ice volume and 65% to temperature decrease, left the ocean's

vertical temperature structure unchanged. Oxygen isotopic

evidence suggests that at this time, glacial ice became

semi-permanent on Antarctica, waxing and waning through the

mid-Oligocene and beyond.









During the Oligocene, repeated Braarudosphaera oozes

were deposited. These unusual deposits may have resulted

from upwelling of cold, low-salinity, nutrient-rich water

originating from melting of an Antarctic ice shelf during

glacial/interglacial transitions.














CHAPTER 1


THE CORRELATION OF GEOMAGNETIC POLARITY TIME SCALES


Introduction



Knowledge in geology advances with time, just as it

does with all sciences. However, geology differs from other

sciences in that the baseline by which all other data are

calibrated, the geologic time scale, changes as the

knowledge changes. In most cases, the ordering of events in

the geologic time scale does not change. However, spacing

between events does change. This causes our understanding

of rate-dependent geologic processes to change (e.g.,

sedimentation rates, frequency of glaciations, tectonic

plate motions, or evolutionary rates) [Kominz, 1984; Delaney

and Boyle, 1988].

The geologic time scale, especially among marine

geologists, is calibrated to the marine magnetic anomaly

pattern (the Geomagnetic Polarity Time Scale, GPTS) for the

past 100 Ma or so. Since the advent of plate tectonic

theory (Vine and Matthews, 1963] when it was realized that

marine magnetic anomalies (MMA) are created during the

seafloor spreading process, several different GPTSs based on

the MMA pattern have been developed, beginning with the










pioneering time scale of Heirtzler et al. [1968]. A GPTS is

based on the concept that seafloor anomalies record the

history of geomagnetic polarity reversals. Time scales

based on MMAs differ in a number of ways, including the

particular anomaly pattern used, the method of dating tie-

points within the pattern, the method used in dating

anomalies between tie-points, the radiometric decay constant

used for radiometric dates, and the particular anomalies

included or excluded from the time scale.

Since 1968, at least 12 different, comprehensive,

geomagnetic polarity time scales covering the last 80 or 100

Ma (back to Anomalies 33 or 34) have been constructed, along

with a host of partial time scales which cover smaller parts

of the polarity record. As a result, comparing studies

which use different time scales can be difficult, because

events may be dated at different times which may be

simultaneous, or events which appear to be simultaneous may

in fact be diachronous.

Unfortunately, the problem is worsened by the fact that

in very few of the revisions are direct comparisons made

between the revised and older time scales. Usually, the

ages of periods of normal polarity are given, often with

little or no identification. Since revisions often distort

some of the polarity intervals, and insert or delete others,

it becomes a non-trivial matter to determine which polarity

interval of one time scale is correlative to a polarity

interval of another time scale.










This paper attempts to give a comprehensive correlation

of 12 geomagnetic polarity time scales over the past 26

years. It does not in general discuss the relative merits

of each time scale. Comparisons of preceding time scales

generally can be found in each successive revision, but

especially in Ness et al. [1981] and Harland et al. [1990].

This paper instead is meant to be a resource by which

investigators can calibrate studies which use different time

scales, or to bring older studies up to date. It presents a

brief description of the method used for calibration of each

time scale to geochronology, and a graphic and tabular

correlation of the time scales to one another. The means by

which time scales are correlated are discussed, and

ambiguous regions are noted. In addition, during the

preparation of this paper, a total of 6 errors were

identified in 5 published time scales. They are listed in

Table 1-1 along with the way that the errors were

identified. In most cases, it is unlikely that researchers

re-entering the time scales for their own uses would not

notice them, but they are presented here for the sake of

completeness. Finally, to help others avoid the onerous

task of typing in each of the time scales for their own use,

a copy of the table which correlates the time scales will be

placed on-line (see Appendix Al) so that it may be

downloaded for the use of other researchers (see Appendix Bl

for instructions).














TABLE 1-1

List of typographical errors in published
Geomagnetic Polarity Time Scales


Datum

Within Ir

Top 5111

Top 12





Within 17

Within 17

Top of

Jaramillo


Incorrect

Date

1.93

12.49

33.32





38.57

38.64

0.92


Correct

Date

0.93

16.49

33.22





39.57

39.64

0.90


How Error Was

Recognized

Out of order

Out of order

Not correctly

converted from

LKC77.

Out of order

Out of order

Supposed to be

identical with

MD79.


Time

Scale

HDHPL68

TM76

MD79





NLC81

NLC81

GTS82









Calibration of the Time Scales


The Heirtzler et al. [1968] (HDHPL68) time scale was

based on a single magnetic anomaly record from the South

Atlantic, the V-20SA magnetic profile from about 300S, on

the western side of the Mid Atlantic Ridge. Heirtzler et

al. (1968] examined several magnetic profiles from the

Atlantic, Pacific, Indian, and Antarctic Oceans and

concluded that the South Atlantic was most likely to have

undergone spreading at a constant rate. This anomaly

profile was to become the standard, with modifications, for

most subsequent time scales. Only two dates for the profile

were used: 0 Ma for the center of the Mid Atlantic Ridge,

and 3.35 Ma for the Gauss/Gilbert boundary [Doell et al.,

1966]. This resulted in a calculation of 1.9 cm/yr as a

spreading rate, which they extrapolated to the entire

anomaly pattern. The result was an age of 76.33 Ma for the

top of Anomaly 32, and a slightly adjusted age of 3.37 Ma

for the Gauss/Gilbert boundary. The ages of normal polarity

intervals are presented in their Table 1 without

identification, although the major anomalies are identified

graphically in their Figure 3 (which in some cases differs

from the table; see the last two normal polarity intervals

in their table and figure, for instance). Despite the

extrapolation of constant spreading rates to more than 22

times the length of the dated portion of the anomaly

pattern, the HDHPL68 time scale has remained within a few

percent of succeeding time scales.










Tarling and Mitchell [1976] (TM76) revised the HDHPL68

time scale in two major ways. First, they introduced

additional radiometric calibration points from sedimentary

sequences for the anomaly pattern. They dated the top of

Anomaly 6 at 19.5 Ma, with Anomalies 13, 24, and 30 centered

at 35, 48, and 64-65 Ma, respectively. They then linearly

corrected all reversal boundaries to these tie points.

Secondly, they inserted additional polarity intervals into

Anomaly 2 based on reversals found in deep sea sections, and

into Anomaly 5 based on reversals found in deep sea sections

and in MMA patterns by Blakely [1974]. They designated

additional polarity intervals with superscript roman

numerals such as Anomalies 21, 51V, etc. As with the

HDHPL68 time scale, ages of normal polarity intervals are

not identified in the tabular listing of ages, only

graphically.

A widely used major revision of the GPTS was developed

by LaBrecque et al. [1977] (LKC77). They started with the

original HDHPL68 time scale, and added an additional

calibration point at the base of Anomaly 29 of 64.9 Ma.

This was based on an age of 65 Ma for the

Cretaceous/Tertiary (K/T) boundary, which was found just

below Anomaly 29 in a section at Gubbio, Italy (Lowrie et

al., 1976]. They also used a more recent [Dalrymple, 1972]

age for the base of the Gauss (Anomaly 2A) of 3.32 Ma,

adjusted the length of Anomaly 4A, included the short

reversed intervals in Anomaly 5 of Blakely [1974], inserted










an extra reversal in Anomalies 23 and 24, spliced in a

revised sequence from Anomalies 29 to 34, and most

importantly, removed Anomaly 14 from the GPTS, which was not

found in most anomaly patterns. In addition, LKC77 used

radiometric ages directly for anomalies younger than 2A

[Klitgord et al., 1975), and extrapolated the GPTS through

Anomaly 34, to 108.17 Ma. The new time scale was presented

with brief identification of periods of normal polarity in

both tabular and graphic form. This time scale re-appeared

in the Larson et al. [1982] time scale, with extensions to

the Mesozoic by Larson and Hilde [1975] and Van Hinte

(1976a,b].

Later in 1977, new decay constants were adopted for

K/Ar dating [Steiger and Jager, 1977]. All subsequent time

scales used the new decay constants in calculating K/Ar

dates. To take into account the new decay constants,

Mankinen and Dalrymple [1979] (MD79) revised the LKC77 time

scale by recalculating each of the LKC77 time scale datums

to new values. They presented their results in a side-by-

side comparison to the LKC77 time scale, along with anomaly

identifications, simplifying conversion from one to the

other time scale. MD79 further revised the late Neogene (5-

0 Ma) based on additional radiometric dates later in their

paper (their Figure 3); those dates are presented in the 5-0

Ma portion of this compilation.

A comprehensive review of previous time scales, along

with the development of a new one, was presented by Ness et










al. [1980] (NLC80). The new NLC80 time scale was based in

general on the LKC77 time scale corrected to the new K/Ar

decay constants, but used segments from several other time

scales to adjust the length of anomalies, and to insert a

few new reversals. Yet another terminology was used in the

NLC80 time scale, with decimal numbers, and then primes, to

designate successively smaller intervals after the main

anomaly numbers. A total of 4 calibration points, plus the

present, were taken from a wide variety of sources, and

dates were interpolated between them. Anomaly 2.3'(o) (for

"older end") was fixed at 3.40 Ma, with dates extrapolated

to Anomaly 3.4; Anomaly 4.1'(y) was fixed at 7.81 Ma;

Anomaly 24(o) was fixed at the Paleocene/Eocene boundary

(54.9 Ma); and the K/T boundary, just below Anomaly 29(0),

was fixed at 66.7 Ma. The study compared in graphical form

the dated anomaly sequences and time scales used in all the

previous studies discussed in the present paper, as well as

several more fragmentary sequences. A synthetic anomaly

profile was generated which facilitates comparison to other

anomaly sequences. Major anomalies and correlations between

the various time scales were made (and in general are

followed in this paper) although a few ambiguities remained.

Dates attached to the anomaly patterns were unfortunately

printed in a font so small as to be nearly unreadable.

Another time scale appeared just a year later by Lowrie

and Alvarez [1981] (LA81). This time scale used a total of

11 biostratigraphically defined calibration points (geologic










period boundaries), from Italian pelagic sections (as well

as the 0 Ma point), to modify the LKC77 time scale,

corrected for new K/Ar decay constants. The

biostratigraphic correlations allowed them to adjust the

positions of the anomaly boundaries to agree with additional

radiometric dates. As in most other time scales, the ages

between calibration points were determined by linear

interpolation. Most of the anomaly boundaries were

identified in the table of ages provided.

A geologic time scale covering the entire Phanerozoic

was published in 1982 by Harland et al. (GTS82). It dated

MMA from the present back into the Mesozoic, although for

the sake of consistency it is reported here only through

Anomaly 34. From 5-0 Ma, radiometric dates were used to fix

reversal boundaries. From 83-5 Ma, GTS82 is based on MMA,

modified from LA81. Two modifications were made. First,

the Paleocene and Eocene boundary calibration points were

removed because they implied improbably rapid spreading rate

changes. Points at the beginning and ends of the Paleocene

and Eocene were kept, however. The second modification was

to place the K/T boundary at 65 Ma; some other adjustments

were made to reduce apparent spreading rate changes as well.

Overall, NLC80 was stretched between the new calibration

points from 10.3 to 83 Ma, NLC80 was used directly for 3.4

to 10.3 Ma, and MD79 was used directly from 0-3.4 Ma.

An extremely widely used time scale was next published

by Berggren et al. [1985a] (BKFV85). The same time scale










has appeared as Berggren et al. [1985b,c] separately for the

Paleogene and Neogene, and, with extensions, in Kent and

Gradstein [1986] (back to the Jurassic). The major advance

of this time scale was that it integrated biochronologic,

magnetochronologic, and radiochronologic data into a single

time scale. The BKFV85 time scale used fewer calibration

points than some of the recently preceding time scales, only

6 (plus the present). The younger ends of Anomalies 2A, 5,

12, 13, 21, and 34 were fixed at 3.40, 8.87, 32.4, 34.6,

49.5, and 84.0 Ma, respectively, by reference to a host of

different studies. Three best-fit line segments were

constructed from these data, from 0 to 10.42 Ma (the base of

Anomaly 5), from 10.42 to 56.14 Ma (the base of Anomaly 24),

and 56.14 to 84 Ma (the top of Anomaly 34). Ages for the

other paleomagnetic boundaries were interpolated from LKC77

and reported in tabular and graphical form, with most

intervals of normal polarity identified. Perhaps most

importantly, BKFV85 also correlated marine calcareous

planktonic biostratigraphic zonations to the GPTS for the

entire Cenozoic. BKFV85 used the chron nomenclature

introduced by LaBrecque et al. [1983], with a "C" prefix

designating the time during which a given normal and

reversed anomaly couplet was created, and a "N" or "R"

designating the polarity of the Earth's field at the time

within the chron.

Haq et al. [1988] (HHV88) recognized that using a

single MMA pattern or a patchwork of MMA patterns strung










together suffers from the problem that spreading rate

changes may not be taken into account. In addition, errors

in radiometric dates of calibration points may introduce

errors in calculated spreading rates. To overcome these

uncertainties, HHV88 plotted numerous radiometric dates

against MMA patterns from 3 ocean basins the North and

South Pacific and the South Atlantic. They calibrated the

MMA to best-fit age-to-length lines, and then stacked (i.e.,

averaged) the ages of the tops of the major anomalies from

each basin to arrive at their time scale. The time scale

has a coarse resolution, with only the tops of Anomalies 1

to 32, 5A to 5E, and 6A to 6C listed.

Harland et al. [1990] (GTS89) modified the original

HDHPL68 anomaly spacing by splicing in parts from the

several different sources, and adjusting spacing using MMA

pattern segments from fast spreading ridges. To date this

modified HDHPL68 record, GTS89 used numerous radiometric

dates to define 3 best-fit line segments, from 0 to 3.4 Ma,

9 to about 50 Ma, and about 50 to about 85 Ma. No data was

used for 3.4 to 9 Ma. These 3 line segments were used to

constrain and convert the modified MMA pattern of HDHPL68,

resulting in the new time scale. No correlation to previous

time scales is shown. However, through an intermediate step

(2 tables), GTS89 does provide an exact match to HDHPL68,

except for Anomalies 4A through 5, which are "very

different."









Cande and Kent [1992] (CK92) used two new approaches in

constructing their new time scale. One of the techniques

was the creation of a synthetic MMA pattern in order to

remove the problems associated with using a single MMA

pattern (V-20SA) as a standard. Cande and Kent [1992]

stacked different segments of MMA patterns from the South

Atlantic to reduce the noise and error associated with a

single MMA pattern, using from 5 to 9 different MMA patterns

for every part of the time scale from Anomaly 1 to 34. They

next assumed that spreading rates were not necessarily

constant but instead smoothly varying, and fitted 9 points

(finite rotation poles from Cande et al. (1988]) plus the

ridge axis with a cubic spline function. Their aim was to

reduce the apparent spreading rate changes in the synthetic

MMA record. They also identified numerous "tiny wiggles"

(their Table 7) which they did not include in the main time

scale (their Table 6). Some of the tiny wiggles they

identified (C2r.2r-l, C5n.2n-1 to -3, C5Dr-l, and C13nl) are

included in Appendix Al of this work, because they are

commonly observed in previous time scales. The second

technique use by CK92 was to use two astronomically tuned

dates (0.780 Ma for the Bruhnes/Matuyama boundary and 2.600

Ma for the Matuyama/Gauss boundary) of Shackleton et al.

[1990] and Hilgen [1991]. These dates were obtained by

assuming that Milankovitch periods seen in geologic records

are known more precisely and accurately than radiometric

dates over the same intervals, and counting back to magnetic









reversals in these records. Interestingly, tuned ages used

in this time scale and the Cande and Kent [1995, in press]

time scale are all older than radiometric ages by 5 to 10%.

It has been suggested that this is due to an error in the

decay constant of Steiger and Jager [1977] [Shackleton et

al., 1990] or to diffusional loss of radiogenic Ar [Hilgen

1991].

Cande and Kent [1995, in press] (CK95) is modified in

two ways from the CK92 timescale. The later part uses a

total of 19 additional tie-points added according to the

astronomical time scales of Hilgen [1991] from the base of

C3n.4n to the base of C2An.ln and Shackleton et al. [1990]

from the top of C2An.ln to the present. The early part was

adjusted so that the Cretaceous/Tertiary (K/T) boundary

falls at 65 Ma, and another cubic spline used to fit the

data from the top of C34n to the base of C3n.4n as with

CK92. Because of the use of different cubic splines in CK92

and CK95, the difference between the two time scales varies

sinusoidally between 0.2 my except near the K/T boundary,

when CK95 is over 1 my younger than CK92. This time scale

will be incorporated into a comprehensive revision [Berggren

et al., in preparation] of Cenozoic microfossil datums

[Kent, personal communication].


Correlation Between Time Scales


Ideally, every time a new time scale was published,

each period of normal and/or reversed polarity would be









completely identified so as to facilitate comparison to

older and newer time scales. Failing that, it would be

helpful if the polarity intervals in each time scale were of

equal number and of roughly equal relative size. Both of

these possibilities are rarely seen.

Another option would be to have each time scale

correlated by referring back to the original anomaly pattern

each is based upon. Unfortunately, the anomaly patterns are

rarely published, and are presented in only 3 of the papers

discussed here (HDHPL68, NLC80, and CK92). It is clear that

these papers, each published 12 years apart, are easily

correlated by comparison of the anomaly patterns (Figure

1-1), despite the different methods used to construct them.

In lieu of the above choices, pattern matching must be

done, similar to the way magnetostratigraphic records in

sedimentary sequences are matched to the GPTS. The

comparison is made easier because, at least in this case,

the entire record is in fact present. This has been done

graphically in two of the papers discussed, NLC80 and CK92.

NLC80 presented their new time scale correlated to other

partial and comprehensive time scales (their Figure 2), with

a few lines of correlation, and much of their technique is

followed here in Figure 1-2. Few lines of correlation were

presented, leaving a great deal of ambiguity in the

correlations. CK92 used a different method, displaying each

polarity record as a series of lines shifting back and forth










C~j 0
OJ 0

Sz


10
aO) )0


o o
4- Q) 0




0 )0 )4 0




U Q) I









I ; "0 4-
: O -4 .0J0
(00 X 04)r.



S0 Q
O C a) U) C-.t U)
X --4 0 (0 V4
~ 4J < r 44






0 ) 40 -)
0 0 a 4) a
0 Q) 0
co0 C4 Mr. U
w 1E (0 4
>. 0 Q)
B41 Sr-4 U) U4
41 0 03

0 T3 01 0

e O U3 0)
C w 411 )
0 0 0Q)
In "-4
1 Q) r. 41 *
4.1C -4)I 44 0

SE-4 r 0 r.
c0 PO M r 0
w, r- 04m
M -r-< 4-4 V w








-'Z 4J w U -) w
- Q) ) PIc







) CO Q) 44 4- U
Q) 4C -4-


(0 a U 0 =














~U Q) U)UM0
3 ) r.) J Q) 4


4C Q 44 U W 4)
















.r4 m 0 w.1-4 r.
44 04 U 44 4J V 4-)




















Figure 1-2: Correlated paleomagnetic intervals from 12
time scales. Black indicates an interval of normal
polarity; white indicates an interval of reversed
polarity. Lines indicate tops and bottoms of equivalent
polarity intervals. In the region from 34 to 64 Ma, some
lines are dotted or dashed. The differences are merely to
help guide the eye. The abbreviated headings are
explained in the text.












(D
oI r- m0 S ZOCX
OMa




5










15




20Ma




25


40Ma











Figure 1-2--continued


m
to

Q- 0 r 0 Z
r i-_I
T~_Z


in
- CO >*0 C I~l
O (A U. > (A 0) 0M
i0 M T 0 U U


40Mc






45






50


55






60Ma






65


80Mo










Figure 1-2--continued


0o
_ S r-- o < f =s c I
ri- 3 : Z 0 O 0D I 0









85





-90





95





100 Ma


105


1 10


1 15









from the right (Normal polarity) to the left (Reversed

polarity) for their time scale and 5 preceding time scales.

In this case, no lines of correlation were used, and

the figure was published at such a small scale that fine-

scale correlations cannot be made.

A hierarchy of methods is used in this paper to

correlate the time scales. First and foremost, the authors'

designation of a polarity interval, if present, was used to

make a first-order correlation. Although not all time

scales were published with identifying labels in the table

of ages, generally accompanying figures could be used for

the same purpose. Second, in the absence of sufficient

labeling (and it should be noted that only GTS82, CK92, and

CK95 time scales had sufficient unique identifiers for all

periods of normal and/or reversed polarity), the relative

lengths of the polarity intervals (i.e., pattern matching)

were used to make correlations. This method succeeded in

making almost all of the detailed correlations left after

the first-order correlations. The ambiguities that remain

generally result from multiple short reversed or normal

events which are not present in all time scales, and whose

relative position cannot be used to determine which of the

multiple events may or may not be present in a particular

time scale.

The result is Appendix Al and Figure 1-2. Within

Figure 1-2, lines of correlation are drawn for the tops of

each major polarity chron, with additional lines drawn to










clear up any residual ambiguities. Asterisks (*) indicate

the ambiguities which remain.

Discussion


A GPTS is usually used to date stratigraphic sequences.

An age model for a sequence is developed using identifiable

magnetostratigraphic events within the GPTS, or

biostratigraphic datums that are calibrated to the GPTS.

Sample ages are then calculated by linear interpolation from

the position of samples within the stratigraphic sequence.

Linear interpolation will also generally be used to make the

conversion between one time scale and another. Two factors

must be taken into account when converting between time

scales, however.

First, in constructing a conversion table of two GPTSs

from Appendix Al, only dates (i.e., paleomagnetic chrons)

common to both time scales can be used. For example, if

conversion is made between the BKFV85 and CK92 time scales,

Chron C2r.2r-l is present only in CK92, not in BKFV85, and

must be left out of a conversion table. Similarly, Chron

C23R-1, found in NLC80, GTS82, and GTS89, is not present in

either BKFV85 or CK92. Obviously, Chron C23R-1 cannot be

used in the conversion table, either. Therefore, Appendix

Al must be edited to remove the unusable chron intervals.

Second, even an edited Appendix Al cannot in general be

used to convert ages from one time scale to another directly

from the nearest two dates common to both time scales. The










reason is that the age model used to date the samples,

usually generated from paleomagnetic boundaries, usually

only uses selected polarity transitions in a time scale.

Since the relationship between time scales is usually not

linear, the age model must be converted, and that converted

age model used to determine ages directly from stratigraphic

sequences. For example, suppose columns 2 and 3 of Table 1-

2 show the depths and ages of several samples in a given

stratigraphic sequence, using the BKFV85 time scale. If the

datums listed in column 1 are known, then the ages of the

samples can be easily converted by using the relationship of

BKFV85 to CK92 for all points, and are listed in column 4.

But if only 2 datums, the top of C12 and the top of C20, are

known (e.g., because of coring disturbance or lithologic

problems in the section between those datums), then the age

model used to calculate ages from the depths would consist

of only 2 points, 32.46 Ma at 12.46 m, and 44.66 Ma at

24.66 m, and ages calculated by linear interpolation between

those points. Converting those dates to 30.452 Ma and

42.629 Ma of CK92, respectively, and using linear regression

to calculate the new dates would result in the ages given in

column 5, which differ from the column 4 ages by up to

0.844 Ma (Column 6), a potentially significant error. The

reason that the differences exist is that the relationship

between BKFV85 and CK92 is not linear in this interval

(Figure 1-3).























140-
39- Linear Trend
S38

E 37
36-
v 35-
34-
33-
32
31-
30
30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45
BKFV85 Time Scale (Ma)

Figure 1-3: The CK92 time scale plotted against the BKFV85
time scale from 30-45 Ma. Note that the relationship is
not linear.









For a similar reason, if biostratigraphic datums are

being used that have originally been dated

paleomagnetically, then to re-calculate the ages of those

datums, the original citation must be re-examined, and the

ages of the biostratigraphic datums re-calculated based on

the age model used in the original citation. Then the new

ages of the biostratigraphic datums can be used in a given

age model to calculate ages at a particular site.

Therefore, the following steps must be taken in order

to convert a series of dated samples from one GPTS to

another. First, Appendix Al of this paper must be edited to

remove polarity boundaries not common to both time scales in

question. Second, using linear interpolation, the age model

originally used to calculate ages for the samples must be

converted from the original time scale to the new one.

Finally, the newly converted age model may be used to

calculate ages of the samples.

The set of steps listed above raises two additional

points. In many papers, ages are given without reference to

a particular time scale and/or the age model used to

calculate those ages. Without this information, it is

impossible to evaluate the synchroneity or diachroneity of

events within one study with those of another. Therefore,

in any geologic study, both the time scale and the age model

used should be explicitly stated.

It is of course highly unlikely that the CK92 time

scale will be the final word in GPTSs. The utility of










TABLE 1-2

Example of a conversion from one time scale to another


5

Age-Model

Age

(CK92.


1





Datum

top C12

top C13

top C15

top C16

top C17

top C18

top C19

top C20


Difference


2

Sample

Depth

(m)

12.46

15.29

17.24

18.10

19.53

21.29

23.60

24.66


3



Age

(BKFV85)

32.46

35.29

37.24

38.10

39.53

41.29

43.60

44.66


4

Direct

Age

(CK92)

30.452

33.050

34.669

35.368

36.665

38.500

41.353

42.629


30.452

33.277

35.223

36.081

37.509

39.265

41.571

42.629


0.000

0.227

0.554

0.713

0.844

0.765

0.218

0.000


DiffPrAnop










future time scale revisions would be greatly enhanced if

they included a detailed point-by-point comparison to at

least one previous version of the GPTS, probably the most

recent preceding one.

Conclusion


Since the development of the first comprehensive GPTS

in 1968, numerous time scales have been developed based on

the sequence of reversals of the Earth's magnetic field.

Twelve GPTSs are examined in this paper and correlated to

each other as well as possible. The correlations are

presented graphically and in table form. Some ambiguities

remain, but the correlations should enable conversion from

one time scale to another.

Direct conversion from one time scale to another cannot

in general be used to re-calculate ages within a

stratigraphic section. Instead, the original age model used

to calculate ages must be converted, and the new age model

must then be used to calculate ages within the stratigraphic

section. The reason is that the relationship between GPTSs

is rarely linear, and it would be very unusual for every

reversal boundary to be identified in any given

stratigraphic section.

The need for this paper results from the fact that most

revisions of the GPTS do not include detailed, useful

comparisons to previous versions of the GPTS. Users of

future revisions would be helped if the authors of those





27



revisions would include a detailed comparison to at least

one pre-existing version.














CHAPTER 2


CONTROLS ON THE 87Sr/86Sr COMPOSITION OF SEAWATER FROM THE
MIDDLE EOCENE TO OLIGOCENE: HOLE 689B, MAUD RISE, ANTARCTICA

Introduction



The time from the middle Eocene through Oligocene saw

major changes in the Earth's climatic and biotic realms.

During this time, the transition from "Greenhouse" to

"Icehouse" conditions occurred [Miller et al., 1991a,b].

For example, ice rafted debris (IRD) from Antarctic glaciers

first appeared on the Kerguelen Plateau in the earliest

Oligocene at about 35.8 Ma (Berggren et al., 1985 time

scale), coeval with the world wide increase in 6180 values

of marine carbonates [Zachos et al., 1992b; Breza and Wise,

1992; Barrera and Huber, 1993]. By the late Eocene, the

planetary thermal gradient increased [Keigwin and Corliss,

1986; Zachos et al., 1992a]. Major plate tectonic events in

the middle Eocene to Oligocene include the opening of the

Drake Passage [Barker and Burrell, 1977], the progressive

closure of the Tethys and uplift of the Alps [Dercourt et

al., 1986], collision of the Indian subcontinent with

Eurasia and subsequent uplift of the Himalayan-Tibetan

Plateau [Harrison et al., 1992], and a middle Eocene change









in the direction of motion of the Pacific plate [Van Andel

et al., 1975; Clague et al., 1975; Kennett et al., 1985].

The ratio of 87Sr to 86Sr in the oceans is a function

of the balance between continental weathering via riverine

input, dissolution of marine carbonates, and hydrothermal

exchange with mid-ocean ridge basalts [Palmer and

Elderfield, 1985]. When biogenic marine carbonates

precipitate, the 87Sr/86Sr ratio of ocean water is preserved

with no fractionation as Sr substitutes diadochically for

Ca. The Sr isotopic composition of the oceans is constant

at any given time due to the long residence time of Sr in

the oceans (about 2.5 Ma [Hodell et al., 1990]). The
87Sr/86Sr ratio thus has great potential as a stratigraphic

tool. Although the controls on the Sr budget of the ocean

are thought to be well understood [Palmer and Elderfield,

1985], the actual factors that influence changes in the

87Sr/86Sr seawater curve are often difficult to determine

[Hodell, 1994].

The ratio of 87Sr to 86Sr in marine carbonates has

changed irregularly through the Phanerozoic [Burke et al.,

1982, Hess et al., 1986], and increased, with some intervals

of stasis, throughout the Cenozoic [Palmer and Elderfield,

1985, Koepnick et al., 1985, Hess et al., 1986]. During the

Eocene, oceanic 87Sr/86Sr ratios were approximately

constant, although somewhat variable, and averaged from

about 0.7077 to 0.7078 [DePaolo and Ingram, 1985, Hess et

al., 1986]. In the late Eocene, oceanic 87Sr/86Sr ratios









began to increase, rising fairly linearly to about 0.7082 at

the end of the Oligocene with a slope reported at from

34x10'6/m.y. [Miller et al., 1988] to 29x10-6/m.y. [Hess et

al., 1989]. The transition to more rapidly increasing

Oligocene values is not well documented, however. Lower

resolution studies [Burke et al., 1982; Koepnick et al.,

1985; Palmer and Elderfield, 1985; Hess et al., 1986] place

the change between 40 and 35 Ma, according to the Berggren

et al. [1985] time scale. The most complete published
87Sr/86Sr records available for the Oligocene do not reach

the transition and/or are not paleomagnetically dated

[Miller et al., 1988; Hess et al., 1989]. In addition, Site

522 (analyzed by Miller et al. [1988]) is at least partially

recrystallized [Hess et al., 1989]. Finally, the resolution

of the two studies was not particularly high (about 0.4-0.5

m.y./sample), especially in light of recent studies that

suggest short term 87Sr/86Sr variation [Capo and DePaolo,

1990; Dia et al., 1992; Clemens et al., 1993].

The cause of the inflections in the 87Sr/86Sr curve in

the late Eocene is not known. There are a number of

tectonic and environmental factors that may have influenced

the input of Sr into the oceans near this time. Glaciation

occurred on Antarctica in the earliest Oligocene, which

could have increased the weathering rates on that continent

[Zachos et al., 1992b; in preparation]. Uplift of the

Himalayan-Tibetan Plateau following the collision of India

with Eurasia may have increased continental weathering rates









[Richter et al., 1992]. In addition, major sea level drops

in the middle Eocene, earliest, "mid", and late Oligocene

occurred [Haq et al., 1988], and the Carbonate Compensation

Depth (CCD) deepened near the Eocene/Oligocene boundary [Van

Andel et al., 1977], possibly reflecting a change in oceanic

circulation, shallow/deep calcium carbonate fractionation,

or continental weathering rates. Plate reorganization in

the Pacific may also have affected the hydrothermal

component and/or direction of plate convergence [Lyle et

al., 1987].

This paper presents the 87Sr/86Sr record from the

middle Eocene to Oligocene of ODP Hole 689B (Maud Rise, near

the coast of Antarctica; Figure 2-1). The purpose of this

study is (1) to refine the Sr isotope curve during the

middle Eocene-Oligocene, (2) to help calibrate Antarctic

biostratigraphy to the Geomagnetic Polarity Time Scale

(GPTS), and (3) to improve our understanding of geochemical

cycling of Sr during that time. In addition, additional

oxygen and carbon isotopes of Cibicidoides were analyzed in

order to cross calibrate the results from the work of

Kennett and Stott [1990] and Chapter 3, and to improve the

resolution of the 6180 and 613C records from Hole 689B.

This site contains a nearly continuous and well preserved

calcareous and siliceous biostratigraphic record of the

Eocene to Oligocene, from very high latitudes. As such, it

constitutes a vital tie-point for biostratigraphic


























































Figure 2-1: Location map, showing ODP Site 689 and nearby
Site 690.









correlation. It also is an important site for biogeographic

and isotopic paleoceanographic studies (Barker, Kennett et

al., 1988, 1990; Chapter 3 of this dissertation].


Methods And Materials


Stratigraphv


Cenozoic stratigraphy is currently in a state of

change, with the standard time scale of Berggren et al.

[1985] (BKFV85) being replaced by Cande and Kent (1992]

(CK92). Ages in this paper will be given using the newer

time scale, although the BKFV85 ages are reported in the

data tables as well as the newer CK92 ages. Where

appropriate, BKFV85 ages are given in parentheses.

The age model used is based on the magnetostratigraphy

of Spiep [1990] and the age model of Kennett and Stott

[1990], modified to the CK92 time scale (Table 2-1, Figure

2-2). Modifications have been made to these models in a few

instances discussed below.

First, an unconformity exists at 161.54 mbsf that was

not recognized by Kennett and Stott [1990] but was

identified by Spiep [1990]. Without this hiatus, it is

difficult to reconcile the pattern of normal and reversed

polarity events between about 135 and 190 mbsf. This

unconformity has been added, with the ages of the upper and

lower surfaces (defined by extrapolation of sedimentation










TABLE 2-1

Age Model of ODP Hole 689B


Preferred

Model

Datums


BKFV85

Age

(Ma)


CK92

Age

(Ma)


60.40, Top C5Dr-2 18.14 17.893

61.76, Base C5Dr-2 18.56 18.317

65.50, Hiatus, ----------------------------------------------


65.50, Hiatus, by extrap.*1

66.86, Hiatus, 5Base C6n


Hiatus, 2Top C7nlr

Top C7n-2

Base C8n

Top C9n

Base ClOn

Base Clln

Base C12n

Top C13n

Base C13n

Top C15n

Base C15n

Top C16n

Base C16n-1


20.05

20.45


25.60

25.67

27.74

28.15

30.33

32.06

32.90

35.29

35.87

37.24

37.68

38.10

38.79


19.770

20.162



24.772

24.826

26.533 SK

27.004 SK

28.716 SK

30.071 SK

30.915 SK

33.050 S

33.543 SK*3

34.669

34.959

35.368 K

35.977


Depth

(mbsf)


66.86,

67.11,

75.97,

79.46,

91.93,

103.37,

106.87,

116.71,

119.69,

124.09,

125.07,

128.32,

134.02,










TABLE 2-1--continued

Preferred BKFV85

Model Age

Datums (Ma)


135.75, Top C17n

152.73, Base C17n3

153.70, Top C18n

161.17, Top C18r-2

161.54, Hiatus, by extrap.*2


161.54,

163.16,

165.55,

171.11,

183.23,


Hiatus, by extrap.*l

Base C19n

Top C20n

Base C20n

Top C21n


39.53

41.11

41.29

42.73

42.80



43.66

44.06

44.66

46.17

48.75


36.665 K*4

38.183 SK

38.500 SK*5

40.221 S

40.312 S



41.387 S

41.617 SK

42.629 SK

43.868 S

46.284 S


S: Datum from Spiep [1990].
K: Datum from Kennett and Stott [1990].
*1: Age calculated by assuming constant sedi-
mentation rate from underlying section.
*2: Age calculated by assuming constant sedi-
mentation rate from overlying section.
*3: Datum listed by KS90 as Top Cl3n.
*4: Age listed by KS90 as 39.52 Ma.
*5: Age listed by KS90 as 41.21 Ma.


Depth

(mbsf)


CK92

Age

(Ma)













0)

0
w
Q4
0

0
0
00
.I

0

co
H
6
V




r.



CO



CO
Ia



0
(0

.V-4
0
4.)l
uP





3
41
V





4U
$4

4)








4J4)
Or
.r4













>4
0




10









$44
0













4.)-4
r4) t
0 O
kC
tp+
**-i a





ON to
BOC


4P.)
o co



1 0


$M4J
I 0

a


fa u































































in '^ CO CM i- O CD CO f^
CO CO CO CO CO cO h r'- r
oo0000000

dd6dddddd


T- cJ CO


T-

E E



T- Q)
-0










rates to the position of the unconformity) converted from

the BKFV85 to CK92 time scale.

Second, there is a disagreement between the age models

from 128.33 to 137.77 mbsf (lower upper Eocene). The series

of polarity intervals identified as Chrons C15n-l/2 to C15r-

2 by Spiep [1990] are identified as Chrons C16n-l to C16r-3

by Kennett and Stott [1990]. The CP15b/CP15a calcareous

nannofossil boundary (recognized at 131.7-132.7 mbsf by Wei

and Wise [1990]) is closely associated with the C15n-1/2 to

C15r-2 boundary [Berggren et al., 1985], and this controls

Spiep's [1990] position of that paleomagnetic boundary at

134.02 mbsf. However, Wei [1991] and Wei and Wise [1992]

showed that the first occurrence of Isthmolithus recurvus,

which defines the CP15b/CP15a boundary, occurs somewhat

earlier in the Southern Ocean, within C16. Using Spiep's

[1990] identification also introduces large distortions into

the sedimentation history, with sudden jumps in the

sedimentation rates within C15n and C16n. As a result, I

accept the chron assignments of Kennett and Stott [1990] in

this interval. This chron assignment also re-defines the

normally magnetized interval between 124.09 and 125.07 mbsf

from a normal within C13r-2 [Spiep, 1990] to C15n,

considerably improving the pattern match between the

magnetic stratigraphy and the seafloor magnetic anomaly

record.

Finally, two unconformities occur in the upper part of

the section: one at approximately 66.86 mbsf and the other










at about 65.5 mbsf. Only the lower one is recognized

paleomagnetically [Spie3, 1990]. The second falls within a

normally magnetized interval and was identified only through

the use of biostratigraphy [Gersonde and Burckle, 1990].

Neither hiatus is recognized by Kennett and Stott [1990].

However, both are clearly identified by the use of Sr

isotope stratigraphy (see RESULTS). The 87Sr/86Sr ratio of

sediments within the short normally-magnetized interval

between the unconformities (about 0.70843) indicates that

the age is approximately 20 Ma [Hodell and Woodruff, 1994],

assignable to Chron C6n [Cande and Kent, 1992]. The

sediments immediately above the upper unconformity have a

87Sr/86Sr ratio of about 0.70851, and are approximately 18.7

Ma [Hodell and Woodruff, 1994], assignable to Chron C5En

[Cande and Kent, 1992].


Materials


Benthic and planktic foraminifera from the >250Mm,

>212.m, or >150Am fraction (depending on abundance) were

picked from washed (>63Mm) samples from Hole 689B.

Preservation of foraminifera is generally good [Stott and

Kennett, 1990, Kennett and Stott, 1990]. In addition,

analysis of interstitial waters from Hole 689B [Egeberg et

al., 1990] indicates unusually low recrystallization rates

of carbonate. To further check the degree of preservation,

Sr/Ca ratios were measured as well (see Results).









The sampling interval is somewhat variable, ranging

between 0.5 and 3 m, or from 50 to 660 ky. The average

sampling interval was 0.96 m or 166 ky. Below 142 mbsf, the

sampling interval was coarser, with one sample per 1.44 m or

211 ky, while above that depth, the sampling interval

averaged one per 0.81 m or 134 ky. Because the amplitude of

the 87Sr/86Sr signal is much smaller below 135 mbsf, the

lower sampling rate below 142 mbsf is adequate to delineate

the 87Sr/86Sr curve.

The samples were ultrasonically cleaned in methanol to

remove adhering fine debris and dried in a 60'C oven. They

were then dissolved in 1 to 2 ml of 0.25 N HCl for about 1

hour at 200-250"C, then dried under a laminar flow hood.

Sample sizes range from about 1.3 to 10.0 mg. Strontium

concentrations were about 2000 ppm according to isotope

dilution techniques, and about 1350 ppm according to atomic

absorption spectrophotometry, giving sample Sr masses of 1.8

to 20 AIg.

Ion exchange columns, made of Teflon shrink-tubing and

packed with ElChroM Industries, Inc. Sr-selective crown

ether resin, were used to extract strontium from the

samples. All extractions were made under a laminar flow

hood. A modification of the method of Pin and Bassin [1992]

was used with slightly differing volumes of HN03 and H20

(Appendix B2). Yields were approximately 75%, with

excellent separation of Sr from Ca [Pin and Bassin, 1992].









Mass Spectrometry


Strontium isotopes


All Sr samples were analyzed in a VG Micromass 354

thermal ionization mass spectrometer in dynamic mode

(instructions for operation are listed in Appendix B3).

Samples were loaded onto oxidized Ta single filaments by

dissolving the extracted Sr in 2-3gl of 0.5 N H3P04 acid.

An analysis of NBS-987 was run with each turret of 14

samples. Analyses of 17 NBS-987 standards averaged

0.710240, with an uncertainty (2 standard deviations or 2cr)

of 0.000022. All samples are normalized to NBS-987 =

0.710235 and 86Sr/88Sr = 0.1194. Samples for which the
86Sr/88Sr ratio was determined to be less than 0.1194 are

assumed to have been fractionated and those values were not

reported.

To examine reproducibility of the results, two

different types of replicates were analyzed. Separate

analyses of planktonic and benthic foraminifera were made in

17 samples. The 87Sr/86Sr ratios of planktonic foraminifera

averaged 8x10-6 (a=20xl0-6) higher than those of benthic

foraminifera, showing no systematic difference between Sr

isotopic ratios of the different types. Because in some

cases less than perfectly preserved specimens of

foraminifera had to be picked due to lack of sample, I also

analyzed the difference between well preserved and poorly

preserved (visibly damaged, probably dissolved) specimens









when possible (5 cases). The 87Sr/86Sr ratios of well

preserved specimens averaged only 2x10-6 (a=13x10-6) higher

than those of poorly preserved specimens, which, despite the

small number of analyses, suggests that no systematic

difference exists between Sr isotopic ratios of

differentially preserved specimens. The poor preservation

of some foraminiferal tests is attributed to dissolution

rather than recrystallization. The average difference of

all 22 replicates is 16x10-6 (u=12x10-6), not much larger

than the standard deviation of the NBS-987 standard.


Oxygen and carbon isotopes


Well-preserved specimens of the benthic foraminiferal

genus Cibicidoides were picked from the washed >150Am

fraction of 35 samples from Hole 689B. Standard methods

were used to prepare and analyze the carbonate (see Chapter

3). The samples were analyzed in a VG Prism triple

collector mass spectrometer with an automated preparation

system. Of the 35 samples run, 14 were replicate analyses

of samples run by Kennett and Stott [1990] in an attempt to

calibrate their data with that of Chapter 3. The analytical

precision of standards (la) was 0.029 for 613C and 0.067

for 6180.


Atomic Absorption Spectrophotometrv


A Perkin-Elmer 3100 Atomic Absorption Spectrometer was

used to measure Sr/Ca ratios. About 50 Ag of carbonate










(typically 3-6 foraminifera) was dissolved in 5 ml of 0.5 N

Optima (trace metal grade) nitric acid. The Ca

concentrations (on the order of 3 ppm by weight) were

measured in flame mode, while the Sr measurements (on the

order of 10 ppb by weight) were analyzed using the graphite

furnace, on the same aliquot of sample. Both sets of

measurements were compared to standard absorbance curves to

calculate concentrations. Results are reported in both mass

and molar ratios (Appendix A2).

Results


Strontium Isotopic Measurements


From the base of the section analyzed in this study

(183 mbsf) to 133.2 mbsf, the value of the 87Sr/86Sr ratio

shows little variability, averaging 0.707735 (a=0.000018)

(Figure 2-2, Appendix A2). When plotted against age, the

87Sr/86Sr ratios remained steady or possibly decreased

slightly from an average of 0.707754 (a=0.000014) at 46.2 Ma

to the middle Eocene hiatus beginning at 41.4 Ma (162.54

mbsf) (Figure 2-3). By the time sedimentation resumed at

40.3 Ma, the 87Sr/86Sr ratio had dropped to near 0.70771.

The 87Sr/86Sr ratio averages 0.707728 (a=0.000015) between

40.3-35.9 Ma (161.4-133.25 mbsf), but rose slightly during

this interval, at a rate of about 8x10-6/m.y. This rise is

poorly constrained, with a regression coefficient (R2) of

only 0.459. None of the variability within this section is












a)

ow
0 )

0 <




O,0



o *a)W


o o



0 0 )
10
0 to0








04J -
,dao a

0 0 414


a) 01

0) ON A
04 kc 4


r. Q) 44
00C: P



t t -C C




V H) Pr.CO
0 04-) HN


0 --0 0

04J1 3 -4 a)
a) 0 0r



o to a -a

Id *.-l )



('4 9 W 4J
4->)O t 0



-4 u -.-i W ) 0
r4 4) 5p 0
lo r. Q) C q



P4 (a 0 Uk
0<^ nS'6U















(K'Lw) Alu!eijeoun e6v


TQ w

\~~ -J^2W
Si+ 0
c C >+
e a, T z

ca 0 i i+ u
C QX + 0
co zO

)II < U


c^ n 0) -
0 + +






10 < 4-7 >

0' 4- ww

o x +- + z
a) +) y i-
cm co x
< (
LO
Cy ++ +

ca c

CDL
Lo <:
x O





1+ <
d LULi
x z

IIco
0 CO +t


co Oda -
+3
Q0


1c

0
Nr
6O


o
o 0

d d

o6ley JS98/JSZ8


I I
o

I-
63


0)
Cv
co
cv
c-

co
-(0
Co

1O




"CO
O



C1)



C'Ja
coV
-C 2
"CO 0)


CO
^0

Co
0)

co
CNJ





CO
-Cm

-m
.cI


,,









likely to be of stratigraphic utility, because the total

range is so small.

During the late Eocene, a brief increase occurred in

four samples from 35.8-35.6 Ma (132.52-130.25 mbsf), with

the 87Sr/86Sr ratio rising to a constant 0.70778

(a=0.000005) before falling again to about 0.70775 (Figure

2-3). Although the increase is very consistent in this

site, it must be confirmed before any interpretations can be

drawn from it.

A steady rise began in the late Eocene at about 35.5 Ma

(128.8 mbsf, 38.2 Ma BKFV85), with 87Sr/86Sr ratios rising

to near 0.70812 by 25.2 Ma (68.8 mbsf), at a rate of

40xl0-6/m.y. This increase is well constrained, with R2 =

0.972. Two samples above 68.8 mbsf (25.0-24.8 Ma; 68.12 and

67.37 mbsf) show somewhat higher 87Sr/86Sr ratios of about

0.70818, suggesting the possibility of another small hiatus.

Across the hiatus in sedimentation from 24.8 to 20.2 Ma

(66.86 mbsf), 87Sr/86Sr ratios increased sharply to 0.70843

at 20 Ma. Another sudden jump in the ratio occurred across

the unconformity from 19.8 to 18.8 Ma (65.50 mbsf), to an

average of 0.70851 between 18.8 Ma and the top of the

section at 18.5 Ma (62.87 mbsf), confirming the unconformity

identified by SpieP [1990] and Gersonde and Burckle (1990].


Oxygen and Carbon Isotopic Measurements


In general, the purpose of the additional 35 isotope

measurements was to improve the resolution of the curve of









Kennett and Stott [1990], in particular near the earliest

Oligocene oxygen isotope shift. This was successfully done,

constraining the shift to between 118.44 and 121.30 mbsf

(33.33 to 33.96 Ma CK92, 35.63 to 36.37 Ma BKFV85) (Appendix

A2, Figure 2-2). The shift in this section has a duration

somewhat longer than reported elsewhere [e.g., Kennett and

Shackleton, 1976; Oberhansli et al., 1984; Zachos et al.,

1992a; Barrera and Huber, 1993; Chapter 3), lasting about

0.6 m.y. During this time, the 6180 values rise from

approximately 1.190/oo to 2.510/oo PDB, with the increase

delineated by 4 separate samples (two of which are

replicated from Kennett and Stott [1990]).

The difference between samples measured in 14

replicates by this study and those measured by Kennett and

Stott [1990] is statistically insignificant. Values for

6180 averaged 0.070/oo (a=0.29) lower in this study than

Kennett and Stott [1990], while values for 613C averaged

0.030/oo (a=0.18) higher in this study than in Kennett and

Stott [1990]. These differences are insignificant compared

to an analytical error of 0.10/oo for stable isotopic

measurements.


Strontium/Calcium Ratios and Diaaenetic Alteration


Strontium to calcium ratios were measured to assess the

effects of diagenesis in this section. Recrystallization

of calcite results in lower Sr/Ca ratios than in biogenic

calcite because of the lower distribution coefficient of









authigenically precipitated calcite present in slowly

altering limestone as compared to rapidly precipitating

biogenic calcite [Lorens, 1981]. Measurement of Sr/Ca

ratios can therefore be used to evaluate the degree of

recrystallization [Hess et al., 1986]. Sr/Ca ratios of 42

samples were measured from ODP Hole 689B (Figure 2-4,

Appendix A2) to evaluate recrystallization in this section.

The Sr/Ca ratios of the samples averaged 3.4 (a=0.6) Ag/mg

(mass ratio) or 1.5 (0.3) mmol/mol (molar ratio). The

Sr/Ca ratios rise very slightly up section, from 2.9 (0.3)

Ag/mg (mass ratio) or 1.3 (0.1) mmol/mol (molar ratio) in

the bottom 20 m to 3.7 (0.5) Ag/mg (mass ratio) or 1.7

(0.2) mmol/mol (molar ratio) in the top 20 m of section.

These values are comparable or higher than those reported by

other studies (Graham et al., 1982; Hess et al., 1986,1989;

Barrera et al., 1991; Egeberg et al., 1990] Figure 2-4 and I

therefore conclude that the 87Sr/86Sr ratios of the samples

used in this study have not been altered by diagenesis.

The 87Sr/86Sr and Sr/Ca ratios of bulk carbonate

(Egeberg et al. [1990]) correspond well with values measured

on cleaned foraminiferal calcite (Figure 2-4) as well. Bulk

carbonate, with a large component of fine fraction

carbonate, tends to be more susceptible to diagenetic

alteration than coarse fraction carbonate [Garrison, 1981].

The fact that 87Sr/86Sr ratios of bulk carbonate (at least

for the 4 measurements of Egeberg et al. [1990]) and
















C3 CI 0 Sr/Ca Ratio

] ug/mg [M U


X mmol/mol X. -1
(a)
0.7086- ( -0
0.7085- W (b)
0.7084-
0.7083- W
0.7082- W
0.7081- Sr/ "Sr Ratio
0.7080- +- --
0.7079- ,
0.7078-
0.707-7
upper Ollgo. lower Oligocene upper Eo. middle Eocene
60 70 80 90 100 110 120 130 140 150 160 170 180 190
Depth (mbsf)



Figure 2-4: Strontium geochemical measurements.
a: X's and empty boxes are Sr/Ca ratios in mmol/mol and
Ag/mg measured on cleaned foraminiferal samples (this
study), respectively, while solid triangles and boxes are
Sr/Ca ratios in mmol/mol and jg/mg measured on bulk
carbonate from Egeberg et al. [1990].
b: Crosses are 7Sr/8Sr ratios measured on cleaned
foraminiferal samples (this study), while W's and B's are
interstitial water and bulk carbonate 87Sr/86Sr ratios,
respectively, from Egeberg et al. [1990].









87Sr/86Sr ratios of foraminiferal calcite are so similar

suggests that little or no diagenetic alteration of these

carbonates has occurred. Lastly, the high 87Sr/86Sr ratios

of interstitial water (Figure 2-4) and relatively low

concentrations of interstitial Sr also argue for low rates

of recrystallization in ODP Hole 689B [Egeberg et al.,

1990]. If recrystallization had occurred, porewater

87Sr/86Sr values would be much lower and bulk carbonate

ratios would be higher. Egeberg et al. (1990] concluded

that pore waters at Site 689 displayed an almost complete

lack of recrystallization.

Discussion


Strontium Isotope Stratigraphy


Strontium isotope stratigraphy is used today in two

primary ways: for determining ages of marine samples based

on their measured 87Sr/86Sr ratio, and for investigating

geochemical cycling of strontium on the Earth's surface

[Hodell, 1994]. The variation in 87Sr/86Sr has been

described using regression analyses, either linear [McKenzie

et al., 1988; Miller et al., 1988, 1991a; Hess et al., 1989;

Hodell et al., 1989, 1990; Oslick et al., 1994] or

curvilinear [Miller et al., 1991a; Hodell and Woodruff,

1994; Oslick et al., 1994]. Oslick et al. [1994] suggested

that in the absence of a theory mandating a polynomial

(curvilinear) fit, the simpler linear regression should be









used. The present study uses line segments and linear

regression to model the data because of the relative

simplicity of the method and because linear regression seems

to fit these data well.

I use two linear regressions to model the data. First,

a regression for the Site 689 data from 40.3-35.8 Ma (42.8-

38.6 Ma BKFV85) was calculated. This regression, with 35

data points, is


87Sr/86Sr = 0.708024 7.8588x10-6 x Age (Ma) (1)


but gives a rather poor fit (R2 = 0.459) as previously

noted, and a standard error of 11x10-6. The equation can be

inverted for the purposes of age calculation to


Age (Ma) = 90093.124 127245.848 x 87Sr/86Sr (2)



and is valid for values of 87Sr/86Sr from 0.707707 to

0.707743. Second, I calculated a regression for the data

from 35.5-24.8 Ma (38.3-25.4 Ma BKFV85). This regression

equation, with 86 data points, is


87Sr/86Sr = 0.709154 39.6584x10-6 x Age (Ma) (3)


and gives an excellent fit (R2 = 0.972), and a standard

error of 21x10-6. The equation can be inverted to









Age (Ma) = 17881.554 25215.329 x 87Sr/86Sr. (4)



and is valid for values of 87Sr/86Sr from 0.707746 to

0.708171. The two regressions cross at a calculated age of

35.535 Ma and a 87Sr/86Sr ratio of 0.707745. I do not

evaluate the 4 data points from 35.8-35.6 Ma, which fall off

of both these trends. Figure 2-3 shows the fit of this

model to the data.

In order to evaluate the reliability of this curve in

examining controls on the marine 87Sr/86Sr ratio and in

predicting ages of samples, two statistical analyses of the

regressions were performed. The first analysis calculates

the 95% confidence interval of the regression itself (i.e.,

the uncertainty in the estimate that the regression

correctly models the actual increase in marine 87Sr/86Sr

ratios). The second analysis calculates the uncertainty in

age calculations made from one or more 87Sr/86Sr analyses of

an unknown sample. In the first analysis, I follow the

procedures outlined in Draper and Smith [1981], as suggested

by Miller et al. [1991a] in calculating uncertainties in the

regression. I use the exact solution in calculating

uncertainty in age estimation due to regression uncertainty

rather than the approximate solution used by Miller et al.

[1991a] because the calculations show that the approximate

solution slightly underestimates the uncertainty. In

addition, (2F)1 was used instead of the t statistic. This

allows the 95% confidence curve for the entire regression to









be calculated rather than the 95% confidence interval for a

given point in the regression [Draper and Smith, 1981, p.31]

(Figure 2-3). In the second case, the procedure outlined in

Miller et al. [1991a, equation 6] is used to calculate the

uncertainty in age determinations.

A caveat in these calculations is that they are based

on the assumption that 87Sr/86Sr measurements are

distributed normally about each point on the regression. If

there is small-scale structure along the line, as suggested

by several recent studies [Capo and DePaolo, 1990; Dia et

al., 1992; Clemens et al., 1993; Oslick et al., 1994], then

deviations from the regression are systematic. In this

case, calculations of ages and uncertainty using a

regression will be invalid.

The linear regression for the interval from 40.3 to

35.8 Ma shows a rather low regression coefficient

(R2=0.459). This, coupled with the low slope in this

interval, results in a high uncertainty of the regression

and raises the question of whether the slope differs

significantly from zero. A t-test of the regression [Draper

and Smith, 1981] was used to test this hypothesis. The t-

test shows that the slope is significantly different from

zero, at the 99.9% confidence level, despite the low

coefficient. A t-test of the interval from 35.5-24.8 Ma,

with a much greater slope and higher regression coefficient,

not surprisingly also shows that the slope differs

significantly from zero at this confidence level.









The uncertainty in the regression from 40.3 to 35.8 Ma

ranges from a minimum of 4.9x10-6 near the middle of the

interval to 8.6x10-6 and 10.9x10-6 at the upper and lower

ends of the interval, respectively (Table 2-2). The

corresponding age uncertainties are not quite symmetrically

distributed about the regression line because of the curve

of the lines of uncertainty, but are about 1.37, 0.71, and

1.76 m.y. for the upper, middle, and lower parts of the

interval, respectively (Figure 2-3). These calculations do

not take into account the uncertainty in 87Sr/86Sr

measurements, however. Given the 2a uncertainty in

87Sr/86Sr measurement of 22xl106 (comparable to a 95%

confidence interval), and a slope of 7.86x10-6/m.y., we can

divide the uncertainty by the slope to obtain an estimate of

the measurement uncertainty. We arrive at an uncertainty of

about 2.80 m.y. due to the measurement uncertainty.

Applying the formula given by Miller et al. [1991, eq. 6],

which more rigorously calculates the uncertainty in age

calculations based on a number of 87Sr/86Sr measurements,

the uncertainty for one measurement of an unknown sample at

the 95% confidence interval is approximately 3 m.y., and

decreases by about a third for 3 measurements. Clearly, the

uncertainty in this interval (with a total duration of only

4.5 m.y.) makes this regression useless for precise age

calculations.

For the upper (35.5-24.8 Ma) interval, the

corresponding uncertainties due to the regression in





















ON m 0 o r-I
8 C! C-4 0
r-IH m N r 4
a >i +1 +I +1 +1

0 C 'o Q
0 co t 4)
14 1 r- ON M 0 r- '0
4 in o + 1o r
4) c H V
I X -q +1 r-I N C H
S0 co O +1 +1 +1
C *-. co

u O N CI; *


0) +I N ( H
to p +1 +1 +1
0
0
0 m
4J 44 o r-I I co IO
0 '0o iC o0 o
S3 c *
N -I 0 > +1 +1 +1 +1

c- o %0
W 0 1 0) N 0C l ') (MN
CSI -4U 0 i (M0 1O A 0
S W I X v +1 0 0 e-
E-1 00 I) 'n o *w-i +1 +1 +1

Ln) I A
$4 m .4 w m 10 co qw
D i 0 o
*r.J H 0 0 H; 0
+1 +1 +1 +1
0 a)
4)



00 04 a) o0 I 0 r. 4+)
4J c 0 rq -H o + r. > 4) > 0 c
toi r-H 4.) w H 0 4) U A 0 )
0 CO Cr M U X VP 0 w 0 0 0
S( o 4I 0 0 0 14i 0
N 0 4) 0I () H )
1 co >I 44 0 (d 0 )
co 0 40
40 4.)
to 0 0 p C n
rN UN p H
co C 4 (N









87Sr/86Sr are somewhat larger, at 10.6x10-6, 5.7x10-6, and

12.0x10-6. Due to the higher slope, however, the age

uncertainties due to the regression are smaller, and are

about 0.27, 0.14, and 0.30 Ma for the upper, middle, and

lower parts of the curve, respectively (Table 2-2, Figure 2-

3). The age uncertainty obtained by dividing 2a uncertainty

by the slope is 0.55 m.y. Uncertainty based on 1

measurement is about 1.1 m.y. at the 95% confidence level,

decreasing to about 0.6 m.y. for 3 measurements, suggesting

that regression error is less important than errors in

87Sr/86Sr ratios in calculating ages for this part of the

curve.


Comparisons with Other Records


Three studies have used 87Sr/86Sr records for the

latest Eocene to Oligocene to calculate linear regression of

87Sr/86Sr ratio vs. age [Miller et al., 1988; Hess et al.,

1989; Oslick et al., 1994]. In addition, a fourth study

[Zachos et al., in prep.] has also generated a record from

the middle Eocene up to the earliest Miocene. In comparing

these records (from DSDP and ODP sites) to the Site 689

data, two modifications were made. For records originally

calibrated to the BKFV85 time scale, sample ages were

converted from the BKFV85 time scale to the CK92 time scale

in one of two ways. If magnetostratigraphy of a site was

available, a new age model tied to the CK92 time scale was

used to calculate ages. If magnetostratigraphy of a site









was not available, sample ages were calculated by comparison

of the magnetic anomaly tie points used by the BKFV85 and

CK92 age models. The new time scale can significantly

change the pattern observed in the marine 87Sr/86Sr ratios.

For example, the inflection identified by Hess et al. [1989]

at 26 Ma (BKFV85) is even more notable when the Hess et al.

[1989] record is calibrated to the CK92 time scale, where

the inflection is dated at near 24.8 Ma. Coincidentally,

this is near where the continuous portion of the record

ends, and so the slope change is not recognized in the data

from Hole 689B. Second, all 87Sr/86Sr results from these

studies were normalized to the standard value of 0.710235

for NBS-987. For example, Miller et al. [1988] reported a

87Sr/86Sr ratio of 0.710250 for NBS-987; therefore, 0.000015

was subtracted from their data. Hess et al. [1989] used the

Eimer and Amend standard, which converts to a value of

0.710220 for the NBS-987 standard; therefore 0.000015 was

added to all of their data. Oslick et al. [1994] reported a

87Sr/86Sr ratio of 0.710255 for NBS-987; therefore, 0.000020

was subtracted from their data.

In general, the records are quite similar, with the greatest

differences occurring during the early Oligocene and late

Eocene (Figure 2-5). With one exception, the data suggest

that the increase in 87Sr/86Sr ratios began earlier, but

initially at a lower rate, than previous studies. Data from

ODP Hole 748B [Zachos et al., in prep.] also suggests that a

lower rate of increase in 87Sr/86Sr began in the middle or












.44 *
0 ri-I
O0





4) ..



4. -) I- 1
in OW
n0 o0



0 o
4) 0 4)





ot
0o <

O 0% r-4i
Co rOc
Q)0 0)




M r.4 r_4
0 r -l
Iu 4) '

















%---
I0 0 a
%00
0 OQ) 0)
0 A'-')










00 0
40 V
,n oq -







0 0
V 4) -10
cWoD r1 0

+J 0) 9-(



.4 ra- tM -
V-C 00 -
to 0) 0
* 0-4 r-I 4
$14 to r-

Srfl r 1.4
in 4) 0














+ (0
--- + +I
CC) 0) a) r
00 oD 0) +
(3) 0) U

a +
__z xt2 i
+ 5



06 X > u -
a) a, =o
a, U) C) a) 0 o
U) = C) 0
v, xc o o + +r
I 0 I N H (






__ __ __ __ S cv,
? x -J-
cv,



x Lx CM
w
w z
x w cr)



o





..ia I :;C
a -
cli

J J W

U of
0-L~

OcoJ


(O N- (0 U) CO CM T- 0 C) N0 m6
co co co co (0 (0 (0 (0 (0 r- t- P%
0 0 0 0 0 0 0 0 0 0 0 0
0000000 00 666

o!4eUJS99/JSz89u euej/'









late Eocene, but the resolution of this record is low

compared to that of Hole 689B. I recalculated the linear

regressions of Miller et al. [1988] and Hess et al. [1989],

and include the regressions of Oslick et al. [1994] for

comparison to this model (Figure 2-6). The regressions of

the different records give quite similar results, with the

exception that the regression of Hess et al. [1989] does

not match the other regressions in the late Eocene to early

Oligocene. This suggests that there may be a problem with

the age model used by Hess et al. [1989]. Alternatively

(see Figure 2-5), there may not be enough points in the data

set to adequately constrain the early part of the records

from sites 516 and 563. It is interesting to note also that

there is a very consistent offset between the data of Miller

et al. [1988] and the data, even after normalization. The

data of Oslick et al. [1994] are offset on average by about

the same amount. The results in the present study are

consistently about 15x10-6 below those of Miller et al.

[1988] and Oslick et al. [1994]. This offset is in the same

direction, but not quite as large as that observed by Hodell

and Woodruff [1994]. Oslick et al. [1994] attributed the

offset between data from the University of Florida (Site

588; Hodell and Woodruff [1994]) and Rutgers (Hole 747A) to

errors in stratigraphic correlation. However, Holes 689B

and 522 [Miller et al., 1988] are well constrained by

magnetostratigraphy. This does not guarantee that the
















rU-




0

00


uO


H 0

0

ri Q)
Q)



4)
Xa)
r-4
,'-I
co.C




.a)



0I4
a)
0) r-"

r-4 a\

1-I




O




a) H
0




o










rlq 0P U
k V %0























4Qd
(I)
U)
0)
= /


oijel JS9e/JSL9 auUiel/ lapoij


C3-
CO
C(0


1Y

ce)
CO





cY)
d




C)
COl


CM
0,


-CO
oC)
CD)
cr
0) <
-Cr,
co
(0
oS




-CM


z

0
0
LU


w
LU
z
Lu
C.)
0

0i


LU


Ctl









correlation between these sites is correct, as small

unconformities and mis-identifications can be present which

would not be visible in the magnetostratigraphy, but it

makes it less likely that mis-correlations are the cause of

the offset between the holes. As yet, the reasons for the

offsets between laboratories remain unknown.

Some small scale variability can be seen in the record

from Hole 689B, with local peaks in 87Sr/86Sr values at

35.8-35.6 (as previously mentioned), 31.0, 29.0, and 26.6-

27.5 Ma, and a period of rapid rise from 33.0 to 32.1 Ma

(Figure 2-3). All of these values are within the expected

measurement error, and thus their significance is suspect.

However, they are all represented by more than one

measurement. Very small scale, high frequency variability

in 87Sr/86Sr ratios has been reported in the Plio-

Pleistocene and related to glacial/interglacial changes

[Capo and DePaolo, 1990; Dia et al., 1992; Clemens et al.,

1993], but these variations remain controversial. In

addition, Oslick et al. [1994] identified several episodes

of small scale 87Sr/86Sr variability in the Oligocene and

Miocene, and suggested that they lagged deglaciation events

by approximately 0.9-1.4 m.y. Whether the specific 32 and

28 Ma events identified by Oslick et al. [1994] correlate

with the 31 Ma (or 33 to 32 Ma) and 29 Ma events identified

here is unknown, but the high quality magnetostratigraphy in

the two sections involved (DSDP Hole 522 and ODP Hole 689B)

makes it unlikely. It is possible that finer scale









variation does exist in this core, but more detailed studies

will have to be undertaken to confirm this.


Controls on the 87Sr=86Sr Curve


Several possible controls on the strontium isotopic

composition of the ocean have been identified. Most of them

depend on changing inputs or 87Sr/86Sr ratios from

continental erosion (i.e., riverine input) of one type or

another. Some sources of strontium may be dismissed as a

control almost immediately. Dissolution of marine

carbonates only buffers seawater 87Sr/86Sr ratios from

changing, and cannot be invoked as a control [Hodell et al.,

1989). This also eliminates changes in the CCD as a forcing

mechanism. Input into the ocean from continental erosion

has a 87Sr/86Sr ratio considerably above that of modern

seawater (0.7119 vs 0.7092, Palmer and Edmond, 1992], and it

has been argued that as a result it is much easier to change

the seawater 87Sr/86Sr ratio by altering that source than by

altering hydrothermal input or carbonate dissolution.

Different proposed ways of changing the Sr input from the

continents are: (1) by opening a previously isolated Arctic

Ocean, with inputs from the Canadian and Siberian shields,

(2) by increasing erosion through the effects of the changes

in glaciation [Armstrong, 1971; Hodell et al., 1990; Miller

et al., 1991a; Oslick et al., 1994; Zachos et al., in

prep.], or (3) by increasing erosion via uplift of mountain

ranges such as the Himalayan-Tibetan Plateau [Raymo et al.,









1988; Hodell et al., 1989, 1990; Edmond, 1992; Richter et

al., 1992]. Hydrothermal input is usually considered to

change too slowly to be effective [Palmer and Elderfield,

1985; Elderfield, 1986]. However, the difference between

the 87Sr/86Sr ratio of hydrothermal input (0.703) and the

Eocene seawater 87Sr/86Sr ratio (0.7077) is even greater

than between riverine input and seawater 87Sr/86Sr ratio

(0.0047 vs. 0.0042), so hydrothermal input is also

considered as a way of significantly changing the marine

87Sr/86Sr ratio.

To investigate these controls, I used the method of

Hodell et al. [1989], which solves the box model of Palmer

and Elderfield [1985] for non-equilibrium conditions, to

investigate possible changes in fluxes. In calculating

initial conditions, I assumed present day fluxes for

carbonate dissolution and hydrothermal exchange, use the

present day 87Sr/86Sr ratio for hydrothermal exchange and

riverine input. The 87Sr/86Sr ratio for riverine input

(0.7119) should probably be lower in the absence of input

from the Himalayan-Tibetan Plateau [Palmer and Edmond,

1992]. However, by using the higher value, river flux is

allowed to vary more conservatively. I assume that prior to

the increase in marine 87Sr/86Sr in the middle Eocene, the

oceans were in equilibrium at 0.707707. This allows us to

use 0.707707 as the 87Sr/86Sr ratio of carbonate dissolution

(Table 2-3). I can then calculate the equilibrium riverine

flux as a starting point. At equilibrium, the riverine flux







66



f-I

0) C%

topLn go f- M r r>
0


P3 c4 ; 4 :r1)r
M M M0 00 0


>1
0 0 Io I
N r? N N
*c~L Vi Q in n-1 m r. r
+* H0My < 0 0 0



C. .
>O H C' 0 0 0 0
> 4
0 0
r4 A rN
(Mo mO'
co 0 riH 0 0 0


pq H a 0 0 0
0 4
*.-4 c 0 0 0
C 0 r4 g H e rH N rN
0 3 0O N V H 0 0O 0
U N> N^ N N
o 0 C n .
H0 0 0 0 0





4)
U) *
10 Ni Nt
o) H 0 0





41 0 H, O U .4
4)> r.i o' ( N

o .,1 0 0 H- H N N






to 9 9 dd
t3 (B O. ^ r r '
M r- n

c 1 H CU1






to co M cn co co co r-(
0) i 0 n N 0% om
*V V) HU 0 0 0

0) 0 U CO. NO On U N
1-i cO '0 CO 0) H C. NN
0 C.) N C-l H 0 0 0



0


$4 $4C; ro
0 Pd 4
m 00 C14 0H
co 0 (8 H H
H 0 U 0 Co co U c co
0l *d *-4 0 cn H- N .(T
ra4) to 0 o H m H O 0
o U U N N^ N' N-
E N M C14 0 0 0 .
0 0 0 0 0


14 0
00 U 0 Hp
0 1 f c ) V O4

VM Hj V. C4 O H- 00 0 H V


Hl Q ) *r(
o C) H 0 ) 00 0 0 0H 0%

H w C 1 0 A I
0

0 0 m 4 to > v 10 4 CO d


H r4 >1 $4 1 rN o to oH 0
r4 co 0 ( z
rl -r al 1p u cQ O -
CrP~UcP 43fl V 4






67








4-4
A



0
,-' >-
Um
-4 -. 3
a go c
0 I g
04


r-l 00 0 4
to- r. 144
> 0 .

o u m


I 4 14cl M 00


m
4 J J 0 -4

4. 0 -41r! 0
0 00 0
U 1M 4J 4) 0) O A

1 ( 4M 0 0


S-i (0 (A 0 .




) to V ir-q e

4o 4) 4J 0 P 4 t
ou -)a 44.o
-H f t -r o 10"$4
PWr O Q O 4 l04l
CD 4 A W 440) $4
r-4 -4. 0 .i 4 0 B 4)

HQ) o10 00 4 (0

10 3Or- (a *o O>
0 1 r-4 CD itl
0 O (drl V O0 (0
1 0 0O 4 H 10
WO CfSU- 41i0H 00)
S.H- Q) ) 4- 0 H 0
tC r0 m C i 0 0
k 0 H4-) 0 0 r-4 H 0 0 4)
SU0 0V-4 4.) H40 0 0V) 41
to to u4) H o 0 to v

*0 0 U0 *0 0 *0 .0

0 OC U0 44CO


(O CQ QiU'








in the middle Eocene is calculated to be less than 15x109

moles Sr/yr, about half of today's riverine flux of 33.3x109

moles Sr/yr.

I model the marine 87Sr/86Sr curve as a series of

straight line segments. This method of modeling can create

artificially abrupt changes in fluxes at changes in slope of

the marine 87Sr/86Sr curve, but is simpler to model and

easier to understand than using a continuously varying

model. Each step in the model is approximated by a portion

of an exponential curve. As a result, it is important to

use a time step which is small relative to the residence

time in order to accurately approximate the model. I use a

step of 0.1 m.y., less than 5% of the residence time of 2.5

Ma for the modern ocean [Hodell et al., 1990] to 3.4-4.0

m.y. for the Eocene and Oligocene (this study, as a result

of smaller riverine fluxes).


Arctic opening model


Plate tectonic reconstructions of the opening of the

Norwegian-Greenland Sea suggest that opening occurred along

a fracture zone, leading to a gateway effect in which the

Arctic was suddenly put into communication with the North

Atlantic Ocean over a short period of time. Prior to 72 Ma

(Cretaceous), the North Atlantic Ocean had not yet begun to

form and the region between Scandinavia and Greenland was

closed with no intervening ocean floor. The first extension

in the Arctic region is believed to have begun at Chron 30









time [Rowley and Lottes, 1988] and the oldest seafloor

anomaly to be identified in the North Atlantic and Eurasian

basins is Anomaly 24 [Vogt et al., 1981], at about 53 my

[Cande and Kent, 1992], of early Eocene age. A close

connection is maintained between Greenland and Europe

(Svalbard) during the Eocene as Svalbard was transformed

eastward relative to Greenland along the Greenland fracture

zone. At Anomaly 13 time (33.4 Ma), the contact between

Svalbard and Greenland was tenuous and the final separation

between the European and American plates may have occurred

at this time. This disagrees with interpretations of Lawyer

et al. [1990] which suggested a later opening (15-10 Ma).

However, as we shall see below, only a small opening would

be necessary to have a large effect on marine 87Sr/86Sr

ratios. Because the seafloor in the North Atlantic and the

Arctic Ocean would have been at depths of over 4000 m, the

opening may have occurred rapidly, over hundreds or

thousands of years. The motion of the Arctic basin may have

occurred similar to a sliding door, allowing water from the

surface to 4 km depth to be released suddenly into the

Atlantic Ocean.

This sudden influx of cold deep water should produce

severe erosion of abyssal sediments in the North Atlantic

Ocean. Severe erosion is, in fact, observed beginning in

earliest Oligocene time (C13) (e.g., the reflector R4 of

Miller and Tucholke [1983]). If the Arctic had been

isolated up to that point, it would then have mixed with the









world ocean, and added its riverine input, hydrothermal

exchange, and carbonate dissolution to those of the World

Ocean.

By using a few assumptions about those sources (Table

2-3), we can calculate what the equilibrium 87Sr/86Sr ratio

of a closed Arctic Ocean would have been (0.709811) and what

riverine input into the world ocean (in the absence of

Arctic input) would be at the equilibrium marine isotopic

value of 0.707707 (14.328x109 moles Sr/yr, as opposed to

14.818x109 moles Sr/yr with Arctic input). At the point

when the marine 87Sr/86Sr ratio begins to increase, I model

the connection between the Arctic Ocean and the world ocean

as a small flux between the two oceans, on the order of

1/100 of the flux of Sr carried by the Gulf Stream. I then

use the non-equilibrium model of Hodell et al. [1989] to

model the response of the world ocean to these new inputs

(Figure 2-7a).

Rather than recreating the marine Sr isotope curve, it

is clear that there are fundamental differences between the

modelled and the real curve. First, there is an immediate,

sudden jump in the world ocean 87Sr/86Sr ratio, as the

isotopically more radiogenic Arctic Ocean mixes with the

world oceans, even with the relatively small flux between

the two. Secondly, the modelled marine 87Sr/86Sr ratio

asymptotically approaches equilibrium with the new riverine

input from the Arctic Ocean, rising steeply at first, then

less steeply as time goes on. The modelled Arctic Ocean



















0 .
0 0






O U4




4-4 -,-4





O )U 04




#r4 4
-. 0 0
) V 0
D '.-4.
oo U V




4) ) t 0







-.4 ,-Iq 0
s-I 4. v







0 0







--I0 -4
I) 4)
0 0 44
t, uP 0 p
0 000
co *d IC


w Uk40
0 r. r0
r. ( t
(0 r UT



0 .. 0 .I.*


a 0A0 r
r0 -. l 0 0








t^. Ylr~~













(o01OJv) olel JS9g/-ISL8
0 LO 0 LO
C 0 0 0 0
0 co o 6
d d d d


o!.el JS98/ISL8















(JA/lOW 60 ) xnlij jS leioel

L) O O i C) 0 LO O L) 0) L) O
Lo i < c ci c c\i T- ,-: d d



w
co







I -- --
00



8 aco






0,- C )
V:4 1
a:o
CII I ~'/< 5i o


(Y) cM V- o 0) co N^
o O o co -
o Co o o o oI
0 0 0 0d d d


o!Pet JS99/JSLS


















(JA/IOwU 60) xnlj jS euuJA!l

LO O I O 0O o L0 O LO lO
6 6 c6 c6 r i- c6 (6 L6 L 4


o!,ey JS99/JSL9


cv
LO

I-


co,

cr)
Co



cv,
CD


cfO
c, -

cv,
Cr3
CM E3)
CO S



















(JA/Iow 600) xnlj JS IewJa4ioJpAH


LO
CM


h i/







``I j IL

O
", //
















o."o
/ / <


y ~/^






I,1
0
N~~Qf



C)
0

~ I.-

I,\


0
co
0
rN
d


o o
r. ',
0 0
I',. r,-
d d


OI1BtJ JS9IgjSLg


0 0)
CO Cj


CO)
c \


L-
(/)
co
CO





0w
w


f>- D(
CV CU


CO C
co c00
0 0
Nd r
6 d


4.-'


I I............ I I i


0

C)
c')
00
COl

C,
CD
co
CO

cLO



-co0

cm
CM





-0)
CM
co
CM

N.
1-c'
CM
-C"
iLO
-CM
qTr





-(l
-CM


j









87Sr/86Sr ratio decreases suddenly and then asymptotically

approaches equilibrium with the world ocean. It is clear

that this model cannot adequately explain the nearly linear

increase in marine strontium isotope composition from 35 to

26 Ma. It is also clear that strontium isotopic

measurements of Arctic carbonate sediments may help to

constrain the time of connection of this basin to the world

ocean.


Glaciation model


Recent evidence suggests that glaciation on Antarctica

began near the time that the marine 87Sr/86Sr ratio began to

increase, and that glaciation may in fact be a cause for the

rising 87Sr/86Sr ratios of seawater [Armstrong, 1971; Hodell

et al., 1990; Miller et al., 1991a; Hodell, 1994; Oslick et

al., 1994; Zachos et al., in prep.]. It has been known for

some time that a major increase occurred in 6180 at about

33.4-33.6 Ma (35.7-35.9 of BKFV85). The cause was

originally suggested to be the onset of cold bottom water

production [Shackleton and Kennett, 1975] and more recently,

due to the onset of Antarctic glaciation [Matthews and

Poore, 1980]. A pulse of IRD has been found on the

Kerguelen Plateau [Zachos et al., 1992b; Breza and Wise,

1992; Barrera et al., 1994] and in lesser amounts on the

Maud Rise [Ehrmann and Mackensen, 1992], which coincided

with the earliest Oligocene increase in 6180 values. This

suggests that glaciation on Antarctica reached the coastline









at least briefly at about 33.4-33.6 Ma. At this time,

benthic foraminiferal 6180 records suddenly increased by

about 1 /oo, then decreased by about 0.3-0.4 /oo [Miller

et al., 1987; Zachos et al., 1992b; Chapter 3]. This

suggests a sudden advance, followed by a retreat of

continental ice sheets.

Other evidence may point to a somewhat earlier onset of

glaciation. Sedimentological studies of ODP Legs 113 and

119 [Ehrmann and Mackensen, 1992] suggest that glaciers may

have affected Antarctica as early the late middle Eocene,

43.3 Ma (45.5 Ma BKFV85), with a further expansion near the

middle/late Eocene boundary at about 37.1 Ma (40 Ma).

Somewhat controversial evidence of terrigenous grains first

occurred on the Kerguelen Plateau and possibly the Maud Rise

in the middle Eocene from 43.3-39.1 Ma (Wise et al., 1992],

suggesting ice-rafting events during the Eocene. In

addition, massive diamictites drilled in Prydz Bay,

Antarctica, and interpreted as till deposits [Ehrmann et

al., 1992] have been tentatively dated between 37.9 and 34.8

Ma (40.8-37.5 Ma), suggesting a late middle to late Eocene

age for Antarctic glaciation. An increase of chlorite and

kaolinite at the expense of smectite on the Maud Rise in the

late Eocene is suggested to have resulted from increased

physical weathering on Antarctica attributed to increased

Antarctic glaciation [Ehrmann and Mackensen, 1992; Robert

and Chamley, 1992], although a moderate, humid climate is

still postulated for that time. The major sedimentologic









evidence cited by Ehrmann and Mackensen (1992] still

suggests the major onset of glaciation in the earliest

Oligocene at about 33.5 Ma. This is the time when the

largest change from smectite to illite occurs in marine clay

mineral assemblages, indicating a change from chemical to

physical weathering conditions on Antarctica [Ehrmann et

al., 1992].

In modeling the effects of glaciation on Antarctica,

equilibrium values for riverine input, hydrothermal

exchange, and carbonate dissolution (Table 2-3) are used.

As a rough average 87Sr/86Sr of typical "Gondwana" shields,

0.715 is used by averaging the values for rivers from

Australia and southern Africa [Palmer and Edmond, 1989]. I

use this value for a new, glacial erosion input (holding

riverine input, hydrothermal exchange, and carbonate

dissolution constant at their equilibrium values; Table 2-3)

and calculate the glacial erosive flux of Sr necessary to

produce the observed record, using the non-equilibrium

method of Hodell et al. [1989] (Figure 2-7b).

In this case we can compare the model's predicted

glacial activity with the observed indications of

glaciation. The comparison is not encouraging. The onset

of modelled glaciation occurred at 40.4 Ma, defined by the

time when the marine 87Sr/86Sr began to increase, and rose

from a low of just 0.24x109 to 0.52x109 moles Sr/yr by 35.5

Ma (Figure 2-7b). The modelled glacial Sr flux then jumped

to 1.50x109 moles Sr, about 1/10 of the riverine flux of









14.8x109 moles Sr/yr, reflecting the increased slope of the
87Sr/86Sr curve, and finally rose to a high of 5.19x109

moles Sr/yr by 24.8 Ma. In contrast, most sedimentologic

and oxygen isotopic evidence points to a somewhat later

onset of glaciation, at about 33.5 Ma, 2 m.y. later than the

first increase in marine 87Sr/86Sr [Rea, 1992 and this

study]. This glaciation, rather than starting at low

intensity and gradually increasing over the next 8 m.y. or

so, as suggested by the strontium isotopic modelling, seems

to have been episodic based on oxygen isotopic evidence,

with maxima at about 33, 29, and 24 Ma (35, 31, 25 Ma;

Miller et al., 1987, 1991a; Wise et al., 1991].

Oslick et al. [1994] also examine the possibility that

glaciation and the 87Sr/86Sr curve are related. They

identified 6 events in the Oligocene and Miocene in which

very small scale 87Sr/86Sr (from ODP Hole 747A) increases

lag deglacial events by 0.9 to 1.4 m.y. The deglaciation

events were defined by decreases in 6180 values that

followed previously identified glacioeustatic episodes

[Miller et al., 1991b, Wright and Miller, 1992]. They

attribute the overall rise in marine 87Sr/86Sr ratios in the

Oligocene and the shorter scale 87Sr/86Sr increases in the

Miocene to glaciation, with the overall rise in the Miocene

due to uplift of the Himalyan-Tibetan Plateau. The 0.9 to

1.4 m.y. lag in response is attributed to the long residence

time of the oceanic Sr reservoir.









I disagree with this interpretation on three points.

First, the long residence time of Sr will create a lag

between the initial 87Sr/86Sr response and its final

equilibrium value, not between the driving mechanism

(weathering of the Antarctic craton during deglaciation

[Oslick et al., 1994]) and initial response of marine
87Sr/86Sr values. If in fact the lagged association between

deglaciations and 87Sr/86Sr increases is real, another

mechanism must be invoked to delay the release of radiogenic

Sr to the ocean until 0.9 to 1.4 m.y. after the

deglaciation. Second, as stated above, the continued rise

in marine 87Sr/86Sr ratios implies a gradually increasing

level of glacial weathering throughout the Oligocene (Figure

2-7b). Evidence for glaciation in the Oligocene instead

suggests a more episodic behavior. Unless the amplitude of

these episodic glaciations slowly increased (which is not

implied by 6180 records) or unless weathering on the

Antarctic craton gradually increased during each deglacial

episode (perhaps through dissolution of gradually increasing

deposits of finely ground, easily weathered cratonic till

[Hodell et al., 1990, B. Opdyke, pers. comm., 1994]), it

would not be possible to produce the observed 87Sr/86Sr

curve from Oligocene glaciation. Third, the inflection to a

high slope of the 87Sr/86Sr curve clearly begins before the

first glaciation event cited by Oslick et al. [1994] (about

35.5 Ma vs. about 33.5 Ma). The time of onset of the slower

rise in 87Sr/86Sr began even earlier, around 40 Ma.









Even if interpretations of sedimentologic evidence

suggesting earlier onset of glaciation are correct, it may

be hard to ascribe the increase in marine 87Sr/86Sr to

glaciation. Two factors support this conclusion. First,

the 87Sr/86Sr ratio in river water draining continental

shields under most circumstances is likely to be lower than

the bulk ratio of the shield rocks, because the minerals

containing highly radiogenic 87Sr/86Sr ratios are more

resistant to erosion [Palmer and Edmond, 1992; Edmond,

1992]. Higher (more radiogenic) 87Sr/86Sr ratios are

present in minerals with high Rb contents (i.e., containing

incompatible monovalent cations such as potassium feldspar,

sodic plagioclase, and muscovite) which are also fairly

resistant to weathering. Biotite has been suggested as one

mineral containing highly radiogenic Sr which is easily

weathered in glacial deposits, but the amount of Sr

contained in biotite is low [Blum and Erel, 1995]. Less

radiogenic 87Sr/86Sr ratios are generally present in more

easily weathered minerals containing divalent cations such

as calcic plagioclase and olivine. Strontium isotopic

ratios derived from shield areas are therefore usually

either not high enough, or with strontium fluxes too low, to

control the marine 87Sr/86Sr ratio [Edmond, 1992]. For

example, the marine 87Sr/86Sr ratio of rivers draining the

Canadian Shield averages only 0.7111, with more radiogenic

ratios present in rivers with lower Sr concentrations

[Wadleigh et al., 1985]. Special circumstances must be









invoked, such as deep-seated metamorphic homogenization, for

rivers with high 87Sr/86Sr to also have large Sr fluxes

[Edmond, 1992]. Second, although the increase in illite in

the late Eocene suggests climatic deterioration possibly

associated with glaciation, it is suggestive of glaciation

precisely because it indicates a change in the mode of

weathering, from chemical to physical. This, however, may

not necessarily reduce the amount of Sr weathered from the

craton, because mechanical grinding by glaciers may lead to

enhanced chemical dissolution [Armstrong, 1971; Hodell et

al., 1990]. There is evidence that chemical denudation

rates of cations in glacial environments are high [Reynolds

and Johnson, 1972], and it is likely that the temporary

glacial events suggested for the Oligocene were temperate

(wet based) in nature.


Uplift model


The Himalayan-Tibetan Plateau system is unique in a

number of ways. First, it is the highest mountain range on

Earth, with an average elevation of 5000 m [Raymo and

Ruddiman, 1992]. Second, rivers draining it have high

87Sr/86Sr ratios for the flux of Sr that they carry [Palmer

and Edmond, 1989, 1992]. It has been suggested that the

anomalous co-occurrence of high 87Sr/86Sr ratios with high

Sr fluxes in Himalayan rivers is due to unique tectonic

aspects of the region [Palmer and Edmond, 1992; Edmond,

1992]. Uplift of the Himalayan orogen followed unusually









deep seated metamorphism due to the collision of India with

Eurasia and subsequent crustal thickening, allowing

radiogenic Sr (formerly trapped within hard-to-weather Na-

and K-bearing silicates) to be remobilized into easier-to-

weather Ca-silicates [Edmond, 1992]. Uplift exposes more,

and deeper, continental crust to weathering, and because of

mass wasting in areas of high relief, fresh rock is

repeatedly exposed to continued weathering. As a result,

several studies [Raymo et al, 1988; Hodell et al., 1989,

1990; Raymo and Ruddiman, 1992; Richter et al., 1992; Hodell

and Woodruff, 1994] have suggested that uplift of the

Himalayas controls the marine 87Sr/86Sr ratio during the

Cenozoic by increasing the amount of continental weathering

and delivery of radiogenic Sr to the world's oceans.

Raymo et al. [1988], Hodell et al. [1989, 1990], and

Raymo and Ruddiman [1992] suggested Himalayan control of

marine Sr isotopic ratios during the Neogene. Richter et

al. [1992] suggested that this control extended back to the

Paleogene, with the first effects of the uplift on marine

87Sr/86Sr ratios occurring at about 37 Ma (40 Ma BKFV85,

their date for the first increase in the Sr isotopic ratio).

They modelled the increase using a non-equilibrium model in

which riverine input of Sr controls the marine 87Sr/86Sr

curve. I re-calculated the curve for this time period for a

number of reasons. First, the model is based on the data

from Hole 689B, which is more highly refined than previous

records. Second, the time-step used by Richter et al.









[1992], 1 m.y., was too small to accurately portray the

large changes in flux implied by the data. An example of

this is the change in 87Sr/86Sr ratios that occurred in the

late Eocene. The sudden change in slope of the marine
87Sr/86Sr curve implies that large and rapid changes in

marine 87Sr/86Sr ratios occurred at this time. By using a 1

m.y. time step, they underestimated these changes. Third,

the model used by Richter et al. [1992] ignores the effect

of carbonate dissolution in calculating fluxes. Although

small, this flux is significant, and leaving out its

buffering effect [Raymo and Ruddiman, 1992] on the Sr

isotope budget gives underestimates of the fluxes necessary

to change the marine 87Sr/86Sr curve. Fourth, I re-

calculate the curve based on the new Cande and Kent [1992]

time scale. These calculations are based on rates-of-

change, which are sensitive to changes in time scale

[Delaney and Boyle, 1988].

I model the change in two ways. Richter et al., [1992]

noted that in increasing erosion from the Himalayan-Tibetan

Plateau, both riverine Sr flux and 87Sr/86Sr may change. I

first assume that world riverine 87Sr/86Sr stays constant at

the current value, 0.7119 [Palmer and Edmond, 1989]. The

result is that a small jump in flux occurred, from the

equilibrium value of 14.82x109 to 15.06x109 mcles Sr/yr at

the beginning of the increase in 87Sr/86Sr at 40.4 Ma. The

riverine flux gradually rose to 15.34x109 moles Sr/yr by

35.5 Ma. At that time, a sudden increase in riverine









strontium flux occurred, from 15.34x109 to 16.32x109 moles

Sr/yr, an increase of about 6.4%. The sudden jump is

probably an artifact resulting from the modelling of the

data as having a sudden change in slope, from nearly stable
87Sr/86Sr values to steadily increasing values. If instead

of suddenly changing, the marine 87Sr/86Sr ratio ramps

gradually up to increasing values, the increase in riverine

flux would be more gradual. The model flux grows steadily

from 35.5 Ma to a maximum of 20x109 moles Sr/yr at 24.8 Ma

(Figure 2-7c, Table 2-3). For the second model, I assume

that the riverine Sr flux began to increase slowly, as the

Himalayas were uplifted, but that the 87Sr/86Sr ratio of the

rivers draining the Himalayas was higher, such that the Sr

flux increased from an equilibrium value of 15.06x109 moles

Sr/yr. In this case, it is a simple equilibrium calculation

to determine what the riverine 87Sr/86Sr ratio would have

been before the increase. Using equilibrium calculations,

if the Eocene riverine Sr flux was 15.06x109 moles Sr/yr,

the riverine 87Sr/86Sr ratio must have been 0.711833. These

results are similar to the results of Richter et al. [1992],

except that (1) the calculated fluxes are higher, and (2)

the timing of course is different, in part because of the

different time scales, and in part because of the different
87Sr/86Sr curves employed.

Superficially, this seems to be a reasonable simulation

of the events leading to the change in marine 87Sr/86Sr from

the Eocene through the Oligocene. The next step is to









determine whether this is a realistic model for the uplift

of the Himalayas. Richter et al. [1992] cited evidence for

the collision of India with Eurasia beginning at about 50 Ma

(BKFV85), and increased uplift at about 20 Ma (BKFV85).

Convergence rates between India and Eurasia decreased

suddenly about 50 Ma (BKFV85)[Richter et al., 1992]. This

might suggest that the uplift could have begun somewhere in

between these times, specifically at the time when the

marine 87Sr/86Sr ratio began to increase. Richter et al.

[1992] calculated that enough crust has eroded from the

Himalayan-Tibetan Plateau since the collision to account for

the increase.

The next point to consider is whether the timing is

correct. Several recent studies on tectonism in the

Himalayan region suggest that the uplift did not begin until

the latest Oligocene or Neogene. These studies use

40Ar/39Ar closure dates [Harrison et al., 1993; McFarlane,

1993] and U-Pb dates [Parrish and Hodges, 1993] to date the

first large episode of Himalayan deformation at 27-19 Ma.

In addition, Rea [1992] examined mass accumulation rates

(MAR) in the Indian Ocean and found that most of the

accumulation occurred after 16 Ma. Even after 16 Ma, there

was little correlation between MAR and the Sr isotope curve.

He concluded that runoff from the Himalayan region cannot

explain the increase in marine 87Sr/86Sr before 16 Ma, and

is not the only control after 16 Ma.