Oceanography 540

Oceanography_540.
doc
Oceanography 540--Marine Geological Processes--Autumn Quarter 2002
Oceanography 540:
Marine Geological Processes
Principles of thermodynamics, heat and mass transfer, fluid mechanics, continuum mechanics, and time-series analysis
applied to marine geological and geophysical data. Applications to heat transfer from the seafloor, to near-ridge Pleistocene
sediment records and to global sea level.
Ocean 540 (i.e., Ocean 549b) continues major revision in Autumn Quarter 2002, making a transition from 5-credit format to 3-credit format. This
web will make extensive use of the 2001 course web. In addition a tgz file of an earlier version of the course web from Spring 1997 will remain
on-line indefinitely. Please let me know of any problems (content, formatting, broken links): mcduff@ocean.washington.edu (9/30/2002)
Syllabus
Overview
Reference Materials
Lectures
Grading
Computer Use
Lecture Schedule (Autumn 2002)
Old Notes Index (2001)

Problem Sets
Problem Set #1 Due Monday, October 14
Problem Set #4 Due Monday, November 4
Final Exam Information

UW Library Links
Course Reserve Materials
UW Libraries Catalog
GeoRef Database Search (UW access only)
Ocean 540 (i.e., Ocean 549b):

Marine Geological Processes
Autumn Quarter 2002
Class Meetings:
211 Ocean Teaching Building
(and occasionally in Spatial Analysis Lab, 111 Ocean Sciences Building)

Monday, Wednesday and Thursday--1:30 p.m -2:20 p.m.
Instructor:
Russ McDuf
119 Ocean Teaching Building, 3-3058 (usually)
or
270 Marine Science Building, 5-1947 (occasionally)
mcduf@ocean.washington.edu
Open Door Office Hours (Details)
Course Web http://www2.ocean.washington.edu/oc540/
Course Overview
Marine Geological Processes is required for students in the MG&G option and is also intended as general course for students outside the
option. The focus of the course will be examination of a small set of selected themes involving the geologic history of earth and ocean
processes. In particular we will focus on processes involved in 1) the development of the sedimentary record in the vicinity of the mid
ocean ridge and 2) the record of global sea level on a broad range of time scales. The approach taken is to apply principles of
thermodynamics, heat and mass transfer, fluid mechanics, continuum mechanics and time series analysis to modeling and understanding
of observational data. In other words, the course is not meant to be encyclopedic, but rather illustrative of the methodology used in
attacking significant contemporary problems in the marine geosciences. Lectures and discussions will be supplemented with reading from
the literature and problem sets.
Reference Materials
Over the past decade, Ross Heath and I developed a set of web-based lecture notes which have functioned as the text
for the old 5-credit version of the course. This lecture archive is available from this index. Reading from these notes will
be assigned in advance of related lectures. This quarter I will build a new set of lectures notes representing the content
of the three-credit format. I will bring copies for you each class session, although they may still be in relatively rough
form.
The course content is not based on any specific textbook. If your background in earth sciences is limited, the text The Ocean Basins:
Their Structure and Evolution published by the Open University (ISBN 0750639830) may be a valuable supplement to the lecture
material. Another excellent introductory reference on the earth sciences, is Understanding Earth by Frank Press and Ray Siever (ISBN
0716731339). A comprehensive, though mainly descriptive, introduction to marine geology emphasizing the history of the ocean basins
can be found in Marine Geology by Jim Kennett. Many have found Turcotte and Schubert's Geodynamics (ISBN 0521666244 paperback)
useful for portions of the course. All four of these books will be on reserve in the Fish-Ocean Library.
References to additional relevant material are listed with the associated lectures.
Lectures
The lecture schedule will be evolving. Changes will be announced in class and posted on the web.
Grading
Problem Sets (75%)
There will be weekly problem sets distributed at the Thursday class session and due the following week. You are
welcome and encouraged to discuss problems with your classmates, but the work you turn in should represent your
own understanding of the problem. Late work will be discounted 10% per day, unless arrangements have been made in
advance..
Final (25%)
There will be a one-hour individual oral final.
Computer Use
We will make use of computing tools in illustrating various concepts. For problems involving time series analysis we will
use MATLAB, a matrix oriented interactive computing environment with powerful graphics capabilites and a collection
of scientific toolboxes, e.g., the signal processing toolbox. MATLAB is available on the PCs in 102 MSB, the Fish-Ocean
library and the SAL (111 OSB) and can be run via XWindows from the UW uniform access machines, e.g.
mead.u.washington.edu. I will also make use of Mathematica as a tool for problem solving, but it is unrealistic to think
that you would surmount the learning curve during this course (but let me know if you'd like help!).
Oceanography 540
Lecture Schedule
Autumn 2002
Monday
Wednesday
Thursday
Sep 30
1--Course Overview; Geologic Overview,
Topography and Age of the Ocean Basins
Oct 2
2--Marine Sedimentation,
Sediment Thickness and Type
Oct 3
3--Conduction; Conductive Plate Model
Oct 7
4--Subsidence of the Lithosphere, Sea Level
Oct 9
5--Mantle Flow and Crustal
Generation
Oct 10
6--Phase Diagrams
Oct 14
7--Petrological Relationships in Generating
Ocean Crust
Oct 16
8--Convection and Rayleigh
Criteria
Oct 17
9--Porous Media
Oct 21
10--Convective Flow in Porous Media
Oct 23
No Class
Oct 24
11--Distribution of Convective Heat Flux
Oct 28
12--Determining Hydrothermal Fluxes
Oct 30
13--Determining Hydrothermal
Fluxes (continued)
Oct 31
14--Time Series Analysis
Nov 4
15--Time Series Analysis (continued)
Nov 6
16--Sediment Accumulation
Nov 7
17--Sediment Diagenesis
Nov 11
Veterans Day
Nov 13
18--Applying Matlab to Core
Analysis
Nov 14
19--Oxygen and Carbon Isotopes as
Paleotracers
Nov 18
No Class--Faculty Retreat
Nov 20
20--Orbital Geometry
Nov 21
21--Pleistocene Records: Frequency Domain,
Time Domain, Phase Relationships
Nov 25
22--Pleistocene Glacial-Interglacial Cycles
Nov 27
23--Pleistocene Responses: The
100 ky Problem
Nov 28
Thanksgiving
Dec 2
24--Boundary Shear
Dec 4
25--Settling of Particles;
Threshold of Motion
Dec 5
26--Particle Collisions and Aggregation
Dec 9
27--Suspended Load
Dec 11
28--Sea Level in the Frequency
Domain
Lecture Index | Search the Ocean 540 Pages | Ocean 540 Home
http://www2.ocean.washington.edu/oc540/lec01-1/
Structure of the Oceanic Lithosphere, Plate Tectonics and Seafloor Spreading, Seafloor Topography
I. Layering of the Earth
The Earth accreted from the solar nebula ~4.5 By ago, forming a more or less homogeneous body with two sources of heat: (1) energy associated
with impacts during the early history of the Earth and (2) continuing decay of radioactive elements (U, Th, K). Strong convection coupled with
cooling at surface gave rise to a chemically stratified body with a layered structure.
Seismic layering, marked by sharp transitions in the velocity at which seismic waves propagate (the Mohorovicic discontinuity at the
crust/mantle boundary and the Gutenberg discontinuity at the mantle/core boundary), can be used to recognize the crust (thin beneath oceans (~6
km), thick beneath continents (~30-45 km), mantle, and core:
Figure 1-1. (from (41))
This layering is inferred to reflect a chemical layering:
crust: average composition grandiorite on continents, basalt in ocean
mantle: Fe-Mg-Si
core: Fe-Ni-S
An Aside on Igneous Rocks In the Mantle and Crust
Nature of layering: seismic waves, either natural (earthquake generated) or man-made (explosions), propagate through the earth. These waves are
of two kinds: compressional (P waves) and shear (S waves). P wave velocity and density within the Earth is shown in this figure:
Figure 1-2. The Preliminary Reference Earth Model (PREM2), calculated from the prem2.m MATLAB script in the Coral software package.
P waves propogate at higher speed than S waves, thus the difference in arrival time is a measure of distance from the source and the basis for
earthquake location. As seimic waves propagate they are refracted by density contrasts; in addition S waves cannot propagate through the liquid
outer core. Refraction by major layers gives rise to a shadow for P waves; Refraction and inability to propagate through liquid gives rise to a
larger shadow for S waves:
Figure 1-3. Selected ray paths for P-waves passing through the earth. The P and S shadow zones are marked on the right. The dashed ray paths
represent weak P waves detected within the shadow zone which provide the principal evidence for a narrow zone of high P velocity in the core.
From (2), Figure 2-1, pg 14.
Within the crust and upper mantle there is also an inferred mechanical layering distinguishing lithosphere, asthenosphere and mesosphere.
Lithosphere (from Greek, for stone): cool (<1200C) and mechanically rigid. Asthenosphere (from Greek, for weak or soft): ~600 km thick,
capable of plastic deformation, containing perhaps a few % melt.
The term lithosphere is used in a variety of ways: one can distinguish a mechanical lithosphere, the asthenospheric boundary representing the
transition from brittle to ductile behavior, and a thermal lithosphere whose asthenospheric boundary represents a particular isotherm, usually
equated with the solidus or melting temperature of the rock making up the mantle. The mechanical lithosphere is ~60 km thick, while the thermal
lithosphere is ~120 km thick. There is an extended discussion of this and related semantics in reference (29). We will generally use the term
lithosphere to mean the thermal lithosphere, reflecting our particular emphasis on heat transfer.
II. Gravity, Isostasy
A second tool providing information on subsurface structure is the gravity field. Gravitational acceleration near the surface of the earth is order
9.8 m s . Of interest are small deviations from this average. The unit for these deviations is the gal (honoring Galileo); 1 gal = 1 cm s ,
approximately 1 part in 1000. The practical unit is the mgal; 0.01 mgal deviations (1 part in 108 can be measured with modern instrumentation).
Gravity varies with:
1. latitude for 2 reasons: a) Earth is oblate sphere, bulges at equator about 20 km, so diferences to center of mass; b) but
more important, component of centrifugal force opposing gravity, strongest at equator. The net efect is that gravity is
smaller at equator than pole. The variation is ~0.81 mgal/km of N-S travel. By convention apply latitude correction to
move measurement onto standard geoid, level of constant gravity.
2. elevation of measurement station. This also changes the distance from the center of mass. Applying a free air
correction to sea level yields the free air anomaly. The magnitude of the correction is ~0.31 mgal/m.
3. density of materials affecting measurement. Air does not fill the space between the measurement station and sea level.
To compensate one applies the Bouguer slab correction replacing air with inferred material filling space between
measurement station and standard geoid. For density 2.67 g/cc the correction is ~0.11 mgal/m.
For more information on this topic, see Tom Boyd's Introduction to Geophysical Exploration pages.
The remaining variations are termed the Bouguer anomaly:
when negative, indicates an excess of less dense material beneath measurement station.
when positive, indicates an excess of more dense material.
In this stacked section across the Rocky Mountains of North America note the parallel behavior of the observed Bouguer
anomaly, in the lower panel, with the topography in the upper panel:
Figure 1-4. from (30)
The explanation lies in concept of isotasy or isostatic compensation. Crustal blocks are floating on more viscous material beneath. Continental
rocks are less dense, float higher and have deeper roots (the minimum in the Bouguer anomaly reflecting this excess of light material). In contrast
the oceanic crust is more dense, thinner, and floats at lower elevation (and so is covered with water!).
The response time for isostatic compensation is slow (for example once load of an ice cap is removed, underlying crust rebounds upwards); these
response times help characterize subsurface properties. See for example see W.R. Peltier, Postglacial variations in the level of the sea:
Implications for climate dynamics and solid-earth geophysics, Reviews of Geophysics. 1998; 36(4): 603-689.
III. Plate Tectonics
Movement of earth's surface reflects large scale convection of upper mantle. Compare and contrast these two views, the first very schematic, the
second showing a cross section illustrating active and passive continental margins and the arc environment behind the zone of subduction:
Figure 1-5. Motion of the lithospheric plates in the context of convection of the upper mantle. Note boundaries: constructive with upwelling
mantle and generation of the lithospheric plate; destructive with downwelling plate closing the convection cell; transform offsets.
From (2), Figure 5-3, pg 135.
Figure 1-6. From the right hand side of the figure: Genesis of the lithosphere at a mid-ocean ridge. Spreading causes the ocean basin to grow. A
passive margin between the ocean crust and continental margin is shown, i.e., the continent is also moving away from the ridge (the eastern
seaboard of the United States is an example). In the ocean basin to the left, the plate is subducted back into the asthenosphere. This is recognized
as a trench on the seafloor. Frictional heating along the plate boundary gives rise to arc volcanism (for example the islands of Japan), and a
secondary mantle circulation causing spreading in the back arc basin. From (3), Figure 19-3, pg 461.
Plate tectonics is a modern idea, taking hold as the central paradigm of earth sciences in the late 1960s. Lines of evidence developed at that time
included:
Earthquake Distribution. Relative motion at plate boundaries is origin of major earthquakes, thus expect the locations of these earthquakes to
mark plate boundaries:
Figure 1-7. (slide 17 of the NGDC Relief Globe Slides). Red dots are earthquakes of magnitude >5 since 1980; yellow lines are plate boundaries.
At zones of subduction, the depth to earthquake epicenters marks the dip of the plate returning to the asthenosphere:
Figure 1-8. From (5).
Magnetic lineations. The polarity of the Earth's magnetic field undergoes reversals on time scales of 10 to 10 years. (For more information on
the origin of the Earth's magnetic field and the present state of understanding of the origin of reversals can be found in Physics Today, 49(1), 1719.) The polarity of Earth's magnetic field is recorded in volcanic rocks as cool beneath Curie temperature:
Figure 1-9a,b. Production of magnetic anomaly patterns. From (41)
For example these magnetic lineations off our coast result from spreading away from the Juan de Fuca and Gorda Ridges during alternating
periods of normal and reversed polarity:
Figure 1-10. From (41).
The present configuration of the Earth's surface involves seven major plates and about 20 in all:
Figure 1-11. From (41)
Variable rates of spreading are evident in the map of crustal age below. Spreading rate ranges from 2 cm/y on Mid-Atlantic Ridge to ~20 cm/y on
fastest parts of East Pacific Rise.
Figure 1-12. This rendering and the data set available from Muller
Topography of the Ocean Basins

Relief
Hypsometric Curve
As the lithosphere ages it cools and contracts. On ocean basin scales this cooling is principally by conduction and oceanic topography can be
modelled, as we shall see, by conductive cooling of the lithosphere such that:
In this model, seafloor depth depends mainly on seafloor age, due to thermal cooling and subsidence. However considerable additional
topography is superimposed. Consider these three examples from a range of spreading rates, which help illustrate the role of subsidence,
magmatic supply and tectonics.
Fast Spreading--Southern East Pacific Rise from 13 to 23S
Medium Spreading--Juan de Fuca Ridge near 48N
Slow Spreading--Mid Atlantic Ridge near 22N
The Oceanic Crust
An Aside on Igneous Rocks In the Mantle and Crust
The upper most lithosphere is the oceanic crust, which is constructed of melt from the mantle which is intruded into preexisting crust and erupted onto the seafloor. The following cross section shows the three principal layers of the oceanic crust:
pillow and sheet lavas are volcanic products (rapidly cooled)
gabbros are plutonic products (slowly cooled)
dikes feeder zones for magma to rise to surface.
Figure 1-14. From (2), Figure 7-4, pg 211.

We will consider the structure of the crust in more detail when we discuss heat transfer from magma bodies and cooling rock
to hydrothermal convection cells.
Overview of Marine Sedimentation

The term sediment refers to material composed of particles that have settled to bottom of a liquid. Sediments are classified
by the processes involved in determining their physical and chemical character. Important distinguishing characteristics of
sediments include:
chemistry and mineralogy (reflecting origin of particles)
grain size(s) (reflecting energetics of transport, deposition)
degree of compaction and/or cementation (resulting from burial (physical) and diagenesis (chemical)).
The source particles can be either preexisting rock (detrital sediments) or prec ipitates formed in the solution from which they
settle (chemical sediments). Sources of marine sedimentary material
The material deposited in marine sediments has its origin on the continents:
rivers
o
suspended load: 180 x 10 g/y
dissolved load: 40 x 10 g/y (corrected for cyclic salt component)
wind
o
atmospheric dust: 5 x 10 g/y
Within the suspended load, the materials are those that make up the continents, particularly those that are resistant to
weathering or dissolve incongruently, f or example:
Eq 13-2:
These include quartz, feldspar, and clay minerals such as kaolinite and montmorillonite. These clay minerals are products of chemical weathering
of primary igneous materials and typically very fine-grained .
Transport of Sediments
Physical processes dominate at ocean margins, where they transfer particles eroded from the land to the sea floor.
Active sedimentation processes (where the sediment modifies the properties and behavior of the suspension) include mass wasting and density
currents. Such deposition tends to mask other sedimentary processes. The giant landslides surrounding the Hawaiian islands and the abyssal
plains of the Atlantic are striking deep-sea examples of active sedimentation. Because active sedimentation depends on gravitational energy, it
does not extend seaward of the trenches along convergent plate boundaries (but can affect the entire ocean basin off passive margins).
Passive sedimentationprocesses are those in which the sediment is carried by but does not modify the normal thermohaline circulation.
Examples:
Hemipelagic sedimentation, where it appears that fine sediment moves along isopycnal surfaces high in the water
column. This process is poorly understood (for example, the role of internal waves in resuspending sediment, the role of
squirts and jets in transporting sediment ofshore and the role of biota in sediment deposition have yet to be
quantified). Hemipelagic sedimentation produces a fringe of terrigenous (land-derived) deposits up to a few hundred
kilometers wide around land masses. These deposits are draped uniformly over the sea floor topography
"Drift" deposition along the path of bottom currents. Prominent in the high latitude North Atlantic along the flanks or
ridges. Sediment in transit to drift deposits creates a near-bottom "nepheloid layer." Paleoceanography, 9 (6),
(December 1994) has a dozen papers on drift deposits.
Eolian (wind transported) sedimentation. Prominent where major wind systems cross semi-arid source areas
(ephemeral lakes) or active ash-generating volcanoes. Prominent in the North Pacific (Figure MS-1), North Atlantic and
Arabian Sea
Figure MS-1. Quartz concentration in North Pacific sediments (carbonate-free basis) showing efects of eolian transport
in westerlies and northeast trades.
Ice rafting. Restricted to high latitudes, but can carry coarse particles into the subarctic gyres, far from land.
Sediments formed by physical processes have distinctive acoustic signatures of military interest. Hence they have been much studied during the
past 50 years.
"Passive" sediments record the history of deep currents, volcanism, aridity, wind trajectories, and iceberg abundances and trajectories. Drift
deposits, which can accumulate at hundreds of meters per million years, yield some of the highest resolution paleoceanographic records.
Biological processes dominate sediment formation in areas of high productivity that receive little terrigenous material. The equatorial Pacific and
Southern Ocean are examples.
Riverine dissolved material is distributed by ocean circulation and mixin g. The principal pathway for particle production is
primary production by a variety of phytoplankton and zooplankton. Inorganic debris from dead organisms becomes the source
for underlying sediment oozes. Most of this material is either calcium carbonate or amorphous silica:
CaCO
foraminiferal ooze
o
organism: forams, a protozoan
most common source is genus Globigerina

o
mineral form: calcite
coccolith ooze
organism: algae coccolithophoridae
mineral form: calcite
pteropod ooze
o
organism: planktonic molluscs
mineral form: aragonite
SiO
diatom ooze
organism: algal diatoms
mineral form: opal
radiolarian ooze
organism: radiolaria (protozoans)
mineral form: opal
(The images are local copies from the WWW server at NGDC, National Geophysical Data Center)
Benthic organisms also modify the historical record by actively mixing ("bioturbating") the most recently deposited sediments. Bioturbation,
which is addressed later in the course, is effectively a low-pass filter that supresses or eliminates records of events that create layers of sediment
thinner than the depth of mixing. Rapidly deposited or anoxic sediments provide the only deep-sea records capable of resolving events shorter
than about a millennium.
Chemical processes dominate sedimentation only in deep, low productivity areas shielded from terrigenous material. Precipitates from
hydrothermal solutions emanating from mid-ocean ridges are prominent along the flanks of the East Pacific Rise in the South Pacific. Authigenic
deposits (formed by the very slow precipitation of oxyhydroxides and silicates? from normal seawater) form distinctive sediments in the central
South Pacific as well as ferromanganese nodules (Figure MS-2) and crusts on any surfaces where other sediments are absent or accumulating
extremely slowly.
Figure MS-2. Deep sea ferromanganese nodules on the floor of the South Pacific Ocean (individual nodules are 5-10 cm diameter).
Distribution of Sediments
The physical processes afecting sediment distribution are reflected in this map of sediment thickness:
Figure 13-3, from NOAA/NGDC, additional information is available concerning the underlying database Distribution of Recent (Surface)
Sediments
The global distribution of surface sediment types is shown here:
Figure 13-4, from (2)
Except for areas near to the continents, which are dominated by hemipelagic contiental margin sediments and muds (and at high latitudes, glacial
marine sediments), the predominant sediments in the deep ocean are biogenic oozes and/or fine-grained clays.
The distribution of biogenic sediments responds to oceanic conditions in three ways:
controls on primary productivity, especially those related to the intensity of upwelling
The particle compositions exhibits increasing silica/carbonate ratio as upwelling rate increases; there are also regional
efects (temperature, light) on species distributions
corrosiveness of deep ocean waters to (settling particles and) surficial sediments
Waters at all depths are undersaturated with respect to solid amorphous silica; the driving force for dissolution
decreases as deep water ages and develops higher dissolved silica concentration. (Because waters at all depths are
undersaturated, organisms must spend energy to precipitate amorphous silica.). The siutation is more involved for
calcium carbonate, with surface waters supersaturated and deep waters undersaturated.
topography
Most of the dissolution of biogenic debris occurs at seafloor. Thus the topography influences the temperature and
pressure conditions under which dissolution occurs.
We can illustrate the interaction of these three factors in generating a vertical sequence of sediment types lying on basement
rock created on the East Pacific Rise somewhat south of the equator.
Near to the ridge at which the basement rock is created, metalliferous sediment (of hydrothermal origin) dominates the supply of material. Once
away from this source of sediment, carbonate material raining from above accumulates. As the plate subsides, the dissolution of carbonate
increases such that this material is no longer preserved and only residual clay material accumulates (at a much lower sedimentation rate). As the
site passes beneath the equator, the rate of supply of biogenic silica exceeds the rate of dissolution and it accumulates. Further north beyond the
equatorial band of productivity, red clay dominates again.
Stratigraphy
Sediments place events in time. Their study from this perspective is termed stratigraphy: age relations of rock strata (layers). The objectives of
stratigraphic studies are to
establish markers of a sequence of events and thus relative ages of events,
correlate these markers of events either regionally or globally, and,
assign absolute ages (the basis of these age assignments is radiometric dating of materials) or a chronology to these
events.
The markers first employed were the skeleta of organisms preserved in rocks and so the description of geological time is
closely linked to the evolution of the plant and animal kingdoms.
Earth's history is divided into ages:
Figures 13-6ab
In ocean sediments, the Cenozoic and Mesozoic ages are represented. These ages can be further divided into periods and epochs and stages:
Figure 13-7
Figure 13-8
(Figures 13-6, 7 and 8 courtesy of the University of California at Berkeley Museum of Paleontology, their copyright notice. Their web server
offers an extensive discussion of geological time in relation to evolution of the species.)
In establishing time scales for ocean sediments, stratigraphic correlations can be made in a number of ways:
lithostratigraphy
boundaries between rocks of diferent type or character (variant: seismic stratigraphy recognizing these boundaries
from seismic reflection profiles)
biostratigraphy
based on fossil content (presence, absence of particular indicator species)
chronostratigraphy
based on ages of materials as established by radiometric methods
magnetostratigraphy
based on magnetic reversals recorded in sediments
stable isotope stratigraphy
based on oxygen isotopic composition of carbonate material, a signal dominated by global state of ice storage on the
continents during glacial times
For study of ocean sediments the most important tools are biostratigraphy (applied throughout the Mesozoic and Cenozoic),
magnetostratigraphy (important for the late Cenozoic sediments (past several My)), and oxygen isotope stratigraphy (for
Pleistocene sediments, last 1.8 My)
Indicators of Changes in Ocean Conditions

Our interest in Pleistocene sediments is derived from the record of climate change they contain. There are two important classes of signals:
Deposition Rates
Signals indicating the extent to which material is deposited and preserved. These include absolute rates of net
deposition and indices of changing proportions of materials (e.g., Si vs. C, specific species)
Chemical Signatures
Signals involving properties of sedimentary materials which reflect conditions of original deposition, either directly or as
a proxy for some other property. Some examples are oxygen isotopic composition reflecting ice storage on the
continents (and to a lesser extent temperature of deposition); Sr/Ca reflecting the temperature of deposition (and to a
lesser extent upwelling), and Cd/Ca reflecting upwelling.
Background Information on Sediment Classification

Summarizing the classification of sedimentary rocks:
Figure 13-1
Figure 13-2, from the Geology 202 Pages at University of British Columbia
Chemical Sediments, both allochems (precipitates formed in basin of deposition) and precipitates (precipitates formed in
deposit itself)
Limestone
Chemical
Composition
Minerals
Dolomit
e
Iron formation
CaCO3
CaMg(CO Fe silicate oxide

carbonate
3 )2
calcite,
aragonite
dolomite
Evaporite
NaCl, CaSO4
hematite, limonite, halite, gypsum,

siderite
anhydrite, other salts
Marine sediments are further classified:
median grain size <5m: pelagic deposit (clays and oozes)

o
biogenic component <30%: calcareous or siliceous clay
biogenic component >30%
calcareous component dominant
carbonate component <67%: marl ooze
otherwise: chalk ooze
siliceous component dominant: diatom or radiolarian ooze
Chert
SiO2
Organic Phosph
s
ate
C
Ca3(PO4)2
opal, chalcedony, coal, oil,

apatite
quartz
gas
median grain size >5m: hemipelagic deposit (muds)
Conduction and the Conductive Plate Model

At the scale of an ocean basin, heat transfer from the seafloor is dominated by conduction. Fourier's Law states that the flux
of heat conducted across a plane surface is the product of the thermal conductivity, a property of the material through which
heat is being transported, and the temperature gradient. In one-dimension:
Eq 3-1:
(conductive heat flux, q) = - (thermal conductivity, k) x (temperature gradient, dT/dz)
The negative sign reflects that heat is transported in the direction of lower temperature. For a historical discussion of Fourier's Law and its
influence on the description of other physical systems, see (44). In dimensional terms:
Dimensions for Heat Transfer Quantities

recent literature (SI)
older literature
flux
Wm
cal cm s
thermal conductivity
Wm K
cal cm s C
gradient
Km
C m
The pratical unit from older literature is the HFU (heat flow unit). 1 HFU (heat flow unit) = 41.9 mW m
conductive heat flow is about 2 HFU.
= 1 cal cm s The global average
The thermal conductivity (or equivalently the specific thermal conductivity, (also called the thermal diffusivity) which includes the density and
heat capacity of a material) varies considerably, over about four orders of magnitude in various solid materials:
Table I. Thermal Conductivities of Common Materials

k [W m C ]
=k/( C )
[10-8 m2s-1]
thermal conductivity
specific thermal conductivity or thermal diffusivity
.025
1938
water
0.6
14
ice
2.1
104
aluminum
237
9975
copper
390
11161
fiber glass
0.04
44
surficial ocean sediment
0.9
20
basaltic ocean crust
2.9
77
air
upper mantle
quartz
saturated sand
3.1
80
141
2.7
102
Using observations of the conductive heat flow we can estimate the temperature increase with depth into the oceanic lithosphere. At 100 My
crustal age the heat flow is ~1.1 HFU = 46 mW m , so:
Eq 3-2:
q=-k(dT/dz)
Eq 3-3:
46 x 10 W m = 3.1 W m C (dT/dz)
Eq 3-4:
(dT/dz) = 15 x 10 C m
Thus, at 100 km depth, T ~1500 C. (This is an overestimate, as even after 100 My, the lithosphere is continuing to cool and so the profile is not
strictly linear.)
This approach is reversed in order to make heat flow measurements: the temperature gradient is measured with a probe consisting of a number of
thermistors which is inserted into sediment; the thermal conductivity is measured (by measuring the response after a thermal perturbation) or
estimated from other data; the product yields the heat flow.
To make use of Fourier's Law to understand the rate of change of the temperature of a material, we need to establish a thermal balance for a small
volume element. Again in one-dimension, we balance the fluxes of heat through the two surfaces and the internal production of heat per unit
volume, H, to determine the temperature change in an interval of time dt:
Figure 3-1
Conservation of heat requires that:
Eq 3-5:
Rearranging equation 3-5 and taking to limits:
Eq 3-6:
Is Internal Production of Heat Important Within the Lithosphere?
Well away from the ridge axis where the time rate of change becomes small, it becomes appropriate to consider the steady-state characteristics of
equation 3-6. The goal is to examine whether the internal production of heat, H has a significant influence on the temperature distribution.
At steady state:
Eq 3-7:
Equation 3-7 is solved by integrating twice to find:
Eq 3-8:
The two constants of integration are found by specifying the temperature at the sediment surface:
Eq 3-9:
and the surface heat flow
Eq 3-10:
Combining equations 3-8, 3-9 and 3-10, we find:
Eq 3-11:
If H=0, the temperature increases linearly with depth. For positive internal production of heat, there is a negative deviation from this linear trend
which grows as z squared. (The deviation is negative because of the boundary conditions we have chosen; as H increases more of the surface heat
flux is derived from internal production of heat and less from the flux of heat from below). In the case of the lithosphere, is this deviation
important? Because the elements U, Th and K are partitioned into the liquid as the mantle melts, considering the radiogenic production of heat in
the basaltic crustal layer represents a worst case (i.e., the mantle must have a lower content of U, Th, and K). This production is 7.8x10 W m .
At 6 km depth, for surface heat flow of 45 mW m and with this value of H the temperature deviation is 0.97C from the value without internal
production, 93C. We conclude that we can generally neglect internal production of heat.
Thus for application to conductive heat loss from the lithosphere we neglect H and equation 3-6 becomes:
Eq 3-12:
Conductive Plate Model--Boundary Layer Approach

In the absence of internal production of heat:
Eq 3-13:
Imagine that at the axis of spreading that material is held at the temperature of melting, in other words an appropriate initial condition is that at all
depths:
Eq 3-14:
(While we will not develop it explicitly, 'temperature' here means 'potential temperature'). For now also imagine that the Earth is infinitely thick,
i.e., the geometry is that of an infinite half space and so at great depth:
Eq 3-15:
Finally the upper surface will be will be held at the temperature of the deep ocean:
Eq 3-16:
Omitting the details of solution, we find that:
Eq 3-17:
(Note the k inside the root operator should be kappa.) We can use equation 3-17 to calculate the temperature field of an evolving lithosphere.
The error function is defined by the integral:
and exhibits the behavior:
To illustrate the temperature field, we adopt T =1400 C, the approximate temperature of the liquidus. For depth we will work in units of km,
and for time in units of My so that the appropriate thermal diffusivity is 25 (km )/(My). We will build a grid for 0 to 100 km depth and 0 to 100
My of crustal age in steps of 5 km and 5 My respectively. The solution at t=0 has to be taken directly from equation 3-2 because of the singularity
at t=0 in equation 3-5. The calculations are done with the MATLAB script, plate.m:
kappa=25;
Tmelt=1400;
z=0:5:100;
t=0:5:100;
for I=1:1:21
T(I,1)=Tmelt;
for J=2:1:21
T(I,J)=Tmelt*erf(z(I)/(2*sqrt(kappa*t(J))));
end
end
cs=contour(t,-z,T);
clabel(cs)
(New to Matlab? Here is a link to an annotated version of the script.) This script displays a contour plot of the temperature
field:
If we think of the base of the lithosphere as corresponding to a particular isotherm, say the 1200 C isotherm which is approximately the
temperature of the solidus, we can trace the growing thickness of the lithosphere as the crust ages. However, because of the semi-infinite
geometry adopted, it grows to an unreasonable large thickness, as compared to the constraints provided by seismic studies of the oceanic
lithosphere. The temperature gradient (and the conductive heat flux) decreases, most rapidly at first and then more slowly.
While there are some problems evident in the model: specifically
1. the temperature of the seafloor at the ridge axis is not T m and

2. the thickness of the lithosphere does not grow without limit
it is appears to be reasonable from a qualitative standpoint and so we will use it as a starting point for considering the
relationships among heat flow, topography and crustal age.
Conductive Heat Flow versus Age
By taking the derivative of equation 3-17 to obtain an expression for dT/dz evaluated at z=0 and substituting into equation 3-2 we find that
(details):
Eq 3-18:
The heat flow predicted from the boundary layer model decreases with an inverse proportionality to the square root of age.
For now (we will later refine these parameters) we can estimate the relationship using a value of T
data given in Table I:
Eq 3-19:
480 mW/m is ~12 HFU.
of 1400 C and the thermal conductivity
Conductive heat flow versus age, from (43)
HS is our infinite half space model, PSM is the original Parsons Sclater model (8), GDH1 is the more recent Stein and Stein ((43)) fit to the
global data set. (HS and PSM can only be distinguished on very old lithosphere.)
Lithospheric Subsidence and Sea Level

Topography versus Age
Thermal contraction of the lithosphere increases its density and so isostatic adjustment will result in subsidence of old
lithosphere as it ages. Consider this model:
Figure 4-1
If we compare the masses (we should actually consider the weights, but we will take the gravitational acceleration as constant) of the two
columns A (at the rift axis) and B (at some arbitrary age), we can derive a relationship for the subsidence, w.
Mass of column A
Eq 4-1:
Mass of column B
Eq 4-2:
where w is the water depth below the ridge, z is the thickness of the lithosphere, and is density. The subscripts w, L and m denote water,
lithosphere, and mantle (i.e., asthenopshere), respectively. Equating M and M and rearranging:
Eq 4-3:
The left term is positive, since the lithosphere is cooler and denser than the asthenopshere, and represents the effect of thermal contraction; the
second term is negative and represents the effect of replacing denser mantle with lighter seawater as the seafloor subsides.
The specific volume, v, is a state variable v=v(T,P) and so a variation of v can be represented by:
Eq 4-4:
where is the isothermal compressibility and is the volumetric coefficient of thermal expansion. For an unconstrained material (e.g., a material
undergoing isostatic adjustment) the pressure does not change when the temperature and volume change, i.e., dP=0, and:
Eq 4-5:
so that
Eq 4-6:
Therefore the change in density associated with thermal contraction can be expressed, using plastic, uncooled mantle as a datum (the M
subscript) as:
Eq 4-7:
The temperature distribution in the lithosphere is taken from the conductive boundary layer model for the plate (rewriting the solution in terms of
erfc(x)=1-erf(x)):
Eq 4-8:
Substituting equations 4-7 and 4-8 into equation 4-3 and solving for w we find that:
Eq 4-9:
Because the error function complement term in the integral of equation 4-9 approaches zero at the base of the lithosphere, the upper limit of
integration can be changed from z to infinity without introducing appreciable error. (This is equivalent to saying T-T approaches zero at the
base of the lithosphere, see equation 4-7). With this change of limit and a variable substitution:
Eq 4-10:
equation 4-16 becomes:
Eq 4-11:
The definite integral in equation 4-11 can be evaluated:
Eq 4-12:
so that:
Eq 4-13:
Since the depth to the seafloor is of the form:
Eq 4-14:
the depth-age relationship can be expressed as:
Eq 4-15:
We can make an estimate of w using a value of T of 1350 C, the thermal conductivity of mantle rock (Lecture 3), and the coefficient of
thermal expansion for the mantle:
Eq 4-16:
This relationship is apparent in observations as seen in this summary of seafloor topography versus age, from (43)
Figure 4-2. Seafloor topography versus age compared to models of thermal subsidence. HS is the infinite half space model, PSM is the original
Parsons Sclater model ((8)), GDH1 is the more recent Stein and Stein ((43)) fit to their complilation of global data.
Variants on Boundary Layer Model
Three refinements to the boundary layer approach have been considered in the literature and applied in variations combinations (see Parsons and
Sclater (8) for references):
1. extending the model to two dimensions and explicitly considering the horizontal advection of heat by the plate:
Eq 4-17:
2. With T(x=0)=T , the heat flux was unbounded at the axis of spreading. Refine the boundary condition at the ridge axis
so that heat leaving the axis by conduction and advection matches the heat carried by injected material (i.e., apply a
flux condition)
3. Establish a geometry for the model domain such that there is an asymptotic plate thickness and fix the temperature at
that lower boundary.
We will not explicitly consider the solutions corresponding to these refinements except to note that all of them exhibit the
"one over root t" relationship of heat flow to age at ages from "near to the ridge" (age >1-2 My) to far away from the ridge
(age <70-90 My). Near to the ridge these discrepancies are not particularly important as convective heat loss dominates and
a conductive model alone is inadequate. Far away from the ridge, the diferences are of more interest as measured heat flow
does not continue to approach zero, but rather approaches an asymptotic value, an important constraint on possible models
for secondary convection of the upper asthenosphere.
In the past decade, Stein and Stein (43) have undertaken a substantial effort to revisit various aspects of the plate model in order to take full
advantage of the data sets that have been obtained since the early Parsons and Sclater work. More information is available at the Seth Stein page.
Material properties can be extracted from best fits of improved models (Parsons and Sclater, 8, Stein and Stein (43)):
Fits to Parameters of the Plate Model

North Pacific (8)
North Atlantic (8)
Global (43)
asymptotic lithospheric thickness (km)
125
128
95
boundary temperature (C)
1333
1365
1450
thermal expansion coefficient (C )
3.3 x 10
3.1 x 10
3.1 x 10-5
The agreement is quite satisfying in that each of these three parameters can be related to other observations (the thickness of the lithosphere to
seismic studies of upper mantle structure, the boundary temperature to laboratory observations and thermodynamic calculations of melting of
mantle rock, the thermal expansion coefficient to studies of physical properties of mantle rock).
Figure 4-3. Fits by Stein and Stein ((43)) to various models for the thermal behavior of the lithosphere.
Subsidence and Sea Level

Converted from a Mathematica notebook, subsidence-sealevel.nb
The mean age of an ocean basin is inversely related to spreading rate: slower spreding increases the mean age, thus there is more subsidence, and
sea level falls. How much? The mean subsidence wbar of a basin of width W with spreading rate u is:
In[1]:=
Out[1]=
In[2]:=
Out[2]=
Using km as the distance unit and My as the time unit, we can first compute the subsidence for a basin 6000 km wide as a function of spreading
rate
In[3]:=
Out[3]=
In[4]:=
Out[4]=
and then calculate the change in mean subsidence brought about by a factor of two decrease in spreading rate
In[5]:=
Out[5]=
This change occurs over a time scale ~100 My so that the rate of change is quite small on human time scales (less than a millimeter per century)
Mantle Flow and Crustal Generation

In the next several lectures we will consider a variety of interlinked phenomena that create the oceanic crust, supply heat and
basaltic magma to the zone of crustal accretion, and sustain hydrothermal activity along the mid-ocean ridge. This lecture
sets a framework for considering in more detail:
melting of mantle rock
heat transfer from magma bodies
convection in liquids and in porous rock
fluxes of hydrothermally-derived heat and material
Mantle Circulation and Melt Generation

This schematic of mantle convection provides a plate scale view of the system:
(Figure 5-1, from (17))
Melt that reaches the crust has its origin in the mantle, with melting occurring as ascending mantle material adiabatically decompresses. In this
model
1. asthenospheric material upwells beneath the ridge axis
2. melting is supported by solid material undergoing adiabatic decompression, crossing the solidus at depth z
solidus is the P/T condition at which melting begins for a material of a given composition).
3. melt migrates upward to form the crust
. (The
4. melt is supplied to form the crust in a narrow zone near to the axis of spreading; on a time-averaged basis this supply
must support V z of crustal generation
5. the chemistry of basalts suggests that melting is incomplete, that the magmas are diferentiated from their mantle
source. The degree of melting that can be accommodated based on compositions of basalts is F =20 (+/-10) %
6. The variables in this model that are poorly constrained are the velocity of the upwelling mantle, V , and the width of
the zone of upwelling, x .
The rate of melting has to balance crustal formation (all melt rises to the surface, because only cooling will stop it), and so:
Eq 5-1:
Eq 5-2:
The simplest model for mantle convection involves a diverging plate overlying a semi-infinite medium of constant viscosity. This model predicts
that V ~V . Thus the zone of upwelling is predicted to be rather wide, about 30 km. One of the puzzles being addressed now in studies of mantle
circulation and its relationship to crustal construction is how to focus the melt from this wide zone of melting into a narrow zone of volcanism.
To understand the parameter zsol, we can examine the relationship of an adiabat to the solidus for mantle rock. As a function of temperature and
pressure:
Figure 5-2, from (18))
The solidus for mantle material has a slope of about 12C/kbar while an adibatic path in the mantle has a slope of about 1C/kbar, hence these
eventually cross and melting begins. Once within the field of melting, the effect of the negative heat of melting is to cause the temperature
decrease with decreasing pressure to be steeper than the adiabat. The temperature of the mantle controls both the depth at which melting begins
and the fraction of melt generated.
These thermodynamic relationships can be superimposed on the physical model:
When the mantle is cold, melting is initiated at shallower depth and over a narrower zone. A thin crust is generated and the depletion in the
residual mantle is minimal. When the mantle is hot, melting is initiated at greater depth and over a wider zone. A thick crust is created and there
are greater chemical changes in the residual mantle which extend to greater depth.
Crustal Magma Chambers
Melted material migrates upward from the mantle to form the oceanic crust. When cooled sufficiently from above, this melt
will pond and form a liquid pool or magma chamber. Cracking of overlying rock will open conduits (usually axis-parallel planar
dikes) to feed extrusive basalt flows at the seafloor. Whether a magma chamber is usually present beneath the seafloor is
thought to be strongly dependent on spreading rate. On the fast spreading East Pacific Rise there is extensive evidence from
multichannel seismic studies for an axial magma chamber.
This is thought to be a steady state feature, found on 60% of the ridge between the Siqueiros Fracture Zone at 830'N and 1330'N. Its width
across axis is 1-1.5 km and height 100-500 m. The low velocity zone surrounding this body is thought to be material with at least 3% melt. In
contrast there is no evidence of the existence of persistent magma chambers on the Mid-Atlantic Ridge, although such bodies are likely present
episodically.
Heat Transfer at a Mid-Ocean Ridge
Convective cooling occurs because of movement of fluid in the upper crust--through fractures and cracks in basaltic rock,
along grain boundaries with the rock mass, and in the pore spaces of sedimentary materials. Thus the depth to which
convective circulation occurs is closely tied to existence of pore space. Pore space exists because rock has cooled, so
understanding the overall process involves not only the mechanics of fluid transport but knowledge of the heat source
(cooling magma and solid rock), how pore space is developed as rock cools and how heat is transferred from solid to liquid.
We can sketch the scales involved:
Figure 5-5
The basaltic crust is emplaced as melt in a narrow zone at the ridge axis, at a temperature near 1300 C. At the ridge axis, high temperature
circulation of fluid is established, with fluid temperatures reaching 350-400 C. Thus as the crust leaves the zone of generation, it is cooler than it
would be by conduction alone. As we will see, a reasonable estimate of the mean temperature of the crust a few kilometers off axis is 200 C.
Thus on a time scale of 100 ky (5 km distance with 5 cm/y spreading), the melt has both solidified and cooled considerably. The heat to be lost in
this zone includes both the latent heat of crystallization of the melt and the cooling of the solid rock mass, so that the total heat loss is:
Eq 5-3:
On a global scale, the rate at which ocean crust is generated is 5.4 x 10 g/y so the convected heat is about 2.3 x 10 cal/y. This is about 1/2 the
deficit in observed conductive heat flow relative to the conductive plate model. (Note that this calculation does not account for conductive heat
transfer in this zone.)
On a local scale, we can consider as an example the Endeavour Segment, northern Juan de Fuca Ridge on which the spacing of major sites of
hydrothermal discharge is of order 2 km. The heat loss from a 2 km section of ridge should be of order 1.6 x 1015 cal/y:
(crustal depth) x (cell spacing) x (full spreading rate) x (density) x (heat transfer)
6 km x 2 km x (2x5 cm/y) x 3 g/cm x 433 cal/g
while individual vent fields here discharge 0.6 GW = 4.5x1015 cal/y, i.e. the instantaneous flux is 3 times as great as the long term geological
average. This implies that venting must be short-lived and episodic.
Magma bodies can serve as a source of heat for hydrothermal circulation. Schematically:
Figure 5-6
But how quickly does this molten material solidify?
Is there a persistent magma chamber that continues to provide heat to fluid flowing by it?
Or is there a cycle of magma injection and solidification with the heat transfer from rock to fluid by penetration of the
fluid into uncooled cracking rock?
To resolve this question we will need to compare the convectively driven flux of heat in hydrothermal fluids discharging from
the seafloor with heat transfer from the upper surface of the magma body. We will find that it is difficult to rationalize
observed rates of convective heat transfer with a persistent magma chamber and so we will continue on to examine the
character of the boundary between cracked and uncracked rock in terms of its thermal characteristics.
With this background as a framework we can ask a series of questions to consider:
Why is the extent of melting of rising mantle not complete? Why does it produce basalts of a narrow range of chemical
composition?
What is the nature of convective heat transport and what roles does it play in heat transfer from magma chambers and
through hydrothermal systems?
What are the fluxes of heat and of chemicals, how do these vary in space and time, and how do they interact with the
overlying ocean?
Phase Diagrams
Review of Chemistry and Mineralogy of Oceanic Rocks
Phase diagrams depict phase relationships within multi-component systems. Phases are physically separate regions of homogeneous chemistry.
Possible phases include a liquid phase (multiple liquid phases if the liquids are immiscible), a gas phase, and multiple solid phases. Solid phases
may exhibit solid solution, i.e., have a range of compositional variation.
To introduce phase diagrams, we will consider the characteristics of a series of phase diagrams of increasing complexity, beginning with a binary
mixture of diopside and anorthite, at 1 atm pressure.
Figure 6-1
Some important elements of this phase diagram include its:
solidus: boundary below which no liquid phase exists
liquidus: boundary above which there are no solid phases
two portions of the liquidus in which one of the two solids plus a liquid (of that composition) is present
eutectic point: a point at which both of the solids and a liquid (three phases) coexist
general shape: freezing point depression by foreign substance
reflects Gibbs phase rule, at fixed pressure for a two component system, three phases (two minerals and liquid) coexist
at a fixed temperature, two phases coexist along a particular T-composition curve
As an example consider the path of melting, starting from low temperature, of material of composition An . As heat is added
the two solid phases are present until a temperature of 1274C is reached. At this point, melting produces a liquid of
composition An . The system remains at this temperature until all the diopside present in the original rock melts. Why?
Because the eutectic is the only point at which three phases (two solids plus the liquid) can coexist. The mass balance for
melting at the eutectic point can be used to solve for the composition of the residual solid during this portion of the melting
process:
Eq 6-1:
Eq 6-2:
In these and the following equations f is the fraction of liquid and is the composition with the subscripts E, s and l denoting the eutectic liquid,
the solid and the liquid, respectively. Melting will continue until the composition of the residual solid reaches the value 1, i.e., the diopside
component has melted completely and the residual solid is pure anorthite. Once this happens the system is no longer restricted to the eutectic
point and the heat applied to system will again increase temperature. For this example, the critical fraction of liquid is 0.69.
As temperature rises again, once this critical fraction is reached, the liquid composition follows the liquidus until the liquid is of starting
composition An . The position of the liquidus indicates the composition of liquid at a particular temperature. Thus when one of the two solids is
present in equilibrium with a liquid, mass conservation requires that for an arbitrary initial bulk composition :
Eq 6-3:
Eq 6-4:
In geometric terms:
Figure 5-2. (the right hand lever should be labelled l, not 1)
and so this mass balance is the basis of the lever rule, i.e.
Eq 6-5:
We call the situtation just described closed system, equilibrium melting. What if melt were continuously separated from residual solids instead?
We call this open system melting. Until the critical fraction of liquid is reached, the system behaves in exactly the same way as before--the liquid
composition is determined by the eutectic. However once diopside is gone only the phase anorthite remains which melts at 1553C, a gap of
279C. Obviously the energetics of melting for closed and open systems are quite different.
A number of solids exhibit solid solution, for example plagioclase, so that there is compositional variation in a single phase. The phase diagram
for plagioclase is:
Figure 6-3
The solidus is no longer a horizontal line--the solid composition at equilibrium depends on the bulk composition of the system composition and
the temperature. For a given bulk composition, one finds the solid and liquid proportions by applying a lever ruler:
Figure 6-4
As an example consider the cooling of material of composition An . The first crystals form at 1450C, of composition ~An . With continued
cooling the liquid composition follows the liquidus and the solid composition follows the solidus until the original liquid is fully crystallized at a
temperature of about 1333C. The last liquid composition is about An .
It would be unusual at best for a natural solid to follow this closed system, equilibrium behavior. The solid initially formed will either be
segregated by mechanical settling or protected from further reaction with the liquid by crystals formed later on the surface (a phenomenon called
zoning). We call this path the path of fractional crystallization. In the following figure focus on the curves labelled TSC. These represent the
integrated total solid composition obtained as a function of temperature when the solid produced at any instance is in equilibrium with the liquid
composition at that temperature--more calcic solids produced early in the crystallization process do not back react with the liquid. The endpoints
for the TSC are known, the initial value determined by the gap between the solidus and the liquidus and the final state by the bulk composition of
the original liquid. The key differences from equilibrium crystallization are that the last liquid is much more sodium rich and liquid remains in a
system to a much lower temperature.
Figure 6-5
Ternary Systems
Multi-component phase relationships can be visualized on a triangular ternary plot. Consider an example involving the three components
diopside, anorthite and albite (i.e., there are two phases: diopside and a solid solution of plagioclase).
Figure 6-6
There are several features of this diagram to note:
concentration scale: any point adds to 100% of the three components, measured from base (0%) to vertex (100%)
contours: are temperature of the liquidus for that composition
the boundary in the interior separates fields for diopside + liquid, plagioclase + liquid. Along this boundary (called the
cotectic), diopside + plagioclase + liquid coexist.
along edges: can compare to the binary diagrams we considered earlier: diopside-anorthite and plagioclase.
As an example of the application of this diagram, consider the cooling of material of a composition from within the diopside field, say of
composition Di An Ab . As the initial liquid begins to cool, the liquidus is reached at 1300C and diopside begins to crystallize. The
composition of the residual liquid moves straight down (away from the diopside vertex, because only diopside is crystallizing). When the cotectic
is reached, plagioclase also begins to crystallize with the liquid composition locked on the cotectic. Of what composition is the first plagioclase
formed? This information can't be read from the diagram; we need additional information. The answer is about An for this composition, about
30% higher than liquid, consistent with what we might have guessed from the phase diagram for plagioclase alone. Upon further cooling, the
liquid composition follows the cotectic. Travel down the cotectic stops when the solid has composition Ab=An, i.e., a liquid about An , about
30% lower than the final solid.
Contrast the behavior of plagioclase alone and the situation just considered, that of diopside-saturated plagioclase
Figure 6-7
At 1200C, for plagioclase alone the liquid is in equilibrium with An crystals contrasted with An crystals for the diopside saturated system.
There is a much lower temperature of melting across the entire compositional range and as illustrated in the upper part of the figures, a smaller
gap between liquid and solid compositions except at very high An content.
Returning to the question of crystallization paths in the system diopside-plagioclase, suppose that instead of starting in the diopside field we start
in the plagioclase field. The liquid composition will move directly away from the solid plagioclase composition at that point in the crystallization
path; see the diopside-plagioclase phase diagram. Thus, the liquid composition follows a curved path to the cotectic, with the tangent of the curve
pointing down to the solid composition at that point in the crystallization sequence.
Petrological Relationships in Generation of the Ocean Crust

As a simple model for melting relationships of mantle material we will consider the system forsterite-diopside-quartz. These
three minerals represent about 85% of the chemical components present in peridotite and include three critical minerals of
the natural suite: forsterite, enstatite and diopside. (review: what is peridotite? what are forsterite, enstatite and diopside?)
The principal limitation of this representation is that it does not include iron.
The subsolidus phase relationships for this system are shown here:
Figure 7-1. From (18) (the Hess paper).
The tie-line between the diopside vertex and the MgSiO point on the Fo-Qtz boundary represents solid solution of these components. The
naturally occurring diopside is of the composition marked a and the enstatite is of the composition marked b. Note the immiscibility gap
(indicated by the dashed line) in this solid solution of pyroxenes. The diopside rich member is in the monoclinic crystal system while the enstatite
rich member is in the orthorhombic crystal system, thus they are often referred to as clinopyroxene (cpx) and orthopyroxene (opx), respectively.
The liquidus at 2 GPa (20 kbar, ~67 km depth) is shown here:
Figure 7-2
The bulk composition of mantle peridotite is at the point labelled P--in terms of the minerals Fo, Di(a) and En(b) it has the composition 53:31:16.
Figure 7-2a
How would melting proceed with addition of heat or adiabatic decompression? The first liquid composition will be at point L, providing that the
three phases coexist. This is a peritectic point, i.e., it lies outside the Fo+Di(a)+En(b) triangle. The melting reaction at the peritectic is:
.07 Fo + .93 Di(a) ==> 0.8 liquid + 0.2 En(b)
The melt is generated mainly at the expense of diopside but also minor amounts of forsterite; additional enstatite is produced. The liquid
composition stays at point L as long as the three phases are present, while the residual solid composition moves directly away from point L from
point P toward point R.
Figure 7-2b
Reaching R marks the point at which all Di(a) has been melted (the solid contains a combination of Fo and En(b).
The liquid composition then follows the Ol-En cotectic toward point G. Anywhere along this part of the liquidus, the intersection of the tangent
to this cotectic and the Fo-En(b) tieline gives the proportions in the liquid generated, via the lever rule approximately 6 parts En(b) to 1 part Fo:
0.15 Fo + .85 En(b) ==> liquid
Thus the residual solid composition follows the Fo-En(b) tie from point R toward the Fo vertex. When the liquid composition reaches the point
labeled G, the points Fo and P are in line and at this point no En(b) remains in the solid.
Figure 7-2c
The liquid composition begins to move toward the Fo vertex reaching its final composition P at 100% melting.
Figure 7-2d
To review we can watch the dance of the melting sequence.
In contrast to the equilibrium melting just discussed, fractional melting (remaining crystals are isolated from the liquid) generates a different
relationship between melt produced and the temperature, as shown in Figure 7-3 below. Note the large increase in temperature (heat supply)
required to undergo additional melting once diopside is all melted. In this way the amount of diopside in the original melt acts as a strong control
on the degree of melting that the mantle undergoes.
Figure 7-3.
The phase relationships we have been examining are pressure dependent, as this diagram of the liquidus at two different pressures shows:
Figure 7-4.
In this diagram, the arrows on the liquidus show the direction of lower temperature. As pressure is reduced, the peritectic moves to a lower
temperature (for the 20 kbar drop shown from 1640 to 1386 C) and to a more Di and Qtz rich composition. If melt produced at the 2 GPa
peritectic (point L in the earlier figure) is decompressed it will lie well into the Fo+liquid field. Thus when cooled, olivine will be the first phase
to crysallize. The liquid composition will move away from the Fo vertex, intersecting the olivine-diopside cotectic at a point up from the
peritectic. Diopside will then crystallize moving the liquid composition toward the 1 bar peritectic. When mantle is melted, the particular
composition of the liquid generated will depend on the P-T regime and in particular the point at which the solidus is reached.
Al is not a component within the system Fo-Di-Qtz. Its introduction via melting of garnet is an additional secondary effect as the mantle melts.
However its presence plays a much more important role in the crystallization history of the liquid as the basalts making up the crust are formed.
At shallow levels in the lithosphere, it is more relevant to consider the system Fo-Di-An (plagioclase (represented by its endmember anorthite in
this model is the principal aluminum-bearing phase in basalts):
Figure 7-5.
A primitive melt formed in the mantle would have a composition near point A. That liquid would first crystallize olivine, then anorthite (Al
bearing), then diopside. Petrographic studies of basalts reveal that crystallization of diopside never follows olivine directly, which restricts the
possible composition of the parent melt, i.e., point A cannot be much lower in the diagram. While we will not develop the topic in more depth, it
is through approaches of this kind that the melting history of the mantle is constrained. Important tools are the melting relationships affecting
major elements and the partitioning of trace elements during melting events.
Other Resources:
Mark Ghiorso's MELTS Software
Jon Snow's Oceanic Petrology. These pages include some MELTS laboratory exercises.
Convection and Rayleigh Criteria

To motivate the Rayleigh number, we consider a scaling argument from ((13), p. 208 f.). Imagine two horizontal plates,
separated by a distance h, with increasing temperature downward such that the temperature diference between them is
Consider a parcel of fluid of scale d (d x d x d) located between these plates.
Th.
Figure 8-1
If displaced upwards it will have a buoyancy force acting on it, i.e., its density will be diferent from its surroundings. The
buoyancy force is:
Eq 8-2:
where g is gravitational acceleration and is density. Further suppose that this density contrast is solely due to a contrast between its temperature
and the surroundings:
Eq 8-3:
where is the thermal expansion coefficient and T is temperature. Because the parcel is less dense than its surroundings, it will tend to move
upwards. A viscous force:
Eq 8-4:
will act against the buoyancy force. In this expression is viscosity, z is the vertical spatial coordinate, and t is time. When these two forces
balance:
Eq 8-5:
In addition, the parcel will be losing heat to its new surroundings by conduction, thereby reducing its temperature. The rate at which heat is lost
will be proportional to the surface area of the parcel. Thus temperature (and so the density term) is a function of time:
Eq 8-6:
where is the thermal conductivity. Combining 8-3, 8-5 and 8-6 and rearranging:
Eq 8-7:
How far can the parcel move in infinite time? We integrate from zero to infinity:
Eq 8-8:
to find that the distance traversed, z, is:
Eq 8-9:
The terms in the numerator of equation 8-9 have their origin in the buoyancy force and promote upward motion while increasing either viscosity
or thermal conductivity in the denominator limits upward motion.
To be able to convect heat, there must be movement across the distance between the two plates separated by a distance h in finite time and so:
Eq 8-10:
By scaling the fluid parcel to the plate separation, through an arbitrary factor f:
Eq 8-11:
substituting into equation 8-10 and rearranging it follows that:
Eq 8-12:
The left hand side of equation 8-12 is called the Rayleigh number:
Eq 8-13:
Lord Rayleigh showed through linear perturbation analysis (for a general discussion see (13), section 7.1.2 or the derivation in (11), section 618), that for thermal instabilities to grow in a fluid, the Rayleigh number would have to exceed a critical value (the f scale on the right hand
side). When:
Eq 8-14:
the temperature profiles between the two plates will be conductive, while when:
Eq 8-15:
a cellular convection will be established, displacing the isotherms upward in the zone of rising fluid and downward in the zone of downwelling.
The value of the critical Rayleigh number depends on the particular geometry and boundary conditions. For convection between parallel plates it
lies between 657 (free boundaries) and 1708 (rigid boundaries).
This example is for water convecting between two rigid boundaries for Ra~3000. The temperature field exhibits two narrow boundary layers,
with convective stirring between. This temperature distribution is reflected in the velocity field, with sluggish flow at the boundaries and in the
interiors of the two cells, strong downwelling at the side boundaries, and strong upwelling in the middle of the domain:
Figure 8-2. Temperature field for a system with Ra~3000.
Figure 8-3. Velocity field for a system with Ra~3000.
Nusselt Number
The Nusselt number is the ratio of heat transferred from a convecting system compared to that which would be transferred by
conduction alone. Thus, below the critical Rayleigh number, the Nusselt number is one. Once the critical Rayleigh number is
exceeded, it grows as (Ra/Ra,crit)^1/3. The Nusselt number is computed by integrating the heat flux through the upper
surface of the convecting system.
Figure 8-4. Nusselt number near the critical Rayleigh number.
Will fluids in the ocean crust convect?

Consider a open fracture penetrating to great depth, say a kilometer. The Rayleigh number is:
Eq 8-16:
(Here the viscosity and density term have been replaced with the kinematic viscosity.). For some representative values
(Table VII-1)
the Rayleigh number is about 10 --far in excess of the critical value.
Animation
Animation of Unsteady 2-D Convection
Animation of (Unsteady) 2-D Convection
This animation is of a numerical model of two dimensional convection at high Rayleigh number. The streamlines are shown in
the left hand panel; the temperature field in the right hand panel. The dominant mode involves downwelling in the center of
the domain and upwelling at both sides (i.e., two convection cells). Occasionally however an instability forms in the center of
the domain and is swept into one of the upflows (this short animation is looping and creates the illusion that the instability
always moves to the left).
Animation courtesy of A. Cherkaoui, 1998
Convective Flow in a Porous Medium; Model for Crustal Porous Flow
Recall (from the lecture on Rayleigh criteria for convection in a fluid) that the Rayleigh number is:
Eq 10-1:
Horton and Rogers and later Lapwood extended this analysis to a situation where a porous medium fills the space between two horizontal plates
separated by a distance h. (Details of the analysis can be found in (14), section 5.1, or (11), section 9-9.) Horton shows that the criterion for
establishing convection in the porous medium, i.e., the onset of Rayleigh-Darcy instabilities is:
Eq 10-3:
For values representive of newly formed oceanic crust (a 6 km crust with a temperature gradient of 1200) the critical Rayleigh number is
exceeded:
Eq 10-4:
The length scales, l, for the convection cells scale with the permeability:
Eq 10-5:
In isotropic media, the horizontal spacing of cells should be similar to the depth of penetration of the fluid; spacing of vent systems along the
ridge axis carries information on the depth to which fluid circulates.
Darcy flow can be used as a starting point to model the convection of fluid through the oceanic crust. In three dimensions:
Darcy's Law
Eq 10-6:
conservation of fluid
Eq 10-7:
conservation of heat
Eq 10-8:
In these equations q is the mass flux of fluid, k is the permeability, is the kinematic viscosity, P is pressure, is density, g is gravitational
acceleration, Cp is the heat capacity with the additional subscript m denoting the porous medium (solid and fluid) and the subscript f denoting the
fluid filling the pores, is the thermal conductivity (note non-standard use of kappa), and u is the spreading velocity. In the energy balance, the
three terms on the right hand side of the equation represent repectively the
convective transport of heat by the fluid through the medium
conductive transport of heat through the medium
advective transport by the medium, i.e., spreading plate
Equation 10-8 is identical in form to a refined conductive plate model, except for the first term on the right hand side which
expresses movement of fluid relative to the solid medium.
The system of equations is generally solved numerically. The methodology is to introduce stream functions, i.e., contours of constant q, and cast
the equations as finite differences in order to obtain a solution.
One of the first applications of this approach can be found in (15). The boundary conditions for flux of mass are a free upper boundary and
impermeable sides and bottom. The initial condition for temperature is taken as the conductive temperature distribution. The boundary conditions
for temperature are a horizontal flux condition at axis which matches the heat content of the injected molten material and the conductive heat flux
at bottom boundary.
The model is applied in the context of a heat flow survey on the flanks of Galapagos Rift conducted in 1976-77. For a mid-ocean ridge this is in
an unusual setting in that it is near to the equator where the sedimentation rate is high, about 50 m/My. Because of the high sedimentation rate,
conductive heat flow is measurable on crust as young as 100,000 y. There is virtually a total sediment blanket within the 1 My isochron. Data
from the heat flow survey are shown here:
Figure 10-1
The cases shown are for varying depths of fluid penetration. All have exponentially decreasing permeability with depth though the value at upper
surface is varied:
Figure 10-2.
The modeling approach was to find the combination of parameters that best matched the conductive heat flow distribution in terms of the
measured conductive heat flow and the geometry of the convection cells. The data:
Figure 10-3
The models:
Figure 10-4
Data averaging prevents finding an exact match.
For models with a reasonable match to the observations, the fluid output is about 5 x 10 g of water per year per cm of ridge which translates to 2
x 10 cal/km of ridge/y, much smaller than reference value calculated earlier when discussing local scales, 5 x 10 . There are two possible
reasons for this discrepancy: the slower spreading (factor of 3.5/5) and the possibility that the insulating effect of having a sediment cover
impedes convection. Cooling of the upper crust increases shallow temperature gradients and thus the conducted heat flux from below:
Figure 10-5
The cell geometry appears to be stable despite advection of the plate away from the axis, however this may be an artifact of the impermeable offaxis boundary condition:
Figure 10-6
A final observation is the correspondence of the depth of penetration of the most vigorous streamlines and the 300-400C isotherms. The origin
of this behavior is in the dependence of the Rayleigh number on and . As a function of temperature, the expansion coefficient shows a strong
maximum in this temperature range (near to the critical point) and the viscosity a minimum thereby promoting convection:
Figure 10-7
Time Series Analysis
Time series of oceanic phenomena often contain periodic components related to forcing at a wide range of time scales:
waves, tides and tidal currents, diurnal and annual cycles, ENSO, Pacific Decadal Oscillation, and orbital geometry, its
influence on incoming solar radiation and Pleistocene climate. Time series analysis involves the extraction of information
about these periodic components from time series data.
A periodic signal can be represented by its intensity or amplitude as a function of time:
Figure 15-1
Using this figure we can introduce some definitions:
The peak intensity of the signal (relative to the mean) is the amplitude, A.
The waveform repeats with period T.
Time can be expressed in either time units, t, or as an angle (expressed in radians).
The angular scale is related to the time scale via the relationship:
where is the angular frequency and is the phase angle.
The period is related to the angular frequency by the relationship:
The frequency f is the inverse of the period f=1/T.
Thus =2 f.
With these definitions, the signal can be represented as:
A physical process can generally be described by the values of some parameter h that varies with time, h=h(t). We refer to
this representation as being in the time domain. An alternative and equivalent representation of the same signal would be to
express it as its amplitude, H, as a function of frequency, f: H=H(f) . We refer to this representation as being in the frequency
domain.
Time Domain
Frequency Domain
Figure 15-2. The period of the waveform is 36 time units, so that the frequency is 1/36=.0278. (The vertical axis on the right
hand side is power, not amplitude, see below).
These two representations of the same signal can be related by the Fourier transform:
Eq 15-1:
(recall that
so the Fourier transform has real and imaginary components).
The original time series can be obtained from the inverse transform:
Eq 15-2:
The transforms can be thought of as weighted averages of waveforms of a given frequency; the forward transform to the frequency domain
weighted by the signal amplitude h(t) and the inverse transform to the time domain weighted by the frequency amplitude H(f). In other words,
when the signal h has the same frequency as the waveform, the the integral H is large.
If h(t) is real (we are generally considering real signals) then H(f) is generally a complex number. However under this condition H(-f)=H(f)*. (the
notation * denotes the complex conjugate). This symmetry means that for real signals there is no unique information in the negative frequencies.
The power of the signal, P, is a measure of its squared amplitude. The power at a frequency f is defined as:
Eq 15-3:
and so the total power in a signal is (fix equation, no (f) on lhs):
Eq 15-4:
Equations 16-1 through 16-4 apply to a continuous function spanning all times. In general though real signals will:
be represented by discrete measurements
these measurements will span a finite interval
For such discrete time series, we consider measurements of h taken at discrete intervals,
Eq 15-5:
The quantity 1/ is the sampling rate. For any particular sampling interval there is an frequency f called the Nyquist critical frequency:
Eq 15-6:
What is the significance of the critical frequency? There are two related issues.
The first issue involves a concept called the sampling theorem which says:
If the spectrum of h(t) is zero outside of the frequency interval -fc<f<fc, then all spectral information is obtained if the signal is sampled at
times separated by no more than 1/(2 ). This means that we gain no spectral information by sampling at a rate greater than the Nyquist
frequency. We can see why by examining the periodic signal in this figure:
Figure 15-3
The period of the signal is 36 time units and so the Nyquist frequency is 1/18 time units. If we sample time unit by time unit, we get a
smooth curve. Even if we sample by 18 time units, the Nyquist period, we get (provided we are not exactly in phase with the zero
crossings) the highs and the lows so that we still capture the period of the signal. If we sample by 36 time units though, we only get a
single value--the signal is constant through time.
Insuring that H(f)=0 beyond the Nyquist frequency is achieved in practice by applying a low pass filter (a filter that lets low frequencies
pass the filter, but not high frequency components) to the signal before sampling (as we will see, nature applies a low pass filter through
bioturbation in the surface-most sediments).
The second issue is a phenomenon called aliasing. If there is power outside the frequency band -f
sampling folds that power into this frequency band. Schematically:
f f then discrete
Figure 15-4. The width of the true transform is 2fc; the band from fc to 3fc maps to fc to -fc, the band from 3 fc to 5fc maps to -fc to fc, and
so on.)
To summarize:
1. Unless H(f) is zero once f exceeds 1/(2 ), there will be aliasing by higher frequency components.
2. To avoid aliasing, either the sampling interval must be chosen to insure that this condition is met or the original signal
must be filtered to remove high frequency components before sampling.
To understand the nature of the filtering effect of bioturbation consider as an example a core taken in sediments accumulating at 3 cm/ky and
sampled at 10 cm intervals. Thus, the sampling interval is 3000 y and so the Nyquist frequency is 1/6000 y.
To model bioturbation, consider a well mixed box of height h at the sediment surface with material delivered at rate F and removed as the
product of sedimentation rate times concentration uc . A mass balance can be written:
Eq 15-7:
or rewriting in terms of a periodic variation in the delivery of material:
Eq 15-8:
This differential equation can be solved for some arbitrary initial condition to find that at large times, when initial conditions have decayed, there
is a stationary solution:
Eq 15-9:
This equation can be simplified at the limits of high and low frequency. When f is large (the supply of material to the sediment varies at high
frequency) it reduces to:
Eq 15-10:
In this case mixing averages the fluctations in the supply of rain to the sediment. When f is small (the supply to the sediment varies slowly) it
reduces to:
Eq 15-11:
In this case the periodic signal is preserved. In other words, low frequency signals are preserved while high frequency ones are not; the system
acts as a low pass filter. The boundary between "low" and "high" frequency is effectively f~u/h. For a depth of bioturbation of 2-10 cm in our
sample core, this frequency is somewhere between 1/600 y and 1/3000 y. By carefully choosing the sampling interval, little aliasing is possible.
(And of course we can also apply a low pass filter to be sure.) Conversely, a limit exists for the highest frequency signal that can be resolved in a
core.
Mathematica is a useful tool for solving Eq 15-8 | Download Notebook
The time series analysis of a record consists of a series of computational steps. Depending on the application these might include:
1. detrending and whitening
2. autocorrelation
3. windowing
4. spectral estimation
We will discuss these in the reverse order from which they are performed.
Spectral estimation
If we have N measurements of a parameter h sampled at an interval then these occur at times t =k with k=0,...,N-1. We introduce the notation
for the values at these times: h h(t ) From these N measurements we can derive independent information at only N discrete frequencies--these
are normally spaced evenly within the Nyquist limits:
Eq 15-12:
With these conventions the discrete Fourier transform is written:
Eq 15-13:
The power at each of these frequencies is:
Eq 15-14:
This information can be plotted in the form of a periodiogram:
Figure 15-5
Windowing
A common problem is that these estimates are not within the bin implied by the geometry; they actually have a spread several bins wide. A
process termed data windowing minimizes this leakage between bins. It involves applying a windowing function to the data before forming the
transform:
Eq 15-15:
Then the power density is given by:
Eq 15-16:
The origin of the leakage between bins arises in the process of taking a portion out of a long time series. Consider this time series h and window
function w:
Figure 15-6
Extracting a discrete portion out of the time series h is equivalent to multiplying h by a window function w where w=1 in the region of interest
and 0 everywhere else. The convolution theorem say that for two functions h(t) and w(t), if p=wh then P(f)=W(f)H(f). When the discrete
transform is taken, the high frequency components associated with the steps in the function w are aliased into the Nyquist interval. The
characteristic of a good window is to smoothly turn "on and off", thereby minimizing the introduction of high frequency components into the
transform.
Autocorrelation
In performing time series analysis, one can work with autocorrelated records to produce a correlogram. If the Fourier transform of h(t) is H(f),
and the autocorrelation function of h(t) is r, then the Fourier transform R(f)=|G(f)|2. For N data points taken from two times series g and h, the
correlation as a function of the lag j is:
Eq 15-17:
Thus the autocorrelation of a time series onto itself is given by:
Eq 15-18:
Imagine a periodic function that is shifted progressively to the right:
Figure 15-7
As it is shifted the correlation will drop, then rise to the same value when the function has been shifted exactly one cycle:
Figure 15-8
Thus the autocorrelation produces a different time series, but with the same spectral components.
Detrending, whitening
There are generally long period and d.c. components that give rise to power in a spectra being concentrated at low frequencies. High pass filters
can be applied to remove the d.c. baseline (detrending) and long period, low frequency components (whitening). The effect is to emphasize peaks
at intermediate frequency, which no longer sit on the shoulder of a larger low frequency peak.
Thus far we have emphasized frequency characteristics of a time series. Time series analysis also includes consideration of the frequency
components within the time domain. These are separated by:
1. transforming to the frequency domain
2. multiplying the transform by a bandpass filter. This product contains only the information contained within this
frequency band
3. applying the inverse transform back to the time domain
Thus far we have been considering time series that are stationary, i.e., whose frequency content is not changing with time.
We may also be interested in phenomena that do not satisfy this constraint. A traditional approach is to apply Fourier analysis
to overlapping subsets of the time series in order to examine the evolution of the frequency content. A more recently
developed technique is wavelet analysis (for a good review see (42)). Figure 15-9 is adapted from that reference and
compares Fourier and wavelet analysis of a stationary time series with two superimposed sine waves and a time series
containing signals of the same frequency content but in diferent segments of the observation period.
Figure 15-9
We can review these ideas with this Matlab demonstration.
Some resources for more detail on these topics:
W.H. Press, et al., Numerical Recipes in <C | Fortran | Pascal>

chapter on Fourier and Spectral Applications develops background for some basic algorithms
J.S. Bendat and A.G. Piersol, Random Data: Analysis and Measurement Procedures
survey of data analysis, including spectral methods
L.H. Koopmans, The Spectral Analysis of Time Series

a comprehensive treatment, used for Charlie Eriksen's Data Analysis course
http://www2.ocean.washington.edu/oc540/lec01-12/matlab.html
MATLAB Time Series Example
The MATLAB script timeseries.m can be used to repeat this example or used as a starting point for further study.
In this example we will use a synthetic data set. To create our synthetic data we will make our unit of time 1000 years = 1 ky and sample a
500,000 year record in 2 ky increments. We first create a vector containing the times:
t=0:2:500;
We want this record to have multiple periodic components. To mimic the orbital forcing of the Earth's climate, we create a vector d, of the same
length as vector t and with components that have periods of 100 ky, 41 ky and 21 ky, weighted 2.5:1.5:1, then plot it
echo off
axis ([0 500 -5 +8])
echo on
d=2.5*sin(2*pi*t/100)+1.5*sin(2*pi*t/41)+1*sin(2*pi*t/21);
plot(t,d)
()
We add some normally distributed random noise to the record to create a new record dn, and plot it on the same graph:
dn=d+0.5*randn(size(d));
hold
plot(t,dn,'g')
()
We also add a long term trend to create a third record dnt, and again plot it on the same graph
dnt=dn+3*t/500;
plot(t,dnt,'r')
()
Our time series has 251 elements. To improve computation efficiency we will pad it with zeros and extended it to 256 elements (a power of 2).
This is done implicitly with the fft command. Then taking the Fourier transform and plotting its real part:
DNT=fft(dnt,256);
hold off
plot(real(DNT))
()
By convention, the transform has elements equally spaced in frequency, with the first element corresponding to f=0, then up to the Nyquist
frequency at the middle of the record, then the negative frequencies from large to small in the remaining elements. The negative frequencies
contain redundant information so we want to look only at the positive frequencies. These frequencies are evenly spaced within the Nyquist
interval (for our 2 ky sampling interval the Nyquist frequency is 1/4 ky):
f=0.25*(0:128)/128;
The power spectrum of the record is estimated as:

PNT=DNT .* conj(DNT)/(256*256);
(The norm of a complex number is the number multiplied by its complex conjugate.) We are interested in the parts of this
spectrum corresponding to positive frequencies
hold off
plot (f,PNT(1:129))
()
We easily see the peaks at frequencies of about .01, .025 and .05 per ky with the power in approximate proportion to the amplitude squared
(25:9:4 or approximately 6:2:1). Notice the power present as the frequency goes to 0 corresponding to the long term (d.c.) trend. This can be
removed by detrending the data which for us is equivalent to working with the dn record. Applying the same steps:
DN=fft(dn,256);
PN=DN .* conj(DN)/(256*256);
hold
plot (f,PN(1:129),'g')
()
We see that the low frequency contamination is removed. We will work with detrended dn record from now on. Next we will apply a window-- a
Hamming window in particular
z=hamming(251);
plot(z)
()
Multiplying the dn record by this window:
dnz=dn .* z';
plot(dnz)
()
We now take the power spectra of this windowed time series (PNZ) and compare to the situation with no window (PN)
DNZ=fft(dnz,256);
PNZ=DNZ .* conj(DNZ)/(256*sum(z));
plot (f,PNZ(1:129))
hold
plot (f,PN(1:129),'g')
()
We are also interested in details of the time series itself. To extract parts of record, say the 21 ky band, we need bandpass filter to let through
frequencies between say .04 to .06/ky. Here we use a Butterworth filter, one of the many possibilities:
[b,a]=butter(5,[.04/.25 .06/.25]);
plot (filter(b,a,dn))
()
We find ~500/21=24 cycles of approximately the right amplitude (1) once we are 5 cycles into series. (There are ways of minimizing these end
effects).
We also wanted to look at spectral characteristics of autocorrelated data:
dna=xcorr(dn);
plot (dna)
()
Zero lag is plotted in the middle of diagram, with negative lag to the left and positive to the right. The plot is symmetric because the record is
compared to itself. We find the power spectra of the autocorrelated data:
DNA=fft(dna,512);
PNA=DNA .* conj(DNA)/(512*512);
f2=.25*(0:256)/256;
plot(f2,PNA(1:257,:))
()
We find the same spectral character working with the autocorrelation, the difference being that the power is a measure of the variance due to that
periodic component, i.e., most of the variance is explained by the 100 ky period.
Periodic Phenomena in Vent Environments
One of the more intriguing observations made in vent environments is that temperatures and flow show periodic fluctuations
inferred to be tidally forced. At present there is no accepted explanation of this relationship. Indeed, there are question
whether the observations reprsent basic characteristics of vent processes or simply reflect the influence of tidal currents on
the measurement apparatus.
Tidal Forcing
For a comprehensive discussion of tidal generation refer to a standard physical oceanography text, e.g. (46). The periodic
characteristics of the tide are related to the interaction of astronomical periodic behavior controlling the spacing and
orientation of the Earth, Moon, and Sun:
process
period
Earth spin
24 h
Moon's orbit
27.32158 d
Earth's orbit
365.2422 d
Moon's nodal precesion
18.613 y
Variation of Moon's perigee
8.861 y
Variation of Earth's perihelion
20940 y
The forcing can be related to a set of additive tidal constituents (366 of them), the most important of these being:
period
T
amplitude
K (cm)
M2
main lunar semidiurnal
12.421 h
24.233
K1
soli-lunar diurnal
23.93 h
14.156
S2
main solar semidurnal
12 h
11.284
O1
main lunar diurnal
25.82 h
10.051
P1
main solar diurnal
24.07 h
4.684
Q1
elliptical lunar diurnal
26.87 h
4.684
N2
elliptical lunar semidiurnal
12.66 h
4.640
Mf
lunar fortnightly
13.66 d
4.174
Tidal analysis involves interpreting long duration records to establish the phase and amplitude of these constituents at a
particular location. Harmonic analysis is used, i.e., constructing a least squares fit to waveforms of known frequency. For
details see (47) pg 397-402.
Figure 1 shows the tidal forcing due to the first four constitutents all with phase of zero. The matlab script, tidebuild.m illustrates the
progressive contributions of the components starting with the dominant M2 tide.
Figure 1. Tidal forcing from M2, K1, S2 and O1 tidal constituents.
Tidal Currents
In order for sea lvel to change fluid must move from place to place giving rise to tidal currents. In a general sense, tidal
currents are related to the first derivative of the tidal amplitude. In the open ocean these currents are rotary. To construct
models for tidal currents, records of current direction and speed are interpretted as the additive efects of the underlying tidal
constituents.
In the vent environment there is a strong influence of topographic forcing and of frictional boundary effects, especially when there is a well
defined rift valley. As an examle consider the character of current in the axial valley near Main Endeavour Field, Juan de Fuca Ridge:
Earth Tides
In response to the gravitation fields of the Moon and Sun, the Earth deforms because it has a certain degree of elasticity. The
amplitudes of motion are much smaller than for the water tide, of order 30 cm and in practice only the first four tidal
constituents are considered when analyzing the earth tide.
Mechanisms
Several mechanism have been suggested that would connect tidal forcing to vent behavior:
deformation of the porous netork due to the Earth tide, cuasing a pumping action.
efects of pressure on the properties of the fluid, especially for fluids near to the critical curve for water-sea salt.
mechanical interactions of vent structures with tidal currents.
Some Representative Observations

As an example of periodic behavior consider these observations from Schultz et al. (47), taken with an electromagnetic flow
appartus deployed at the "Peanut" structure of Main Endeavour Field. The appartus consists of a vertical tube sealed by a
gasket to the surface of a vent structure, with measurements made of the flow through and temperature in this tube.
Figure 3. From (47)
Figure 4. From (47)
The power spectra clearly show an influence of the M2 tide in all of the records. However an analyis of the coherence between
the records demonstrates that most of the coherence between these records is at periods >12 h arguing that while tidal
signal are evident, most of the variability is related to long period, low frequency phenomena.
This time series is filtered to pass low frequencies:
Figure 5. From (47)
There appears to be an inverse relatinship between temperature and flow, perhaps with a small lag. There are some plausible
physical explanations for this inverse relationship but such explanation are at this point ad hoc in nature.
Tidally related variability is also observed in high temperature, smoker style vents. These data are from the "Cannaport" vent at MEF:
This example is representative of many records where a 12 hr periodicity is present, though not easily interpretable in terms
of making a connection between forcing and response.
A more intriguing (and perhpas unique) record comes from observations made in 1995 at MEF at the vent structure named "Puffer", a vent poised
on the critical curve (379C at 220 bars):
The strong excursions in temperature come with a period ~12.4 h, the M2 period, and the strongest excursion is in phase with the full Moon.
However in a long time series taken in the following summer only two excursions were recorded. In a dive program conducted the following
summer, the pattern was again detected and an ALVIN dive conducted to measure temperature and collect fluids before, during, and after an
event. A ~20C excursion was observed. The entrained fluid, surprisingly, was not seawater, but is consistent with being the conjugate brine of
the relatively fresh venting fluids, conductively cooled. The mechanism by which these fluids are entrained is likely due to cracking events of
some kind, but it is unclear whether they are induced by the Earth tide or the pressure flucations due to the water tide.
Marine Sedimentation
Marine sediments are the products of a limited number of physical, biological, and chemical processes. The nature of the resultant sediments is
determined by the relative rates of input of material supplied by these processes. In this class we focus on deep sea sediments, but the processes
apply equally to shallow water deposits.
Physical processes dominate at ocean margins, where they transfer particles eroded from the land to the sea floor.
Active sedimentation processes (where the sediment modifies the properties and behavior of the suspension) include mass wasting and
density currents. Such deposition tends to mask other sedimentary processes. The giant landslides surrounding the Hawaiian islands and the
abyssal plains of the Atlantic are striking deep-sea examples of active sedimentation. Because active sedimentation depends on gravitational
energy, it does not extend seaward of the trenches along convergent plate boundaries (but can affect the entire ocean basin off passive margins).
Passive sedimentationprocesses are those in which the sediment is carried by but does not modify the normal thermohaline circulation.
Examples:
Hemipelagic sedimentation, where it appears that fine sediment moves along isopycnal surfaces high in the water
column. This process is poorly understood (for example, the role of internal waves in resuspending sediment, the role of
squirts and jets in transporting sediment ofshore and the role of biota in sediment deposition have yet to be
quantified). Hemipelagic sedimentation produces a fringe of terrigenous (land-derived) deposits up to a few hundred
kilometers wide around land masses. These deposits are draped uniformly over the sea floor topography
"Drift" deposition along the path of bottom currents. Prominent in the high latitude North Atlantic along the flanks or
ridges. Sediment in transit to drift deposits creates a near-bottom "nepheloid layer." Paleoceanography, 9 (6),
(December 1994) has a dozen papers on drift deposits.
Eolian (wind transported) sedimentation. Prominent where major wind systems cross semi-arid source areas
(ephemeral lakes) or active ash-generating volcanoes. Prominent in the North Pacific (Figure MS-1), North Atlantic and
Arabian Sea
Figure MS-1. Quartz concentration in North Pacific sediments (carbonate-free basis) showing efects of eolian transport
in westerlies and northeast trades.
Ice rafting. Restricted to high latitudes, but can carry coarse particles into the subarctic gyres, far from land.
Sediments formed by physical processes have distinctive acoustic signatures of military interest. Hence they have been much studied during the
past 50 years.
"Passive" sediments record the history of deep currents, volcanism, aridity, wind trajectories, and iceberg abundances and trajectories. Drift
deposits, which can accumulate at hundreds of meters per million years, yield some of the highest resolution paleoceanographic records.
Biological processes dominate sediment formation in areas of high productivity that receive little terrigenous material. The equatorial
Pacific and Southern Ocean are examples.
Several taxa of phytoplankton and zooplankton (the latter include benthic and upper water column representatives) secrete either CaCO3
(coccolithophores and foraminiferans) or opal (hydrated SiO2 - diatoms and radiolarians). If not masked by terrigenous material, the tests (shells)
of these organisms can form carbonate or siliceous oozes.
Because the distribution and abundance of the various species are determined by the temperature, salinity, thermocline depth, carbonate
chemistry, and productivity of the waters in which they live, most of what we know about paleoceanography is derived from fossil assemblages
of these organisms.
In addition, the composition of the tests records the oxygen and carbon isotopic compositions and trace element contents of the waters in which
they were secreted. These parameters in fossil tests provide insights to past ice volumes, temperature, productivity, and changes in
biogeochemical cycles.
Benthic organisms also modify the historical record by actively mixing ("bioturbating") the most recently deposited sediments. Bioturbation,
which is addressed later in the course, is effectively a low-pass filter that supresses or eliminates records of events that create layers of sediment
thinner than the depth of mixing. Rapidly deposited or anoxic sediments provide the only deep-sea records capable of resolving events shorter
than about a millennium.
Chemical processes dominate sedimentation only in deep, low productivity areas shielded from terrigenous material. Precipitates from
hydrothermal solutions emanating from mid-ocean ridges are prominent along the flanks of the East Pacific Rise in the South Pacific. Authigenic
deposits (formed by the very slow precipitation of oxyhydroxides and silicates? from normal seawater) form distinctive sediments in the central
South Pacific as well as ferromanganese nodules (Figure MS-2) and crusts on any surfaces where other sediments are absent or accumulating
extremely slowly.
Figure MS-2. Deep sea ferromanganese nodules on the floor of the South Pacific Ocean (individual nodules are 5-10 cm diameter).
Chemical processes also modify biological (biogenic) sediments through the dissolution of CaCO3 and (to a lesser extent) opal in deep water. T
and P both play a role in increasing the corrosiveness of deep waters, but the major effect on carbonate tests is due to the creation of CO2 by the
biologically mediated oxidation of organic matter, which creates biocarbonate ions at the expense of carbonate ions (Equn. MS-1), thereby
driving the dissolution of carbonate tests.
Equn. MS-1. CO2 + CO3= + H2O = 2HCO3In general, the carbonate content of deep sea sediments decreases with increasing water depth. Two important horizons are the lysocline, where
the proportion of solution-resistant tests increases abruptly, and the calcite compensation depth (CCD) which is the boundary between carbonatebearing and carbonate-free sediments (Figure MS-3)
Figure MS-3. Carbonate concentration versus depth showing lysocline and CCD (right) created by simple linear dissolution with depth of
carbonate in a carbonate-opal mixture (left).
The sea floor distributions of both carbonate and opal tests reflect the evolution of bottom water chemistry as it makes the "grand tour" from the
North Atlantic via the Southern Ocean to the northern Pacific and Indian Oceans. As the bottom water "ages" it accumulates carbon dioxide,
which drives equation MS-1 to the right, thereby dissolving CaCO3.
Chemical reactions within the sea floor (diagenesis) can further modify deep-sea sediments, both through isochemical changes (which can
obliterate particle source and paleomagnetic information) and open-system changes (which can modify the isotopic and trace element
composition of sedimentary particles).
Rates of deposition. Sediments can be dated by a number of well established techniques. the most common are:
First and last appearance of fossils
Paleomagnetic correlations
Decay of radioactive isotopes
Oxygen isotope correlations
From dated layers in sediment cores, the rates of sedimentation of the various types of deposits can be calculated. Typical ranges for the products
of the various depositional processes are (note that m/my is equivalent to mm/ky):
Nearshore sediments, turbidites
Up to km/my (kilometers/million years)
Hemipelagic deposits
Tens to hundreds of m/my
Drift deposits
40-400 m/my
Mid-latitude eolian deposits
3 to 10 m/my
Ice rafted material
10+ m/my
Carbonate oozes
Up to 50 m/my
Siliceous oozes
Up to 10 m/my
Hydrothermal deposits (off ridge axes)
About 0.5 m/my
Hydrogenous sediments
Rarely exceed 0.2 m/my
Ferromanganese nodules
0.0002 to 0.005 m/my (0.2 to 5 mm/my)
Sediment Accumulation
Deposition and Accumulation of Marine Sediments

The sediment record that accumulates below the sea floor is rarely identical with the sediment that is initially deposited by oceanographic
processes (Figure SA-1). Only where bottom currents are weak and the oxygen value at the sea floor is virtually zero (thereby inhibiting
bioturbation) are relatively pristine records preserved. These conditions are found at such locations as the Santa Barbara Basin off southern
California, Saanich Inlet, north of Victoria, B.C., Guymas Basin in the Gulf of California, Orca Basin in the Gulf of Mexico, the Black Sea, and
the Cariaco Trench off Venezuela. These sites have been studied intensely to unravel the highest frequency changes in oceanographic and
depositional conditions.
Figure SA-1. Processes that intervene to change the characteristics of freshly deposited sediment prior to its incorporation into the geologic
record.
Bioturbation mixes the new sediment with older material, thereby acting as a low pass filter of the record of depositional change. In addition, if
the incoming sediment is poorly sorted, bioturbation can separate the coarse particles from the fine ones, which can be concentrated at the sea
floor (armoring). Erosion, whether physical, physical plus biological (which can result in the removal of material at very low bottom current
velocities), or chemical, all create gaps or distortions of the preserved record. The change in elevation of the sea floor over time can be described
in terms of the bulk density of the surface sediment, the flux of incoming material, and lateral transport.
Equn. SA-1:
Where the del term is positive (lateral convergence), sedimentation is enhanced by lateral transport (as is the case in the drift deposits of the
North Atlantic, for example). In contrast, divergence (negative del term) results in erosion.
The rate of deposition and the processes that modify freshly accumulated sediment can be unraveled by the use of radioactive isotopes. The use
of 14C to interpret the Holocene dissolution of equatorial Pacific carbonate deposits will be discussed in a future lecture. Other useful radiotracers
are members of the U and Th decay series. In unweathered minerals, the daughter isotopes of the decay of the primordial radioactive isotopes
238
U and 232Th are in equilibrium with the parent (i.e. the absolute decay rate of all isotopes in a series is the same). During the surface
geochemical cycle, however, the daughters are separated from the parents and begin to decay as independent species, each with its characteristic
half life. In the case of 238U (Figure SA-2), daughters such as 234Th (t1/2 = 24 days), 230Th (75,000 years), 226Ra (1,622 years) and 210Pb (21.4
years) have been used to interpret sediment deposition.
Figure SA-2. Chain of isotopes produced by the radioactive decay of 238U. Downward shifts in the table result from alpha decays (loss of a He
nucleus), shift up and to the right result from beta decays (loss of a nuclear electron). Half lives of radioactive isotopes are shown in years (Y),
days, (D), minutes (M) and seconds (S).
For 232Th (Figure SA-3), the daughters 228Th (1.9 years), 224Ra (3.6 days) and 212Pb (10.6 hours) have been used in the same way.
Figure SA-3. Chain of isotopes produced by the radioactive decay of 232Th. Symbols as Figure SA-2.
In radioactive decay, the number of atoms decaying depends only on the number present,
i.e. - dN/dt = lambda * N, where lambda is the decay constant. If the number of atoms of a radionuclide at time zero is No, then the number Nt at
time t is:
Equn. SA-2:
Where C is any measure of concentration of the nuclide. Since the accumulation rate (A) is dz/dt, A is given by:
Equn. SA-3:
Where, in this case, the depth interval
z is measured from the sea floor. This relation ignores the effect of bioturbation, which mixes fresh
and decayed sediment. For a steady state,
Equn. SA-4:
Where the first term describes bioturbation, the second burial, and the third, radioactive decay. This yields:
Equn. SA-5:
Which, by comparison with Equn. SA-3, shows that the accumulation rate will be overestimated if bioturbation is not taken into account. The
two unknowns, A and D can be determined by measuring two isotopes with different half lives. Ideally, one is chosen with a half life close to the
characteristic time for bioturbation and the other with a half life equivalent to deposition over several bioturbated layer depths. The limited
selection of available radionuclides does not always make this possible, however!
Identifying Sediment Sources
Most sediments are derived from two or more independent sources. One of the challenges facing sedimentologists and geochemists is to unravel
these sources, both because of the paleoenvironmental information they carry, and because mobile elements will interact differently with the
different sediment components. Individual minerals differ in their crystal form, lattice spacing, and location of elements within the crystal lattice.
As a result, each mineral lattice diffracts X-rays slightly differently. Such differences are used to identify the minerals present in a sample, and to
estimate the abundance of each. Unfortunately, this technique rarely fully characterizes a sample because of the presence of amorphous or poorly
diffracting phases, and because appropriate standards for many naturally occurring minerals (particularly those which are very fine grained) are
not available. As a result, much effort has been devoted to developing strategies to partition elemental composition data into source components.
This is a classic inverse problem, where the compositions of the sources are rarely known with certainty.
Mixing models. For the case of two end members, A and B, containing a% and b% of some element, the proportion of A in an unknown sample
containing x% of the element is (b-x)/(b-a) (assuming that x lies between a and b). In general, however, more than two end members, a,b,c, ...,
are involved, their exact compositions are uncertain or variable, and there are analytical uncertainties associates with a,b,c, ...,x.
A number of different techniques have been used to address this problem. They include:
Factor analysis. This technique, which is a form of objective analysis, will be discussed in much more detail in the lecture on proxies, where it
will be used to reduce abundance data for a large number of microfossils in a sample to a more manageable number of statistically independent
groupings or factors. Elemental abundances can be substituted for species abundances to yield statistically independent geochemical factors.
The selection of the appropriate number of factors is subjective and is usually guided by geographic or temporal coherence in the distribution of
each factor. A serious limitation of the method is that unless the sample set includes members the compositions of which are close to those of the
end members, the compositions of the factors will bear little relation to real source materials. Thus, the proportions of each factor making up each
sample will be difficult to interpret in physical terms. Various mathematical strategies have been applied to solve this problem, but none are
satisfactory in all cases.
Component stripping. If the concentration of an element is much higher in one end member than in all the others, it can be used to estimate the
proportion of that end member in all the samples. That proportion, together with the known concentration of other elements in that end member
can be used to remove the contribution of that end member to each element in the sample, i.e.
For end member A, consisting of A(a)%. A(b)%, A(c)%, ... of elements a, b, c, ... respectively, where A(a) >> B(a), C(a), ... (the concentration of
a in all other end members), the proportion of A in sample X is X(a) / A(a). A's contribution to elements b, c, d ... in X is (X(a)/A(a)) *A(b), ...
and the unexplained part of b, c, ... is :
Equn. SA-6:
The process can be repeated if the concentration of a second element is much greater in B (or A and B) than in any of the remaining end
members, and so on.
This approach works quite well in an area such as the East Pacific Rise west of South America. There, the sediment consists largely of detritus
from South America that is rich in Al relative to other components, hydrothermal plume precipitates from the Rise crest that are rich in Fe
relative to non-detritus components, biogenic material from the overlying water column that is rich in Zn relative to the remaining component,
and authigenic material precipitated slowly from seawater. The stripping technique yields proportions that are comparable to those from more
sophisticated interpretive techniques.
The method's weaknesses are its requirement for strong elemental differentiation between end members (which is unlikely in the case of two
detrital end members, for example), its inability to convey any sense of the uncertainty in the calculated contributions of the end members, and
the propagation of errors resulting from several successive subtractions like that of Equn. SA-6. In addition, it requires that all the end members
and their compositions be known.
Linear programming. The requirement that elements be strongly partitioned between end members makes the analysis easier, but is not
necessary if the number of elements exceeds the number of end members. In this case, for four end members (as above) and n elements, there are
n equations of the form:
Equn. SA-7:
where i is the ith element, D,H,B, and A are subscripts for pure detrital, hydrothermal, biogenic, and authigenic ratios, respectively, and DAl, HFe,
BSi, and ANi are the (unknown) amounts of detrital Al, hydrothermal Fe, biogenic Si, and authigenic Ni in the sample, respectively, and Ri is the
residual or "error". To solve, we rearrange to:
Equn. SA-8:
For n elements, this yields n+4 equations. A conventional least squares analysis usually results in negative contributions for one or more of the
end members. Linear programming allows for restrictions on the solution. For sediment partitioning, end member contributions are restricted to
positive values or zero, and the sum of Ri over all elements is minimized. DAl can be converted to total detritus if the Al content of detritus is
known. The same for the other end-member elements, allowing the weight fraction of each end member to be determined.
The weakness of this method is its requirement that end member compositions be known, and that Ri be minimized (if one or more end members
are overlooked, for example, the true R may be large). In addition, trace element concentrations have to be scaled so that the major elements do
not dominate the solution to the problem.
Total inversion. The total inversion method, developed for solving geophysical problems (A.Tarantola and B.Valette, 1982, Rev. Geophysica and
Space Physics, v. 20, pp. 219-232) does not distinguish strongly between knowns (the elemental analyses) and unknowns (end member
compositions and abundances). All are considered "data" and are treated as input variables with specified uncertainties. This allows variable
analytical uncertainties as well as uncertainties about end member compositions to be factored into the solution. During the inversion, the
magnitude of the uncertainties control the adjustment allowed for each variable. The method differs most significantly from the others in
allowing the end member compositions to vary from sample to sample, which is probably more geologically realistic than the assumption of
fixed end member compositions. Figure SA-4 shows the end member compositional data and their uncertainties used as input to a total inversion
model of sediment composition of a North Pacific red clay core spanning the last 70 million years.
Figure SA-4. Concentrations and uncertainties in concentrations of elements in eight end members used to partition North Pacific red clay
sediments in core GPC-3 (Figure SA-5).
Figure SA-5. Location of Core GPC-3 showing its displacement relative to the hot spot reference frame during the past 65 Ma (due to absolute
drift of the Pacific Plate) and to present-day quartz abundances in surface sediments.
The inverse model was able to simplify concentration data for 39 elements in 324 samples (Figure SA-6,7) into eight end members that varied
coherently through time in a geologically and oceanographically reasonable way (Figure SA-8). Such modelling allows sediment budgets to be
generated in a reproducible and transparent way and to be modified easily in the light of new information.
Figure SA-6. Profiles of elemental abundances versus depth in core GPC-3. These elements are predominantly of terrigenous origin. The
variations result from movement of the core site and flux variations from multiple sources.
Figure SA-7. Profiles of elemental abundances versus depth in core GPC-3. These elements include substantial hydrothermal (left) and biogenic
(right) contributions.. The variations result from movement of the core site and flux variations from multiple sources.
Figure SA-8. Variations in the fluxes of seven end members to core GPC-3 during the past 70 My, estimated by total inversion of the core sample
and end member compositional data. The green dots indicate samples from the Cretaceous-Tertiary boundary layer, where the model fit to the
data is less satisfactory due to the presence of extraterrestrial material (characterized by high Ir concentrations) associated with the K-T meteorite
impact.
Oxygen and Carbon Isotope Systematics

Oxygen has three different stable isotopes:
O (~99.76%)
O (0.04%)
O (0.20%)
Molecules containing atoms of differing mass behave differently from one another:
1. in chemical reactions: the molecules have different chemical potentials and so have different equilibrium constants
2. during mass transport and in rate-limited chemical reactions: the molecules have different masses--the light molecule is transported or
reacts more quickly
For example, at equilibrium:
Eq 24-1:
and
Eq 24-2:
If a fractionation factor, , is defined as the ratio of the isotopic ratios in two phases, it can be shown that this fractionation factor is equal to K:
Eq 24-3:
This equilibrium fractionation is usually expressed in a different way, as the "del value" for the fractionation (upper case ):
Eq 24-4:
which is usually expressed per mil,
Similarly measurements of the isotopic ratio are generally expressed in a parallel del notation (lower case delta):
Eq 24-5:
also expressed per mil, .
For oxygen isotopic measurements of water the standard is standard mean ocean water, SMOW, while for carbonates either SMOW or fossil
belemnite from the Pee Dee formation, PDB, is used as the standard.
A consequence of the equilibrium relationships is that if two materials equilibrate then
Eq 24-6:
To apply this relationship in interpreting geological record, we are concerned with four possible effects on the temporal variability:
1. the isotopic composition of seawater
2. the temperature of the water in which the calcium carbonate was deposited
3. whether equilibrium is established
4. whether calcium carbonate acts as a closed system once deposited
The isotopic composition of liquid water responds to evaporation and precipitation. There are combined effects of differences in vapor pressure
of the light and heavy molecules, i.e., in the equilibrium between vapor and liquid, and in the mass-dependent transport across the air-sea
interface. The vapor pressure effect is about -8.5 to -10 , while the mass transport effect is an additional -2 to -4 , i.e., the overall shifts are
between -10.5 and -14 , the vapor being isotopically lighter than the water. In approximate terms:
- =-12 =
= -1
-12 =-.012= -1
=1-.012=.988
As water vapor condenses to form clouds there is an opposite effect, the first precipitation is isotopically heavier, leaving lighter water vapor
behind and
~1.012.
The marine atmosphere is fairly well buffered by the isotopic composition of the surface ocean (see below for the range of variation in the
surface ocean). Once vapor is transported over the continents, a Rayleigh condensation begins. This condensation generally occurs at lower
temperature than did the evaporation which provided mositure to the cloud and at lower temperature the fractionation between the two isotopes is
greater. In terms of the fraction of residual vapor, f:
Eq 24-7:
R is the 18/16 ratio in the vapor and the superscript * denotes the initial vapor. Consider a cloud containing vapor with O=-10 losing moisture
as precipitation according to equation 20-7 (As the cloud cools, the value of will progressively decrease; this is not accounted for here). The
progression and the liquid and vapor composition are shown here:
Figure 24-1
During glacial times storage of the isotopically light water produced at moderate (f<0.01) on the continents leaves behind an isotopically heavier
ocean.
By examining the isotopic composition of benthic forams, the effects of changing temperature and the change in the isotopic composition of
seawater can be separated. We look at benthic forams because in the deep ocean the temperature decrease during glacial times is constrained by
the freeezing point of seawater. There is a 1.9 difference in oxygen isotopic composition between glacial and interglacial times with the glacial
value being higher. The maximum temperature shift in the deep ocean is of order of the present day 1C to the freezing point of about -1.8C.
This 2.8C temperature shift can lead to an isotopic shift of no more than 2.8 x 0.22 /C or 0.6 . A shift of at least 1.3 remains to be
explained by the change in the isotopic composition of the water, and a still greater shift if the temperature difference in deep water is less than
2.8C.
Forming a mass balance in terms of the volume of the present day ocean, V , and the glacial storage of ice, V , and assuming that the mean
isotopic composition of glacial ice is -35 (consistent with ice core measurements):
Eq 24-8:
Dividing through by V
and letting x=V
/V :
Eq 24-9:
This equation can be solved to find that x~.035, corresponding to .035 x 4000 m ~ 140 m of sea level change. This decrease in volume by
evaporation leads to an average increase in ocean salinity to 1.035 x 35 =36.2 .
In summary: the oxygen isotopic signal in benthic forams is primarily an ice volume signal (and secondarily a temperature signal).
Another contributing factor to variations in seawater oxygen isotopic composition is the internal distribution of oxygen isotopes in the ocean. In
the present day ocean:
Figure 24-2
NADW maintains the signature of surface waters because of its mode of formation, by cooling-induced convection, whereas AABW does not
because of the salinity increase brought about by salt expelled during formation of sea ice. The isotopic shift experienced by benthic forams can
thus also be influenced by the history of formation of deep ocean waters.
Do planktonic forams have a temperature signal in addition to the ice volume signal? The ice volume effect dominates the overall signal, but at
particular locations an additional component can be recognized amounting to about 2C cooler temperatures in surface waters during glacial
times. However analysis of this signal depends on correctly estimating the surface water oxygen isotope distribution which adds considerable
error.
Carbon Isotopes
The important stable isotopes of carbon are:
12
13
C (~98.9%)
C (~1.1%)
Carbon isotopes are expressed relative to the PDB standard, belemnite carbonate from the Peedee formation, South Carolina. (Oxygen isotopic
composition of oxygen in carbonate may also be expressed on this scale; it is equivalent to +30.6 on the SMOW scale).
C provides a measure of oxidation of organic carbon. There is little fractionation of carbon isotopes between CO dissolved in the water and
solid CaCO :
Eq 24-10:
There is a much larger fractionation when organic carbon is produced:
Eq 24-11:
As deep ocean waters age, the evolve from being nutrient poor, oxygen saturated, CO2 depleted as oxidation of organic carbon and dissolution of
calcium carbonate occur. For all practical purposes, only oxidation of organic carbon is important in the isotopic balance because of the very
small fractionation between dissolved carbon dioxide and calcium carbonate. The concentration of some nutrient, say phosphate, can be used as a
measure of the extent of oxidation of organic carbon. Thus in the present day ocean, the isotopic composition of CO can be related to the
phosphate concentration through the mass balance:
Eq 24-12:
Eq 24-13:
Eq 24-14:
In the present day ocean this linear correspondence of isotopic composition of dissolved inorganic carbon and the phosphate concentration is
observed with the range being about 2.6 (-0.85 x 3 M phosphate).
CONTROL OF CARBONATE PRESERVATION
1. Review of the Seawater Carbonate System
Carbon dioxide is present in the atmosphere:
Eq C-1:
Carbon dioxide is soluble in water:
Eq C-2:
Because water is present in excess, we generally do not distinguish between these two forms, and define:
Eq C-3:
At equilibrium:
Eq C-4:
Carbonic acid is a weak diprotic acid:
Eq C-5:
Eq C-6:
Eq C-7:
Eq C-8:
The equilibrium constants are defined for seawater medium and are functions of temperature, pressure and salinity. For completeness, water also
acts as an acid:
Eq C-9:
Eq C-10:
We have 3 equations with 5 unknowns, therefore to solve the system of equations we need to specify two more conditions. One is mass
conservation:
Eq C-11:
The other is a parameter called alkalinity, Alk:
Eq C-12:
The alkalinity is essentially a restatement of electroneutrality. Consider a system composed of NaCl and NaHCO3 dissolved in water:
Eq C-13:
Correcting for the portion of the sodium with its origin in NaCl:
Eq C-14:
yields an equation analogous to the one defining the alkalinity. In general the left hand side is the sum of positive charge from cations of strong
bases (e.g., NaOH) less the negative charge from anions of strong acids (e.g., HCl). CT and Alk are measurable quantities. If they are known then
the system is completely defined.
These two parameters vary systematically in the ocean:
Figure C-1, from (20)
Why does it have this shape? It reflects the stoichiometry of three important processes. These are
gas exchange:
Eq C-15:
photosynthesis-respiration:
Eq C-16:
carbonate precipitation-dissolution:
Eq C-17:
Our particular interest here is in the interaction of seawater with solid CaCO3 (s). There are two crystalline forms of calcium carbonate, calcite
and aragonite. Calcite is the more stable and more common of the two. Thermodynamically:
Eq C-18:
Eq C-19:
In the ocean, the calcium/chloride ratio is constant (within 1%) so it is useful to introduce an effective constant:
Eq C-20:
To establish where calcium carbonate is stable, we need to compare this value of the carbonate ion concentration at thermodynamic equilibrium
(which will be a function of temperature, pressure and salinity) with the distribution of carbonate ion in the ocean. Consider this profile from the
southern Atlantic:
Figure C-2, from (20). Gridded data files of carbonate ion distribution, and the difference between the observed distribution and calcite solubility
are available along with information on how they were constructed from David Archer, University of Chicago.
The surface waters are supersaturated and the deep waters understaturated. Aragonite becomes undersaturated at a shallower level than calcite,
i.e., calcite is the stable phase at these temperatures and pressures. The oceanic distributions of carbonate ion concentration can be represented
relative to the value at saturation at that same temperature and pressure:
How would this saturation state be recorded in the adjacent sediment? The answer is not as simple as equating the saturation horizon with the
boundary between carbonate and non-carbonate bearing sediments. Rather an important issue is the rate of supply of carbonate to the seafloor
versus its rate of dissolution:
Thus two boundaries exist that can be recognized in the sediment:
the level of the onset of dissolution or lysocline (operationally defined as the depth at which solution-resistant forams
become dominant in sea floor sediments)
the level of "complete" dissolution or calcite compensation depth (CCD).
We next consider the processes controlling the position of these two horizons.
2. Dissolution of Calcium Carbonate at the Seafloor

Emerson and Bender (22) have modeled the processes involved in controlling carbonate dissolution in the sediments just below the sediment
water interface. As a basis for their model we will first consider the transport equations for solutes in pore fluids.
The governing differential equation for the Emerson and Bender model has the form:
Eq C-21:
The three terms on the right hand side are respectively:
difusion of carbonate ion
the efect on carbonate ion of the production of CO 2 by degradation of sedimented organic carbon
the efect on carbonate ion of dissolution of solid CaCO 3
We first need to find the coefficients in the last two terms relating the change in carbonate ion concentration to the amount of respiration and of
dissolution, respectively. The situation can be summarized by examining a plot of contours of carbonate ion concentration drawn as a function of
Alk and CT:
Figure C-6
In the upper left hand corner of this plot a one-to-one correspondence of the carbonate ion concentration with Alk-CT is evident:
Eq C-22:
which is expected at pH > 7.5 when carbonate ion is dominant, but as the lower right hand corner of the figure is reached, the pH approaches 7.5
and the carbonic acid term begins to make a measureable contribution. As degradation of organic material occurs, the solution composition
follows a horizontal vector to the right. Depending on where one is in the diagram, the change in carbonate ion concentration for a unit change in
total carbon varies. In particular the slope is rapidly changing as one approaches the Alk=CT line, i.e., the contours are spreading so there is less
change in [CO3-2 ] for a given change in CT or Alk as the solution composition approaches Alk=CT. Because the slope we are interested in is
changing rapidly in the region of interest to us, we take an alternative approach. Expressing the equilibrium as:
Eq C-23:
Between pH 6 and 9 the dominant form of inorganic carbon is bicarbonate and so we approximate:
Eq C-24:
Combining equations 22-22 and 22-24:
Eq C-25:
We can solve this quadratic to find a relationship between [CO3-2 ] and Alk-CT, a relationship that is not particularly sensitive to the value of CT
(Fig. C-7).
Figure C-7
At [CO3-2 ] = 70 micromolar (close to the saturation value at depth), the slope is about 0.5. We can then use the chain rule to relate this slope to
the coefficient terms in the model:
Eq C-26:
Thus
Eq C-27:
So the general model can be approximated:
Eq C-28:
Jorg and Jcarb are the rates (mol/cm3/sec) of organic matter degradation and carbonate dissolution.
To describe the rate of degradation of organic matter, Jorg, we postulate first order decay:
Eq C-29:
The balance between bioturbation of organic carbon associated with the solid phase (this mixing is described with an eddy diffusion coefficient
K) and rate of carbon degradation is:
Eq C-30:
Solving this differential equation for the boundary condition:
Eq C-31:
At steady state, the rain of organic material, Rorg, has to balance degradation integrated over all depths:
Eq C-32:
so that:
Eq C-33:
To describe the rate of calcite dissolution we adopt a first order reaction rate as a function of the saturation anomaly:
Eq C-34:
Emerson and Bender discuss the limitations of this equation in fitting data in the undersaturated part of the curve, where the rate is non-linear in
the extent of undersaturation. A second limitation is whether precipitation should be allowed under supersaturated conditions.
A series of model curves (a-d) show the expected pore water profiles for sediments at various water depths (with "a" deep and below the
saturation horizon and "d" shallow and above the saturation horizon):
Two sets of models are considered, one with no rain of organic carbon and the other with a 1:1 ratio of rain of organic carbon and rain of calcite,
a value close to that observed in deep sediment traps. The left hand panel in each figure shows the ratio of preservation to rain. The dashed curve
is for the case when calcium carbonate is allowed to precipitate when supersaturated so preservation can exceed the rain rate. In the lower left,
where the effect of organic carbon is included, as bioturbation is increased the effect on preservation is increased. The right hand side translates
this information in the carbonate content of the sediment. When the effect of organic carbon is not included, the lysocline is at the saturation
horizon and the separation between the lysocline and the CCD is about 300 meters. When the effect of organic carbon is included, the lysocline
can be kilometers above the saturation horizon and the saturation horizon and the CCD are coincident.
How do experimental data relate to these concepts?

Peterson, in 1966, deployed a vertical array of polished spheres made of optical grade calcite from a single mineral deposit (Figure C-11) on a
mooring south of Hawaii for just over four months. The spheres were lightly etched in hydrochloric acid prior to deployment to remove any
disordered layers produced by the grinding, and were weighed before and after the deployment.
Figure C-11. Calcite spheres (about 1.5 cm diameter) used by Peterson to measure the rate of dissolution in the deep Pacific.
At a depth of about 3700 m, roughly equivalent to the lysocline, the rate of dissolution of the spheres increased rapidly and continued to increase
to the sea floor (Figure C-12). The scatter in the dissolution rate from sphere to sphere was much larger than the weighing error. The source of the
scatter is unknown, although it may be related to heterogeneities in the calcite or variations in the boundary layer between the spheres and
circulating seawater.
Figure C-12. Rate of dissolution of calcite spheres in the water column south of Hawaii. Circles are rates averaged over five adjacent spheres.
There have been no comparable measurements within deep-sea sediments, so the Emerson-Bender model remains untested by direct observation.
Modeling equatorial Pacific sea floor sediments.
Oxburgh (1998) has taken a different approach by using the compositions and 14C ages of equatorial Pacific sediments deposited during the last
ice age and at the present sea floor to model Holocene carbonate dissolution. As shown by Figure C-13, carbonate dissolution in the bioturbated
layer ("homogeneous dissolution") produces a different distribution of 14C surface sediment ages from dissolution in the water column ("nonhomogeneous dissolution") or chemical erosion (mixing of old sediment up into the bioturbated layer when the dissolution rate exceeds the rain
rate).
Figure C-13. Effects of different patterns of calcite dissolution on the 14C age of surface sediments as a function of water depth.
The Oxburgh model assumes an abrupt change 10,000 years ago from carbonate-rich glacial sediments (Figure C-14) to the present rain rate for
carbonate measured in sediment traps from the region.
Figure C-14. Carbonate content of equatorial Pacific sediment cores as a function of water depth and age (glacial and Holocene, or postglacial).
"Glacial (AMR)" assumes a pulse of rapid accumulation of carbonate at the end of the glacial as proposed by Archer and Maier-Reimer (1994).
The model assumes a mixed layer depth of 11 cm (based on 14C measurements in many cores from the region) and a non-carbonate rain rate of
0.15 +/- 0.05 g cm-2 ky-1. It then calculates the mass of carbonate in the mixed layer every 100 years from 10 ky to the present, using an
appropriate mix of homogeneous and non-homogeneous dissolution to constrain the carbonate content to arrive at the present values (Figure C14), and the surface sediment 14C also to match the present values (Figure C-15).
Figure C-15. Measured 14C ages of surface sediments in the equatorial Pacific.
For a carbonate rain rate of 1.85 g cm-2 ky-1, the time evolution of the carbonate content of the mixed layer is given by a mass conservation
equation, which yields Pc:
Eq C-35:
with the initial condition that
where M is the mass of carbonate in the mixed layer, P is the preservation rate, h is the bioturbation (mixed layer) depth, is density, f is the
fractional carbonate content, R is the rain rate, D is the dissolution rate, B is the burial rate, and the subscript G denotes glacial, c carbonate, nc
non-carbonate, and ml mixed layer.
A separate conservation equation for 14C in the mixed layer, which yields Dnh, is given by:
Eq C-36:
with the initial condition
and where, in addition to the symbols above, is the mean life of 14C (8,267 years), and the subscript h denotes homogeneous, and nh nonhomogeneous. The model is summarized in Figure C-16.
Figure C-16. Terms in the model which determine the dissolution necessary to produce the present day carbonate content and 14C surface ages as
a function of age and water depth for modern equatorial Pacific sediments. See text for key to symbols.
Relative to the steady state flux required to produce the observed carbonate content and accumulation rate, all the model runs for all depths and
all variations of the non-carbonate rain rate (Figure C-17) showed a preservation deficit (i.e. excess dissolution), implying that this area of the
Pacific is currently not at steady state with respect to carbonate deposition for either the AMR or stepwise glacial-Holocene cases.
Figure C-17. Preservation flux from the Oxburgh model relative to the steady state preservation flux as a function of water depth for three noncarbonate rain rates. ND is for the observed late glacial carbonate content, AMR is for the Archer and Meier-Reimer pulse of carbonate
deposition at the beginning of the Holocene. Note that all calculations yield negative carbonate fluxes relative to steady state.
Even if the carbonate rain rate is adjusted to recognize a factor-of-two decrease from the equator to 4 N, all calculated preservation fluxes below
about 4,200 meters remain negative (Figure C-18), suggesting that glacial age carbonates are being actively chemically eroded at the present
time. In all cases, nonhomogeneous (water column) dissolution exceeds homogeneous (mixed layer) dissolution.
Figure C-18. Dissolution and preservation fluxes of carbonate as a function of water depth for a uniform calcite rain rate over the eastern
equatorial Pacific, and assuming a 50% decrease in calcite rain rate as a function of latitude from the equator to 4 N.
The Oxburgh results lend support to the Emerson and Bender (1981) model which is otherwise difficult to reconcile with steady state deposition
in the equatorial Pacific.
One other idea to bring into the discussion is the control of the level of saturation. Ignoring details of the accumulation in the zone between the
lysocline and the calcite compensation depth, assume that the preservation can take two possible values
if the seafloor is above the saturation horizon, then the preservation is equal to the rain rate
if the seafloor is below the saturation horizon, then the preservation is zero
If we balance the river flux of calcium into the oceans with the calcium carbonate preserved in sediments:
Eq C-37:
This implies:
Eq C-38:
This area of carbonate sediment has to be the shallowest 20% of the ocean basin which based on this hypsometric curve would place the CCD at
about 2.2 km:
Figure C-19.
There is considerable uncertainty in the 20% figure. If calculated directly from the above equation using the range of observed rain rates, one
would find values of 7-32%--the higher percentage corresponding to the the lowest rain rate. The actual distribution of sediment types indicates a
number at the high end of the range. This is not unexpected--the sediment trap measurements are biased toward productive areas where the rain
rate is greatest, but zones of upwelling are a relatively small portion of the waters overlying carbonate-bearing sediments.
Consider also the dependence on the river flux of calcium and the rain rate from surface waters. For constant calcium flux to ocean, the CCD
shoals with increased rain rate and deepens with decreased rain rate. With constant rain rate and changing calcium flux, an increased calcium flux
causes the CCD to deepen and a decreased calcium flux causes it to shoal.
ORBITAL GEOMETRY & OCEANIC CHANGE

World Data Center A-Paleoclimatology at the National Geophysical Data Center maintains an archive of data relevant to Pleistocene climate.
These data include the Earth's insolation, the CLIMAP and SPECMAP data sets, and many long records from the Ocean Drilling Program
(ODP).
Long-term climate change (tens to hundreds of thousands of years) appears to be paced and in some cases, forced, by variations in the latitudinal
distribution of incoming solar radiation due to changes in the earth's orbital geometry. This relationship is particularly striking for the pattern of
alternating glacial-interglacial conditions during the Pleistocene. The marine record has provided a critical test of one of these models,
formulated by Milankovitch in the early 1930s.
Components of orbital geometry - Eccentricity
The path of the earth around the sun is slightly elliptical, rather than circular. The eccentricity of the ellipse changes periodically with time, with
components having periods of ~100 ky and ~413 ky. The eccentricity is defined as the ratio of the focal length of the ellipse (the distance
between the foci) to the length of its major axis; the eccentricity of the earth's orbit about the sun has ranged from less than 0.01 to more than
0.05 over the past 600 ky (Figure OG-1, 2):
Figure OG-1. Exaggerated range of eccentricity of the earth's orbit.
Figure OG-2. Variations in the eccentricity of the earth's orbit during the past 600,000 years.
As is suggested by the exaggerated ellipses of Figure OG-1, variations in eccentricity change the flux of radiation received by the entire planet at
different times of the year, but there is no latitudinal effect. In actuality, because the earth's orbit is nearly circular, the radiation effect of
eccentricity by itself is very small.
Obliquity
The rotational axis of the earth is not normal to the ecliptic (the plane of the earth's orbit - Figure OG-3). The present tilt is 23.5 (the latitudes of
the Tropics of Cancer and Capricorn). This tilt or obliquity has ranged from 22.1 to 24.5 over the past 700 ky (Figure OG-4):
Figure OG-3. Exaggerated range of tilt of the earth's orbit.
Figure OG-4. Variations in the obliquity of the earth's orbit during the past 600,000 years.
The exaggerated tilts of Figure OG-3 illustrate how this parameter affects the seasonality of the earth (the length of day as a function of latitude
and season); when the tilt is large, there is more contrast between summer and winter, and the tropics and polar zones both increase in extent at
the expense of the mid latitudes. It is easy to see how this parameter interacts with eccentricity - summers will be hot in the hemisphere tilted
towards the sun when the earth is at its closest approach to the sun.
Precession
Over time, the earth's rotational axis rotates or precesses around the normal to the ecliptic. If the earth's orbit were circular, this would have no
climatic consequences. Because precession changes the locations on the orbit (i.e the times of year) where the hemispheres experience summer
and winter, however (Figure OG-5), the coupling of precession and eccentricity produce variations in the latitudinal distribution of radiation with
periods of 23 ky and 19 ky (Figure 17-6).
Figure OG-5. Exaggerated view of the impact of precession on the energy flux to the earth.
Figure OG-6. Variations in the precession of the earth's orbit during the past 600,000 years.
Why does the earth's orbital geometry change over time? Two factors are involved:
1. Changes in the Earth's spin due to tidal friction (resulting in a lunar recession rate of 10-9 m/sec).
2. Weakly chaotic changes in the orbits of the inner planets.
Estimates of 1. from direct laser ranging to the moon, from Babylonian, Hellenic, Chinese, and Islamic historical data on the precession of the
equinoxes, and from the number of days in a solar and lunar month recorded in corals spanning the past 450 million years, are reasonably well
known and the effects of the remaining uncertainty have been assessed.
Chaotic changes are inherently unpredictable, but their effect back to about 100 to 200 my is small (Figure OG-7). Prior to about 100 my, the
exact motion of the solar system cannot be calculated, so only the ratios of the orbital parameters, not the absolute age of a particular
configuration, can be calculated.
Figure OG-7. Variations in the dominant frequency of precession (left) and tilt (right) of the earth during the past 200 million years (after Berger
et al., 1992)
For the past five or so million years, very detailed insolation curves for 65 N have been prepared (e.g. Figure OG-8 shows the history for the
past 600,000 years) and correlated with the geologic record to allow individual events to be dated to within a few thousand years - an accuracy of
better than 0.1 percent over 5 million years. The relative ages of two closely spaced events can be determined even more precisely. Residual
uncertainties relate both to the orbital parameters, and the offset in time, or lag, between a change in incoming radiation and the response of the
geologic and oceanographic proxies that are measured in sediment cores.
Geological/geochemical proxies for paleoceanographic variables.

The sedimentary record preserves only indirect evidence of past oceanographic conditions. Much of the paleoceanographic work of the past three
decades has focussed on identifying and validating proxies for these conditions. Some of the proxies that we will refer to include:
Proxy
Microfossil census data

(factor analysis, transfer
functions)
Paleoceanographic Variable
SST, SSS, mixed layer depth, bottom water O2, bottom water corrosiveness (CO3=)
Fragmentation of foram tests Bottom water corrosiveness, CO3=*

Oxygen isotopic composition Vol. of continental ice sheets, SST, bottom water T, SSS
of foram tests
Carbon isotopic composition Productivity, water mass mixing
of foram tests
Cd in foram tests
PO4-3 content of SW, water mass formation
Mg in foram tests
SST
Ba in foram tests
Productivity
Opal flux
Productivity
Excess Al in carbonate ooze
Productivity
Quartz (dust) abundance and

Wind patterns and intensities
flux
Alkenone indices
SST
These will be discussed in more detail during the future lectures.
PROXIES: OCEANIC RECORDS OF PLEISTOCENE CLIMATIC CHANGE
Sediment sequences do not record climatic parameters (temperature, rainfall, atmospheric pressure, P-E, seasonality, etc.) directly. They can only
be inferred from proxies which must usually be calibrated empirically against modern data or other geologic variables. The choice of proxies
depends on the time interval under consideration. Examples of the proxies that have proved useful in interpreting Pleistocene (last 1.8
million years) climates include:
1. Oxygen isotopic composition of foraminiferal tests . As forams build their tests, the incorporated ratio of 16O to 18O varies over time
and space. Oxygen isotope systematics were covered last week, so suffice it to say here that the ratio measured in fossil tests in carbonate
sediments varies as a function of:
Isotopic composition of the water from which the carbonate was extracted
Water temperature
Salinity
Degree of dissolution
Degree of bioturbation
Vital efects
At most locations, the downcore variation is dominated by the preferential segregation of 16O in continental ice masses; thus, during ice ages
when large ice caps built up on North America and Eurasia, the oceans were enriched in 18O. This enrichment affected the entire ocean
simultaneously (to within the mixing time of the oceans - about 103 years).
This property of the oxygen isotope record has been used to create a "global standard" chronology (the "SPECMAP stack" - Figure PR-1) which
can be use both to correlate oxygen isotope profiles from around the world and to assign "absolute" ages to these profiles.
Figure PR-1. SPECMAP del18O stack for the past 600,000 years.
Figure PR-2 shows the correlation and age scale used by Imbrie et al. (1992) in their assessment of Milankovitch forcing of glacial cycles. Note
that the correlations are limited by sampling intervals and variable bioturbation, as well as by the mixing time of the oceans, and that the assigned
ages are model dependent.
Figure PR-2. SPECMAP data sets correlated by del18O records.

In addition to their correlation and chronology applications, oxygen isotope data are used to infer paleotemperatures (by subtracting the icevolume part of the signal) and to assess changes in temperature gradients (either vertically or ocean basin to ocean basin) by assuming that the
ice-volume signal is identical in contemporaneous samples, so can be subtracted out.
2. Alkanone indices. Haplophyte algae, particularly Emiliania huxleyi, synthesize unsaturated methyl alkenones (unsaturated ketones) with 37
C atoms and 2, 3, or 4 double bonds. These molecules are designated 37:2, 37:3 and 37:4, respectively. All are resistant to degradation in deepsea sediments.
The proportions of the three molecules are dependent on the temperature of the water in which they are synthesized. Two indices are used:
UK37 = (37:2 - 37:4) / (37:2 + 37:3 + 37:4), and
UK37' = 37:2 / (37:2 + 37:3) (for low latitudes where 37:4 ~ 0)
The indices have been calibrated against temperature using cultures of E. huxleyi from the North Pacific. The relation is linear, i.e.:
UK37 = A + BT, where A and B are empirical constants.
Problems:
The correlation loses precision (and is strain dependent) when 37:4 is high
UK37' is uncorrelated with T for temperatures below 10C in the Nordic seas (where 37:4 exceeds 5 percent of the
alkenones)
E. huxleyi only appeared 285-300 ky ago. Thus, older T determinations depend on the assumption that the modern
calibration is valid for the ancestors of E. huxleyi.
For the last interglacial, T estimates in the northeast Atlantic from alkenone indices are 12-17C versus 3-5C from
foram assemblages.
The abundance of 37:4 depends on salinity as well as temperature
Nevertheless, this parameter yields SST values that are less model dependent than most
3. Relative abundances of species in sea floor sediment samples , which provide insights to the environmental conditions in the
overlying water column where the organisms lived. Because the abundances of different species in a suite of samples are found to be correlated,
and because counting errors depend on the number of each species present, the raw abundance data normally are transformed and aggregated to
yield statistically independent assemblages, each of which can be treated as a single variable. The most common transformation / aggregation
procedure is called factor analysis. It involves:
Counting taxa in a limited size class of each sample
Transforming counts into unit vectors (where the sum of the squares of the numbers of each species in a sample is
reduced to one)
Generating the principal components of the suite of samples (objective analysis)
Rotating the principal components to maximize variance (yielding varimax "factors" or quasi end member samples)
Describing the samples in terms of the varimax factors.
The contribution of the various factors to the samples are found to be related to environmental parameters of the water in which the forams lived.
Figure PR-3a. Variation in the abundance of "polar" foraminifera (selected by factor analysis) with winter temperature.
Figure PR-3b. Variation in the abundance of "subpolar" foraminifera (selected by factor analysis) with winter temperature.
Figure PR-3c. Variation in the abundance of "subtropical" foraminifera (selected by factor analysis) with winter temperature.
Figure PR-3d. Variation in the abundance of "tropical" foraminifera (selected by factor analysis) with winter temperature.
A regression equation ("transfer function") is then calculated for the environmental parameter of interest (most commonly sea surface
temperature - SST) in terms of the factors, their squares and cross products. The error in the temperature estimates (about 2C - Figure PR-4) is
small relative to the range of variation in SST across the oceans, but is problematic in limited areas.
Figure PR-4. Comparison of measured winter temperature and temperature estimated from regression of foram assemblage factors.
Downcore samples are counted using the same procedures, and expressed in terms of the surface factors. The success
in fitting the surface factors to the downcore sample is expressed by the communality (essentially the percentage of
the sample unit vector explained by the factors).
The regression equation is then used to estimate the paleotemperature for that sample
NOTE: If the communality is low, there may be no analog to the downcore sample, so the estimated temperature will be suspect. In the eastern
equatorial Pacific, this problem led to estimates of the glacial SST that were high relative to estimates from adjacent land areas and groundwater.
The problem was addressed by pooling all the surface and downcore samples before factoring, then using only the surface sample to generate a
transfer function. The function, when applied to the co-factored downcore samples reduced the glacial SSTs by 2 or more degrees.
When samples are plotted in species space, the angle between them (expressed as cos theta) is a measure of their similarity (cos theta = 1 for
identical samples). This leads to an alternate method to estimate paleotemperatures: Downcore samples are compared to the surface set in species
space. The paleotemperature is estimated by averaging the temperatures of the most similar 3 or 4 surface samples. If the cos theta values
between the downcore and all surface samples are low, a no-analog situation is again indicated.
Figures PR-5 and PR-6 are examples of profiles of paleotemperature generated by the factor analysis /transfer function technique. In each case,
the original depth-in-core has been replaced by an age scale based on oxygen isotope correlations with the SPECMAP stack.
Figure PR-5. This temperature record from core V30-97 spliced to DSDP 607 is derived from planktonic foraminifera from 41 N in the eastern
North Atlantic (west of Spain), an area which was crossed repeatedly by the boundary between subarctic and subtropical waters. As a result, the
warm-season temperature at the site has varied by more than 15C.
Figure PR-6. This temperature record from core RC11-120 spliced with E49-18 is derived from radiolaria from 44 S in the Southern Ocean
(directly south of India), a site which has been south of the Antarctic Convergence for most of its history.
CAUTION: While the use of the transfer function technique has been largely responsible for transforming paleoceanography from a qualitative
to a quantitative science, the theoretical basis for the approach is limited. Mathematically, the technique requires that for F factors, there be F1 or less controlling physical processes. The validity of this assumption can rarely be established. Ecologically, forams and radiolaria live over
an extended depth range and secrete their tests over limited seasons. Thus, the relation between assemblage composition and SST is likely to be
indirect.
The paper by Watkins and Mix (1998) illustrates the difficulty of unraveling multiple correlated environmental parameters when the number of
factors is small (3 in their case). Although they calculated transfer functions to predict February and August temperatures (Figure PR-7), as well
as seasonal values for mixed layer depths and primary productivity, the significant correlations between these parameters (Table PR-1) casts
doubts on the value of their transfer functions for hindcasting paleoceanographic conditions.
Sample type
Net tows
Cores 24N-24S
# Samples
SST vs Prod
SST vs ML Depth
Prod vs ML Depth
23
-0.70
0.4
ns
141
-0.55
0.73
-0.84
Table PR-1. Correlations between environmental parameters at net tow and core sites where foraminiferal assemblages were used to generate
transfer functions to predict February and August values of these parameters.
Figure PR-7. Comparison of observed equatorial Pacific sea surface temperatures (SST) with values estimated from foraminiferal assemblages in
mixed layer net tows.
This difficulty in using transfer functions to make hindcasts can be expressed through the linked equations:
Assemblage = F1*(environmental variable), and
Environmental variable = F2*SST
which can only be inverted to yield paleo-SST from assemblage data if F1 and F2 are constant over time. This is very hard to prove.
4. Carbon isotopic composition of foraminiferal tests . As with the oxygen isotopes, the carbon isotopic composition of the calcite tests
of foraminifera, which will be discussed in more detail next week, depends on:
Isotopic composition of the water in which they live
Temperature (minor efect)
Salinity
Carbonate chemistry
Vital efects
Even though continental glaciers do not have a direct impact on the carbon isotopic composition of seawater, the signals of oxygen and carbon
isotopes resemble each other. In the case of carbon, however, the primary global glacial-interglacial signal is attributed to changes in the mass of
the terrestrial biosphere. Plants preferentially take up 12C. Thus, during glacials, when the terrestrial biomass was greatly reduced (due,
apparently, to reductions in the biomass of both temperate and tropical forests), the ocean was relatively depleted in 13C.
Note the del convention:
del13C(sample) = 1000 * (13C (sample) - 13C (standard)) / 13C (standard) (units of per mil)
Foraminifera metabolize C and use it to build protoplasm as well as calcite. Not surprisingly, therefore, vital effects are of much greater
magnitude for carbon than for oxygen isotopes. Interspecific differences are large and often swamp the global signal. Because organic carbon is
heavily depleted in 13C, surface waters in productive areas become enriched in 13C whereas regeneration (oxidation) of the 12C carbon in deep
waters leads to progressive depletion in 13C as the waters get older and older.
As with oxygen, changes in lateral and vertical del13C gradients over time provide insights to the past behavior of the oceans. Figure PR-8 shows
the percentage of North Atlantic Deep Water at 3500 m in the eastern North Atlantic west of Spain. This is calculated by a simple mixing model
between a northern North Atlantic benthic record and one from the equatorial Pacific. The relative supply of NADW to this site was greatly
reduced during glacial maxima.
Figure PR-8. Estimated proportion of NADW at DSDP site 607 (eastern North Atlantic) based on fluctuations in the 13C content of benthic
foraminifera (NADW water is 13C enriched).
Figure PR-9, on the other hand, is less easily explained. This plot shows the difference in benthic 13C between an Antarctic core and one from the
equatorial Pacific. As expected from the general movement of bottom waters, the Antarctic today is enriched in 13C relative to the deep Pacific.
During glacial times, however, the Antarctic is relatively depleted in 13C, a topic that we will return to later.
Figure PR-9. Del13C in benthic foraminifera from an Antarctic core relative to contemporaneous samples from an equatorial Pacific core. Both
the secular trend and the negative glacial values are unexplained.
5. The flux of windblown dust (often identified by its quartz content, as we discussed in the first lecture) can be determined if the
accumulation rate and bulk density of a core are known. Figure PR-10 shows that for the area off the Arabian peninsula, glacials were marked by
high deposition rates of dust - an indication of enhanced supply or more capable monsoon winds (or both). Note the inverted scale and the
clipping of the signal where values approach zero.
Figure PR-10. Flux of dust off the Arabian peninsula to core RC27-61. Note the inverted scale (low fluxes during interglacials).
6. The addition of carbon dioxide via the oxidation of organic matter to deep water as it "ages" (moves further from its source)
results in increasing dissolution of foram tests and other carbonate particles. One indication of this dissolution is an increase in the proportion of
fragmented foram tests. Figure PR-11 shows the variation in fragmentation down a core from the Caribbean, which is presently overlain by North
Atlantic Intermediate Water. In this case, it appears that the overlying water was younger (less corrosive) during ice ages than it is today.
Figure PR-11. Percentage of foram fragments (a measure of bottom water corrosiveness) in Caribbean core P6408-9.
Broecker and Clark (1999; Paleoceanog. 14:596) have suggested that the percentage of carbonate >63 m in a sediment sample is a direct
measure of the pressure-normalized carbonate ion content of the bottom water (which they call CO3-2*) when the sediment was deposited.
CO3-2* = CO3-2 + 20 (4-h) mol.kg-1
Where h is the water depth in km and 20mol.kg-2 is the increase with water depth in the solubility of calcite. This is equivalent to normalizing
all CO3-2 values to 4 km. Because of postdepositional fragmentation, the validity of this index earlier than the past few thousand years is still
unclear.
7. Trace elements in foram calcite are in thermodynamic equilibrium with the seawater from which the calcite is deposited. The Cd/Ca
ratio, which is highly correlated with the PO4-3 content of the water, and the Ba/Ca ratio, which appears to be related to biological productivity are
discissed in future lectures. The Mg/Ca ratio incorporated in the tests appears to depend only on the water temperature (Figure PR-12; Lea et al.,
Science 2000, 289:1719).
Figure PR-12. Effect of SST (A) and postdepositional dissolution (B) on the Mg/Ca ratio in the calcite tests of the foraminifera Globigerinoides
ruber (from Lea et al. 2000, op. cit.)
Current analytical techniques allow temperature changes of less than 0.5C to be measured. Sea-floor dissolution preferentially removes Mg-rich
calcite(Figure PR-12), so that data for fossil shells must be corrected to yield hindcast SSTs. By combining Mg/Ca and del 18O data from the same
samples, it becomes possible to separate the temperature and water composition components from the del18O signal.
When all the proxy information is put together (Figure PR-12), there is a high degree of internal consistency.
Figure PR-13. SPECMAP proxy data for the past 600,000 years, correlated by 18O records and showing the strong glacial-interglacial changes in
all environmental variables.
The following changes occur during a transition from a glacial to an interglacial state:
Radiation at 65 N increases
Ice volume decreases
Dust fluxes to the North Pacific, Arabian Sea, and Antarctica decrease
Sea surface temperature increases almost everywhere
13
Carbonate dissolution increases in NAIW but decreases in the Southern Ocean
13
Deep waters are relatively depleted in nutrients in both the North and South Atlantic
13
Atmospheric carbon dioxide increases.
C in surface waters is depleted relative to intermediate and deep waters
C increases in North Atlantic relative to Pacific deep waters
C is enriched in the entire ocean and in Antarctic relative to Pacific deep waters
The transition from an interglacial to a glacial state is more complex (Figure PR-13).
Figure PR-14. Patterns of climate change over the past 150,000 years. The signs of observations identified on the right are reversed as needed so
that the interpretation of each time series has a positive upward sense of change toward an interglacial condition. For variable 20, for example, an
upward shift in the curve means that deep waters in the Antarctic are becoming more 13C enriched than in the Pacific.
During the transition from the last interglacial (which began about 130,000 years ago) to the last glacial, all the proxies show a sharp onset of
interglacial conditions (an event designated termination II, by analogy with termination I, the transition to the current interglacial), but some
proxies indicate a brief (10-15 ky) interglacial, whereas others indicate a much longer (about 60 ky) warm period. The model developed to
explain this paradox will be discussed after the proxy records have been revisited in the frequency domain.
PHASE RELATIONSHIPS OF PROXY VARIABLES
Following the brief introduction to time series analysis three weeks ago, a few points deserve additional emphasis:
For valid spectral analysis, the time series must be statistically stationary (i.e. not evolving through time). This is not
strictly true for oxygen isotope series (see Figure PH-1 for example), but can be handled to some extent by detrending
and prewhitening. Where nonstationarity is one of the phenomena under study, the series can be analyzed by wavelet
analysis or in (overlapping) segments.
Figure PH-1. Oxygen isotope record for benthic foraminifera from ODP site 849 (eastern Pacific) showing that neither the mean nor the variance
of the data has been constant (stationary) over the past 4.5 million years.
Filtering reveals the structure of a time series in a particular frequency band. Because the filters are of finite length,
however, information is lost at the ends of the series. Inasmuch as we are particularly interested in the youngest end of
most paleoceanographic and paleoclimatic time series (i.e. the present), various techniques are used to fill in the ends.
These depend on assumptions about the properties of the time series and should be viewed with reservations.
As Russ mentioned, the autocorrelation function of a geologic time series, rather than the series itself, yields a more
robust spectrum of amplitudes, while preserving all the spectral information.
Substitution of a second time series yields the analogous cross correlation function, which can also be fourier
transformed to yield a coherence spectrum and a phase spectrum (which indicates the degree to which shared spectral
peaks are in phase (Figure PH-2)).
Figure PH-2. Typical information provided by cross spectral analysis. The top panel shows the separate spectra of the two time series, the middle
panel shows the cross spectral coherence of the two time series (with significance levels), and the bottom panel shows the phase relation of the
two spectra (note that for the significantly coherent peaks, there is no phase lag in this case).
Both a significant peak amplitude and a stable phase spectrum over the same frequency band are necessary to provide
confidence in the reality of a cross spectral peak.
As discussed earlier, the SPECMAP group (Imbrie et al., 1992, 1993) developed an averaged del18O curve for benthic foraminifera, which they
compared to the 65 N June history of incoming solar radiation (Figure OG-8) and to numerous time series of proxy variables, a number of which
were discussed during the last two lectures. The results of their comparisons in the frequency domain are summarized in Figure PH-3.
Figure PH-3. Amplitude-density specta of 20 long climatic time series showing the dominance of variance in bands centered at about 100, 41,
and 23 thousand years. The cross hatched areas are the coherence spectra of the time series with the SPECMAP del18O stack, whereas the solid
lines are the spectra for the individual time series. The key to the series illustrated is:
1. Milankovitch orbital parameters
2. SPECMAP del18O against 65 N June radiation
3. Loess (dust) record from China against del 18O
4. Dust record in the Arabian Sea
5. SST at 50 N in the North Atlantic
6. SST at 41 N in the North Atlantic
7. SST in the southern Indian Ocean
8. Delta del13C between equatorial Pacific surface and deep waters
9. Delta del13C between equatorial Atlantic surface and deep waters
10. Delta del13C between equatorial Pacific intermediate and deep waters
12. Fragmentation of foraminifera in the Caribbean
13. Calcite preservation in the southern Indian Ocean
14. Proportions of Atlantic and Pacific deep water at 41 N in the Eastern Atlantic
15. Cd/Ca ratio at 31 S in the South Atlantic
17. Cd/Ca ratio at 42 N in the north Atlantic
20. Delta del13C between the Antarctic and equatorial Pacific deep waters
For both the 41ky (tilt) and 23ky (precession) peaks the coherence is strong and persistent enough for Imbrie et al. to propose a simple linear
forcing of climate by variations in these two orbital parameters. We will return to their chain of logic, which is largely self consistent and
persuasive. However, one should always bear in mind the old statistical axiom that correlation does not prove causation. Cross spectral analysis is
a correlation technique.
For the eccentricity period (100ky), there is strong coherence between del 18-O and the other variables, and between del18O and the eccentricity
of the earth's orbit. Because eccentricity contributes so little to the variation in incoming radiation at 65 N (or any other latitude), however, there
is no significant coherence between this variable and del18O in the 100ky band. Thus, the simple linear forcing attributed to precession and tilt
cannot apply to eccentricity. This lack of a clear correspondence between forcing and response has led to the development of models of nonlinear response, as well as suggestions that resonant oscillations within the earth's atmosphere-ocean-ice system or passage of the earth through
an interplanetary dust ring about every 100,000 years are responsible for the major glacial-interglacial frequency.
The phase relations between the various time series is illustrated by the filtered records in Figures PH-4 (23ky band - eccentricity) and PH-5
(41ky band - tilt). The key is the same as above. The slopes of the dashed lines indicate that the series are not in phase with the presumed forcing.
They have been arranged so that the phase lag increases from top to bottom. The lags in degrees from the del18O stack are shown on the left.
Figure PH-4. The 23-ky climate cycles, bandpassed filtered from Figure PH-3. Vertical arrangement is according to phase difference relative to
del18O stack. Dashed lines show systematic phase offsets relative to 65 N radiation. Superimposed, phase-aligned records show high coherence
with orbital forcing.
Figure PH-5. The 41-ky climate cycles, prepared as the 23-ky cycles (Figure PH-4).
Note that lags in a phase spectrum are expressed in degrees (or radians) rather than in years, as the lag is a function of frequency. Figure PH-6
shows the degree of coherence and phase relation of the Milankovitch peaks in proxy records versus the SPECMAP del18O stack and of the stack
versus orbital parameters and June incoming radiation at 65 N.
Figure PH-6. Statistical summary of cross spectra of climatic time series. Signs have been changed for consistency with Figure PH-3. The del 180
stack is compared to orbital indices; other variables are compared to the del18O stack. Positive phase means that the variable lags the reference
variable. k is coherence (in brackets if not significant at 80% level), phase angles are shown with 80% confidence intervals
To present an enormous amount of data in a compact form, Imbrie et al. make use of "phase clocks" for each frequency band (Figure PH-7) The
vectors on such clocks make it easy to visualize relative leads (counterclockwise offsets) and lags (clockwise offsets) as well as groupings within
multiple data sets.
Figure PH-7. Phases of orbital, radiation, and climatic cycles. As shown in Figure PH-6, some proxy variables lead del18O, others lag. In most
cases, the behavior is the same for all three frequency bands
At all three Milankovitch frequencies, certain sets of proxies tend to group together. In the case of the 100ky (eccentricity) band, the appropriate
"zero point" in the eccentricity cycle is unclear. Imbrie et al. chose "maximum eccentricity," even though there is no a priori reason to suppose
that this is the key point in the cycle. This zero point occurs after a number of important environmental changes, suggesting that minimum
eccentricity or some intermediate point in the cycle may be more significant.
The events that lead minimum ice volume are:
Minimum deep-water Cd/Ca at 31 S in the South Atlantic
Maximum SST at 44 S in the Indian Ocean
Maximum del13C depletion in the surface E equatorial Pacific and tropical Atlantic
Maximum del13C enrichment in the Antarctic
Those that lag minimum ice volume are:
Maximum Caribbean intermediate water carbonate dissolution
Maximum oceanic del13C enrichment
Minimum dust fluxes
Maximum SST at 50 N in the North Atlantic
Maximum ventilation at 3.5km at 41 N
The leads and lags, in years, are the same (within measurement uncertainties) for the 23ky (precession) and 41ky (tilt bands). Note that :
Lag time (years) = Period of the spectral peak * Lag (degrees)/360.
Thus, for the precession peak at 23ky, a lag of 40 degrees converts to 2,560 years. The same lag in years would be 22 degrees for the tilt peak at
41ky.
Imbrie et al. concluded that the pattern of lags requires a four state system to explain it. These states they label:
Interglacial
Preglacial
Glacial
Postglacial
Based on the phase relations from the frequency domain analyses, and the nature and magnitude of the changes at critical boundaries (such as the
terminations) in the time domain records, Imbrie et al. proposed the following attributes for these four states:
Interglacial (Figure PH-8)
Figure PH-8. First stage of a generic Milankovitch gl;aciation cycle, according to Imbrie et al. (1992).Symbols for Figures PH-8 to 11shown
below. AA = Antarctic ocean, NOR = Nordic ocean. For the 23-ky and 41-ky bands, the forcing is 65 N radiation, for the 100-ky band, the
forcing is the mode of ocean overturning.
During this state, both sea ice and northern hemisphere continental ice caps are at their minimum extents. Deep water is formed in the North
Atlantic both by cooling of relatively saline water in the Nordic (Norwegian + Iceland + Greenland) Seas ("Nordic Heat Pump") and by cooling
in the Labrador Sea and open Atlantic ("Boreal Heat Pump"). The deep water moves south to the Antarctic, where sea ice is also at a minimum
and the westerlies are close to the tip of South America. There the NADW and southern deep water masses mix to create the deep and bottom
waters of the Pacific and Indian Oceans. These waters upwell in the North Pacific and Indian Oceans and return to the North Atlantic via surface
currents which transport heat and salt to the northern Atlantic source regions where the "grand tour" began.
Preglacial (Figure PH-9)
Figure PH-9. Second stage of a generic Milankovitch glaciation cycle (see Figure PH-8 caption).
This state is triggered by decreased northern hemisphere radiation, which leads to freshening of the Nordic Seas, thereby terminating the
formation of NADW and shutting down the Nordic Heat Pump. The reduction if heat from the north leads to growth in Antarctic sea ice, which
pushes the wind field and water masses north. This appears to reduce the rate of overturn of Antarctic waters, which in turn could draw down
atmospheric carbon dioxide values (to be discussed in more detail later).
Glacial (Figure PH-10)
Figure PH-10. Third stage of a generic Milankovitch glaciation cycle (see Figure PH-8 caption).
This state results from the continuing growth of the northern ice caps which steer northern hemisphere winds. These produce more convection in
the boreal Atlantic, and possibly Pacific (analogous to that occurring in the Labrador Sea today) thereby providing "younger" NAIW and NPIW.
The loss of water to the ice caps eventually lowers sea level to the point where ice sheets are grounded on large areas of the continental shelf.
Deglacial (Figure PH-11)
Figure PH-11. Fourth (final) stage of a generic Milankovitch glaciation cycle (see Figure PH-8 caption).
As high latitude northern hemisphere radiation increases, the atmosphere and surface waters warm, evaporation increases, and sea ice retreats
from the Nordic Seas. The initial retreat of the glaciers modifies the winds to drive warm saline Atlantic water northward. This accelerates glacial
melting and provides the buoyancy flux necessary to turn on the Nordic Heat Pump and form NADW, apparently to a greater extent than during
full interglacials, thereby transporting heat to Antarctica to accelerate sea ice melting there and enhanced ventilation of the southern deep water,
thereby releasing trapped carbon dioxide and setting the stage for a catastrophic collapse of the continental ice sheets.
The evidence for these states and a model to describe the chain of events in a full glacial-interglacial cycle will be revisited after the productivity,
and NADW sub-systems have been examined in more detail.
CARBON-14 SYSTEMATICS
Because of its half-life of 5,730 years and its involvement in the carbon cycle, 14C is a uniquely valuable "clock" for studying carbonate
deposition and the transition from the last glacial period to the present interglacial.
Source - 14C is produced in the upper atmosphere as a result of incoming cosmic rays (high energy protons). These protons collide with
atmospheric components, generating lower energy neutrons in the process. Some of these neutrons, in turn, collide with the nuclei of 14N atoms:
14
N (7p + 7n) + n =
14
C (6p + 8n) + p
Decay - dN/dt = lambda * N where lambda is decay constant
If N = N0 at t = 0, then N = N0 exp(-lambda*t)
For N/N0 = 0.5, t(1/2) = ln 2 / lambda

Decay occurs by the nuclear reaction:
14
C-e=
14
Until recently, detection has required the measurement of the electrons released by this decay. Because of the small numbers of disintegrations
per hour, this requires extended counting times (days) to get accurate dates, and careful shielding to minimize background counts.
Currently, accelerator mass spectrometry (AMS), which detects the 14C atoms by their mass, rather than their radioactive decay, is the preferred
detection method due to its sensitivity and the small amount of sample required (a thousandfold less than for beta counting). However, the most
accurate and precise 14C activities still come from beta counting.
Problems - The conversion of 14C activities (or concentrations) into ages is complicated by the following phenomena:
Variations in N0.
Will be addressed later.
Differences between 14C/C in the atmosphere and the same ratio in HCO3- in natural waters. The ratio in
HCO3- is always less because of "reservoir efects." On land, these are caused by the dissolution of "dead" (ancient,
non-radioactive) CaCO3, whereas in the ocean, decay due to the overturn time of the oceans (of order 1000 years)
lowers the proportion of 14C.
Thus, all 14C data derived from HCO3- or CO3-2 or oceanic organic carbon must be corrected in order to yield meaningful ages..
For surface tropical and temperate waters, contemporary shells and corals have 14C ages of 400 200 years.
Contamination by young carbon. Because of the exponential decrease in 14C with time, a given fraction of modern
contamination is 16x as serious for a 34,000 year old sample (which has decayed for 6 half lives) as it is for an 11,000
year old one (which has decayed for only 2 half lives). For 45,000 year old samples, even 0.5% contamination seriously
degrades the result, which is why this is close to the upper limit of the applicability of this technique.
Methods to correct/verify 14C ages
Tree rings. By careful cross correlation, a continuous record of tree rings from German oak trees and logs has been
created for the past 11,900 years. By measuring the 14C activities of rings of known age, a curve relating calendar to
years has been built up (e.g. Figure C14-1).
14
Figure C14-1. Corrections to be added to 14C dates to convert them to calendar dates, based on tree-ring analyses.
Lake varves. Annual layers (varves), particularly in lake sediments, can often be correlated within and even between
lakes. The lakes must be anoxic and the varves characteristic enough to allow for correlation. Few varve sequences are
continuous to the present, so must be calibrated by counting varves between 14C-dated levels (to ensure that the layers
are, in fact, annual), and must then be spliced into the tree-ring sequence on the basis of 14C activity of included wood
fragments. Questions have been raised as to the possibility of detecting all hiatuses in varve records
Cariaco Basin varves. Sediments from the Cariaco Basin (an anoxic basin on the margin of Venezuela) contain varved
sediments spanning the last glacial-interglacial boundary (Peterson et al., Hughen et al., Science (2000) 290: 1947,
1951). The surface water "reservoir age" of 420 years matches the open Atlantic and the 14C record from the foram G.
bulloides from the varves can be correlated almost perfectly with the tree ring data from 10,200 to 11,900 years BP
(Figure C14-2).
Figure C14-2.Calendar and 14C ages for German pine data (red) and G. bulloides from Cariaco Basin core PL07-58PC (black) which have been
aligned by sliding the varve data within a moving 1370 year window to maximize the correlation between the two data sets. YD = Younger
Dryas, PB = Preboreal.
In addition, the Cariaco varve sequence extends the record to almost 15,000 years, providing a detailed chronology and 14C record covering the
end of the last glacial to the Preboreal (Figure C14-3).
Figure C14-3. Calendar versus 14C ages for Cariaco Basin varves compared to pine, coral, and Japanese lake varve data (see Hughen et al. op.
cit.). GL = Glacial, B/A = Bolling/Allerod, YD = Younger Dryas, PB = Preboreal.
Ice cores. As snow accumulates on ice caps, it develops an icy crust each summer. This crust persists as a zone of
coarser crystals in the ice that forms as the snow is compressed by later years' accumulations. The annual layers can
be detected visually in young ice, and are apparent from the oxygen isotope record even in old ice (Figure C14-4). The
annual layers have been counted back to about 40,000 years with an error of about 200 years.
Figure C14-4. Annual layers in an ice core revealed by detailed oxygen isotope profiles. Both the thickness of the annual layers and the amplitude
of the oxygen isotope signal decrease with increasing age.
The layers in ice cores cannot be dated directly by
14
C because:
The amount of trapped CO2 is too small, even for AMS analysis.
14
The ice contains ancient CaCO3 dust that adds "dead" carbon to the record
C is created within the ice by cosmic ray bombardment of oxygen nuclei
However, the dates of rapid climatic fluctuations during the last deglaciation (Figure C14-5) can be determined and compared with 14C
dates of the same events in other records.
Figure C14-5. Dated events during the transition from the last ice age to the present interglacial. The names are derived from outcrops in Europe
and appear frequently in the literature. H-1 and H-2 are the first two Heinrich events (enhanced ice-rafted debris in the North Atlantic) which will
be discussed in a subsequent lecture.
Corals dated by 230Th. Corals take up U in almost the seawater U/Ca ratio, but are totally free of Th, one isotope of
which is a decay product (daughter) of 234U. The decay sequence:
238
U (t1/2 = 4.47 billion years) - 234U (t1/2 = 244,000 years) - 230Th (t1/2 = 77,000 years) provides a way for dating corals (by measuring the
ingrowth of 230Th) as well as checking on the extent to which the coral has been closed to U and Th gain or loss since its formation (by
measuring the age-corrected 234U/238U activity ratio to see if it deviates from the seawater value of 1.15).
230
Th dates are not as precise as 14C dates, but are more accurate because the systematics of the system are much simpler. The 230Th dates
on corals overlap with the tree-ring series, and show the same discrepancies between calendar and 14C years (Figure C14-6).
Figure C14-6. Discrepancy between calendar and 14C ages for the past 25,000 years based on tree ring analyses (open circles) and 230Th dating of
corals (solid circles with error bars).
The difference between the 230Th and 14C dates of corals are consistent with the 14C corrections inferred from the ice-core dating of the BollingAllerod warm interval and Younger Dryas cool interval (the correction increasing from about 1 ky at 11,000 years to 2 ky at 14,000 years).
Scandinavian varve data suggest the opposite trend, raising questions as to the continuity of the varve sequence.
The difference between Th/U and uncorrected14C ages create significant differences in the history of the transition from the last glacial to the
present interglacial (e.g. Figure C14-7).
Figure C14-7. Chronology of the rise of sea level from the last glacial to the present based on uncorrected 14C and Th/U dating of corals. The
difference in sea level can be as much as 35 m during the interval of rapid rise.
Why do 14C years not correspond to calendar years? Two possible causes have been proposed:
1. Changing reservoir effects. If the deep waters of the ocean were to be more efectively isolated from the surface
waters, the 14C system would be biased towards the kinds of young ages seen for the deglaciation events, for example.
However, such a change would increase the apparent 14C age discrepancy between contemporaneous planktonic and
benthic foram tests in deep sea cores. While small glacial-interglacial diferences are seen (Figure C14-8) they are
nowhere near large enough to create the 14C errors observed.
Figure C14-8. Relative 14C/C ratios in the atmosphere and deep and surface ocean waters today and during the last glacial. The data require that
the atmospheric 14C/C ratio during the last glacial be about 1.4 times its 1850 value.
2. Changes in the source term. Although the source of cosmic rays is unlikely to have fluctuated on the scale
required to produce the observed variations in 14C production, the number striking the atmosphere is afected by
variations in the intensity of the earth's magnetic field. Paleointensities can be recovered from the magnetic properties
of dated lavas and pot sherds, both of which recorded the intensity of the ambient field as they cooled through their
Curie points. A compilation of the inferred changes expressed in terms of the predicted efect on 14C formation (Figure
C14-9) shows fair agreement with the coral data. More work is needed to validate the reliability of both data sets.
Figure C14-9. Perturbations in 14C ages predicted from the reconstructed record of the intensity of the earth's magnetic field, compared to the
results from 230-Th-dated corals (solid squares with error bars).
There are still questions as to the appropriate corrections prior to about 15,000 years, but the younger corrections, which cover the most recent
deglaciation, are well established.
Uses - 14C, because it adds a time component not provided by isotopic and trace element tracers, has been used extensively for dating carbonatebearing sediments and for constraining models of vertical and horizontal mixing in the oceans.
Atmospheric nuclear testing, particularly during the 1960s, added 14C that swamped the natural input. The origin and magnitude of this
"spike" is well known, so it can be used to measure the rate and locations of penetration of atmospheric carbon dioxide into the upper ocean, an
important parameter in assessing the role of the ocean in taking up greenhouse CO2.
The burning of fossil fuels, which adds "dead" carbon to the atmosphere, has the opposite effect, by reducing its 14C/C ratio. This "Suess"
effect began before the turn of the century. Hence, 14C activities are commonly compared to the 1850 atmospheric 14C/C ratio.
PALEOPRODUCTIVITY
Paleoproductivity, because it places important constraints on past ocean circulation, nutrient distributions, and on the history of the oceanic
carbon cycle, is a major area of paleoceanographic research. The tracers or proxies that have been used to estimate past productivity include:
13
Organic carbon (Corg)
Foram indices
Opal.
Marker organic molecules
Cd/Ca
Sr
Ba
Al
Each has its limitations, and different proxies commonly hindcast different productivities.
Carbon-13
As discussed in an earlier lecture, Corg is about 20 per mil depleted in 13C relative to dissolved CO2, whereas calcite shows only a slight
enrichment (about 1 per mil). Thus, the 13C values of planktonic forams in the sediments beneath productive surface waters will be enriched in
residual 13C whereas the contemporaneous benthic forams will record lowered values because of dilution by 12C from settling Corg that is oxidized
in the deep water. Because the deep water records the input of oxidative CO2 since it left the surface, the difference between planktonic and
benthonic foram 13C values can only be converted to productivity if the "age" of the bottom water is known.
The relation between 13C (deep water - the preformed value in surface water) and deep-water phosphate (approximately 0.85 per mil per
micromole phosphate) does provide a way of reconstructing past phosphate concentrations.
Concerns:
13
C in benthic forams may not provide accurate deep-water records because dissolution seems to preferentially remove 13C-enriched tests.
McCorkle et al. (1995) found that benthic forams in cores shallower than 2.5 km from the Ontong-Java Plateau in the western equatorial Pacific
are about 0.2 per mil enriched in 13C relative to bottom water, whereas the same species in cores deeper than 4 km are about 0.2 per mil depleted.
Benthic forams show marked interspecific differences and vital effects comparable to the range of values found in the ocean. The stability of
these effects over time is uncertain.
Analyses of living benthic forams show lower 13C in infaunal tests than in epifaunal tests. This is consistent with regeneration of Corg within
the sediments, but complicates the interpretation of fossil assemblages.
The paradox of the glacial 13C enrichment of equatorial Pacific deep water relative to Antarctic water was mentioned earlier. The gradient
between equatorial and northeast Pacific glacial 13C values (Figure 28-1) does not suggest that the answer lies in the glacial formation of North
Pacific deep water.
Figure 28-1. Comparison of the
13
C records from equatorial (V19-30) and northeast Pacific (W8709A-8) cores spanning the last glacial cycle.
Organic carbon
Primary productivity is the rate of generation of organic material, so Corg should be the most direct index of paleoproductivity. Complications
include:
1. Loss of Corg by oxidation in the water column. Empirically it is found that Corg flux at depth z = 0.02 (primary productivity)**3/2 *
(100/z + 0.025)
2. Additional oxidative loss in the upper part of the sediment column (Figure 28-2).
Figure 28-2. Typical profile of Corg in a North Pacific sediment core, showing the exponential decrease in the near-surface sediments.
Berger and Herguera give the relation for Corg below the bioturbated layer as:
Log (Corg) = 0.7 log (sea floor flux) - 0.24 (Figure 28-3)
Figure 28-3. Relation between Corg in sediments and the sediment flux to the sea floor.
Berger and Herguera note that the 3/2 and 0.7 exponents of the two equations almost cancel, so that:
Primary productivity (approx.) = K * Corg.

K is about 200, but is location dependent.
Other authors have found that the exponent of Corg in this equation lies between 0.5 and 1. Sarnthein et al. (1988) have related the seafloor
flux to new production (Pnew)by the relation:
Corg = 20.56(Pnew)**0.605 * z**-0.554
3. Continued slower Corg loss in deeper sediments, which has the effect of reducing the amplitude of older climatically induced
fluctuations (e.g Figure 28-4). Emerson and Hedges have shown that even 100,000 year old organic matter continues to oxidize (Figure
28-5), suggesting that no depth of burial is sufficient to preserve Corg.
Figure 28-4. Decline in magnitude and amplitude with increasing age of Corg peaks in equatorial Pacific sediments, suggesting continued
degradation. Numbers above Age scale are oxygen isotope stages.
Figure 28-5. Degradation rate of Corg as a function of its age. Circles are laboratory experiments, squares are field experiments, and crosses are
field data and model.
Even in anoxic sediment, the concentration of Corg declines exponentially with depth (Figure 28-6), suggesting that degradation occurs
even under the best conditions for preservation.
Figure 28-6. Inferred degradation of Corg in anoxic sediments from Santa Barbara Basin.
4. Corg preservation is dependent on sedimentation rate (Figure 28-3), leading to empirical relations of the form:
Primary productivity (approx.) = Corg flux/0.003 (Sed. Rate)**0.3

Again, different authors working in different areas find different values for the constant and exponent in this relation.
5. Terrigenous Corg can make up a significant or even major fraction of the total Corg, particularly along continental margins. This can
lead to overestimates of primary productivity and, because terrigenous Corg tends to oxidize more slowly, produces errors in preservation
corrections. Terrigenous Corg can be identified by its distinctive composition (n-alkanes for example) but this is rarely done in studies of
marine Corg.
Foram Indices
The ratio of benthic to contemporaneous planktonic foraminifera in surface sediments increases with increasing productivity. The
quantification of this observation has been frustrated by the overprint of dissolution (which preferentially removes planktonic tests) and the
geographic variability of the ratio.
The relative abundance of different planktonic species has been used to estimate productivity via the transfer function approach. Again,
dissolution tends to preferentially remove the high productivity index species, and productivity tends to be highly correlated with other
environmental variables, such as sea surface temperature. The stability of such correlations over geologic time is unknown.
Benthic foram accumulation rates. Berger and Herguera report that:
Corg flux (at sea floor) = 19.1 + 0.74 (Benthic foram accum. rate); (Figure 28-7)
Figure 28-7. Correlation between flux to the seafloor of Corg and accumulation rate of benthic forams.
Comparison of glacial and interglacial sediments from the Ontong-Java Plateau (Figure 28-8) shows that glacial benthic foram fluxes were
systematically higher, suggesting that primary productivity in this area was 1.5 to 2X its present value during the last glacial.
Figure 28-8. Comparison of benthic foram fluxes as a function of water depth on the Ontong-Java Plateau for recent and last glacial sediments,
showing the effect of inferred higher glacial primary productivity.
Opal
Biogenic opal (hydrated silica) does not display the extreme depth dependent dissolution behavior of calcite or aragonite, so is assumed to record
production more reliably. Opal is difficult to analyze (it is amorphous to X-rays) and chemical techniques have high blanks due to silica from
volcanic ash and clay minerals).
However, a study of the accumulation rates of opal, Corg and CaCO3 in eastern equatorial Pacific sediments (Rea et al., 1991) showed poor
correlations between opal and Corg (Figure 28-9). Archer et al. (1993) explained the observed opal accumulation rate and porewater dissolved
silica concentrations by suggesting that the ease of dissolution of opal is a function of the production rate, resulting in preservation patterns that
are not easily related to primary production.
Figure 28-9. Opal versus Corg for eastern equatorial Pacific sediment spanning the past 1 million years.
Rea et al. (1991; Figure 28-10), also showed that there is a phase lag between the flux of opal and those of Corg and CaCO3, suggesting that the
opal flux is responding to something other than primary productivity alone. This result suggests that findings from Antarctica, where carbonate is
not preserved and paleoceanographic conclusions are heavily dependent on siliceous microfossils and opal accumulation rates, should be viewed
with caution.
Figure 28-10. Phase wheel for the eccentricity (100 ky) band showing equatorial Pacific productivity-related indices.
Organic Marker Molecules

Primary producers, in the course of their normal metabolic activities, generate an enormous number of distinctive organic compounds. A few of
these are relatively resistant to degradation in the water column and sediments, so serve as useful markers of the occurrence and abundance of
particular taxa on the overlying water. They have been used as recorders both of paleotemperatures and paleoproductivity.
Figure 28-11 is an example of the use of the accumulation of C-37 alkenones (produced by haplophytes, such as Coccolithus and Gephyracapsa
spp.) to infer that productivity in the North Atlantic is forced by precessional changes in the energy budget of this region.
Figure 28-11. C-37 alkanones flux at 43 degrees north in the North Atlantic compared to insolation variations at the same latitude due to
precession of the equinoxes and to ice volume as represented by 18-0 in a planktonic foram. FAPI is a fatty acid preservation index that indicates
that the the alkanone flux variations are not due to chemical breakdown.
Cadmium/Calcium
The vertical distribution of cadmium in the oceans mimics that of phosphorus. Hence, the Cd/Ca ratio in foram tests has been used as a proxy for
past P distributions. This will be considered further in the next lecture.
Strontium
Although strontium is a trace element in seawater, it is used by acantharians (radiolarians) to construct SrSO4 tests. These dissolve below the
mixed layer, and are rarely found below a few hundred meters. Thus, like aragonite, strontianite appears to have limited potential in
paleoceanography.
Barium
The profile of barium in seawater closely resembles that of silica. Dymond et al. (1992) used sediment trap data to show a strong correlation
between Corg and Ba fluxes (Figure 28-12).
Figure 28-12. Correlation between Corg and non-detrital Ba in sediment traps at 1700m and less in the Atlantic and Pacific Oceans.
Corg/Ba decreases with depth from a value of about 200 at 200m to less than 50 at depths in excess of 3000m (Figure 28-13). The accumulation
rate data show that Corg decays with depth, whereas Ba is taken up by settling particles. The uptake rate appears to depend on the concentration of
dissolved Ba in deep waters (currently about 50 micromol/kg at 1,700 m in the western North Atlantic to 145 micromol/kg in the California
current).
Figure 28-13. Decrease of Corg/Ba (non-detrital) in sediment traps as a function of water depth.
The empirical relation between Ba flux, new production, dissolved Ba concentration, and water depth is:
F(Ba) = [2056 (Pnew)**0.665 * z**(-0.476+0.000478Ba)] / 0.171Ba**2.22
Complications in using Ba fluxes for reconstructing paleoproductivities include:
Ba content of ancient deep waters is poorly known
Ba can be added from hydrothermal and terrigenous sources near ridge crests and nearshore regions, respectively. The Ba/Al ratio in
continental detritus can be used to correct for the terrigenous Ba, but the hydrothermal input is more difficult to model.
Addition of Ba by xenophyophorans (benthic protozoa that secrete BaSO4 crystals). These organisms are widely distributed, but their
quantitative significance in the Ba cycle is unknown.
Diagenesis. Sulfate reduction in suboxic and anoxic sediments results in the dissolution of barite:
o BaSO4 = Ba+2 + SO4-2
In less reduced sediments, Mn migration can affect the distribution of sorbed Ba.
The preservation of sedimentary Ba depends on the bulk accumulation rate (Figure 28-14).
Figure 28-14 Correlation between the percentage of non-detrital Ba preserved and the bulk accumulation rate of the enclosing sediment.
Dymond et al. (1992) have taken all these factors into account and developed the following steps to estimate new production from Ba
accumulation rates in sediments:
1. Measure mass accumulation rate (MAR) of sediments

2. Measure Ba and Al concentrations in sediment samples
3. Compute Ba(biol)
Ba(biol) = Ba(total) - 0.0075*Al (varies with location)

4. Compute fraction of Ba preserved
Ba (pres %) = 20.9log(MAR) - 21.3 (varies with location)

5. Compute flux of Ba(biol)
Ba(biol; flux) = Ba(biol) / (0.209log(MAR) - 0.213)

6. Estimate dissolved Ba from Ba/Ca in benthic forams
7. Compute new production
P(new) = [Ba(biol; flux)*0.171Ba **2.218 * z**(0.476-0.00478Ba) / 2056]**1.504
About 30% of the Ba(biol) flux to the sea floor is preserved, which is higher than the values for CaCO3, Corg, or opal, but still leaves plenty of
room for perturbations in the paleoceanographic record!
Paytan et al (1996) have used the relation between the burial flux of Ba and productivity to assess productivity in the equatorial Pacific over the
past several hundred thousand years (Figure 28-15). They find that the productivity during glacials was about twice the present value.
Figure 28-15. Accumulation rate of Ba and inferred productivity of the equatorial Pacific between 110 and 140 W during the past 400 ky.
Numbers above Age scale are 18-O stages.
Dissolved Ba. Lea and Spero (1994) used cultured forams to determine that the distribution coefficient Ba/Ca(SW) / Ba/Ca (tests) is 0.147+/0.004 (Figure 28-16), and is independent of temperature (at 22 and 29 C) and salinity (over a 3 ppt range).
Figure 28-16. Ba/Ca in the tests of a cultured planktonic foram (Orbulina universa) as a function of Ba/Ca in the seawater culture medium.
In Ontong-Java plateau cores, however, McCorkle et al. (1995) found that the distribution coefficient decreases with increasing water depth
(Figure 28-17) and appears to correlate with the level of carbonate saturation, suggesting that dissolution may introduce a systematic bias to Ba
concentrations based on Ba/Ca values in foram tests.
Figure 28-17. Variations in the distribution coefficient of Ba/Ca in benthic foram tests relative to seawater as a function of water depth on the
Ontong-Java plateau (left) and carbonate saturation of the seawater (right).
Aluminum
Al in deep-sea sediments is usually viewed as exclusively non-biogenic and refractory (not affected by dissolution). For this reason, element/Al
ratios in average continental rocks are commonly used to correct for the terrigenous contributions of such elements to marine sediments.
However, Murray et al. (1993) found that equatorial Pacific sediments are greatly enriched in Al relative to Ti (an even more refractory nonbiogenic element), compared to other sediments (Figure 28-18).
Figure 28-18. Al/Ti ratios in equatorial Pacific sediments compared to other sediment types and organisms. The excess is attributed to biogenic
scavenging.
They attribute the excess to scavenging of Al from the water column, primarily by CaCO3. They measured the ratio in sediments spanning the
past 600 ky (Figure 28-19) and suggest that productivity during the last glacial of about twice the present value was necessary to produce the
observed variations.
Figure 28-19. Al/Ti ratio in equatorial Pacific sediment spanning the past 600 ky. Variations above the "upper crust" baseline are attributed to
variations in productivity.
Panama Basin Sedimentation

A 1991 study by Pederson et al. highlights some of the progress that is being made in unraveling past records of productivity and the challenges
that remain.
Background: The level of CO2 was lower in the atmosphere in glacial times as documented by studies of gas inclusions within ice cores (about
190 ppm in glacial times, 280 ppm in interglacial times). A variety of explanations have been offered for this, a common theme being that an
increase in productivity plays a role. We can imagine two mechanisms to accomplish this: storage of carbon directly in sediments (productivity
determined) or a change of the internal distribution within the ocean so as to increase the deep water concentration (productivity and circulation
both play a role).
There is conflicting evidence as to the timing of higher production: some studies suggest that primary production increases are in phase with
increased ice volume and lower atmospheric CO2 concentration, while others indicate that the productivity peak lags these two indices.
Pedersen et al. studied cores from the Panama Basin (Figure 28-20):
Figure 28-20. Location of the cores studied by Pederson et al. (1991)

This is an interesting area oceanographically:
There is strong upwelling in the basin (wind driven at coast, upward deflection of equatorial undercurrent by shallow topography in
southwest basin, divergence at the equator)
There is high primary production, ~325-425 mg C m**-2 d**-1
The surface waters contain nitrate and phosphate suggesting that primary production is limited by some other factor.
The present day surface waters are a source of CO2 to the the atmosphere so a critical question is whether these waters ever become a sink
to draw down atmospheric levels.
Features of the Pedersen et al. data set:

The chronostratigraphy is developed from the oxygen isotopic record:
Figure 28-21. Oxygen isotope stratigraphy of Panama Basin cores. Numbers across top of figure are 18-O stages.
We will focus especially on the glacial to interglacial transition 120-130,000 years ago, (termination II, glacial stage 6 to interglacial stage 5)
There is a definite increase in productivity during glacial times as measured by organic carbon concentration and accumulation rate:
Figure 28-22. Variation in Corg concentration and mass accumulation rate (MAR) over the past 500 ky in core AII54-25PC.
However note that there is no clear relationship to atmospheric carbon dioxide (Figure 28-23); contrast the behavior at 2-1 glacial to interglacial
and the 3-2 interglacial to glacial (Figure 28-24):
Figure 28-23. Comparison of Corg flux in Panama Basin cores with CO2 records from the Vostok and Byrd Station ice cores.
Figure 28-24. Comparison of of Corg flux (a measure of productivity) with the CO2 record in the Vostok ice core.
Why was productivity enhanced? Nutrients are not limiting, so upwelling is likely not to be the most critical factor.
Martin and others have suggested iron limitation as the control on equatorial productivity.
The origin of iron in surface waters of the open ocean is atmospheric dust.
The productivity record shows strong correspondence to a dust index
Figure 28-25. Comparison of Corg accumulation rate in Panama basin cores with a dust index in the Vostok Antarctic ice core.
Figure 28-26. Longer term comparison of Panama productivity with dust in Antarctic ice.
What was the CO2 concentration in surface water?
A potential tracer is 13C.
The concentration of dissolved carbon dioxide is reconstructed through the dependence of the carbon isotopic composition of organic
material on the isotopic composition of the CO2 in the water and the concentration of the molecular form CO2(aq), which is a direct
measure of the partial pressure of carbon dioxide in the surface water.
When CO2(aq) increases, 13C decreases (1-2 per mil for the 80 microatm glacial-interglacial transition)
o
13
C=
13
C(CO2) - 0.02 o/oo per ppm P(CO2)
The carbon isotopic composition of carbonate gives the change in carbon isotopic composition of the CO2. There is no evidence for significant
changes in the partial pressure of carbon dioxide in surface Panama Basin water during glacial times.
Figure 28-27. 13C of organic carbon minus the contemporaneous 13C for tests of Neogloboquadrina dutertrei relative to the mass accumulation
rate of Corg in Panama Basin core P7.
Figure 28-28. Comparison of measures Corg accumulation rate in Panama Basin core P7 with surface water CO2 contentent estimated from the 13C
ratio in organic matter (corrected for the global glacial-interglacial 13C change).
Surface waters were always a source, in fact even a stronger source because of the increased gradient and presumably narrower boundary layer
induced by greater wind stress.
We conclude from this study that while productivity in the Panama Basin was enhanced during glacial periods (due, perhaps to increased iron
input via atmospheric dust), the waters in which this occurred did not become sinks for atmospheric CO2 . The locus of any enhanced penetration
of CO2 into the ocean is more likely to have occurred at high latitudes where the solubility of carbon dioxide is high because of generally lower
temperatures
Cd/Ca,
13
C, and GLACIAL NADW
Cadmium/Calcium
The rate of formation of North Atlantic Deep Water (NADW) is a key variable in theories of climate change. Boyle and Keigwin (31) were
among the first to suggest that the cadmium content and the carbon isotopic composition of benthic forams are powerful tracers of deep water
formation. In what way?
Cd is a trace metal in seawater, present at nM levels and below. Typical profiles (Figure NA-1) show a nutrient-like distribution and a strong
correspondence to phosphate.
Figure NA-1, Typical profile of dissolved Cd from the central north Pacific, (31)
The slope of the element-element plot defines a Redfield ratio:
Eq NA-1:
A record of Cd concentration is seawater can be found in forams -- Cd is incorporated in solid CaCO3 via solid solution. Consider the chemical
reaction:
Eq NA-2:
Eq NA-3:
If these two solids form a solid solution within a single phase, then the activity of the solid scales as some measure of the concentration, e.g. the
mole fraction:
Eq NA-4:
The activity of this mixture is not necessarily ideal. The energetics of this non-ideality are expressed with an activity coeffcient, :
Eq NA-5:
Thus:
Eq NA-6:
Separating the measureable quantities from the other variables:
Eq NA-7:
where D is called the distribution coefficient. To a reasonable approximation the activity coefficients of the two divalent cations are the same.
Also since CaCO3 is a major phase its activity coefficient will be approximately one. Finally, the observed value of D is about 2, thus:
Eq NA-8:
Since K is about 2x10-5, the activity coefficient is about 10-5. In energetic terms the free energy change would be RTln or about 14 kcal/mol to
substitute the much smaller (0.17 nm) Cd ion into the Ca (radius 0.23 nm) lattice position.
Given a distribution coefficient of 2, we would expect to find a Cd/Ca ratio in the CaCO3 of a benthic foram deposited in equilibrium with deep
water of concentration 1 nM Cd (typical of the Pacific) of:
Eq NA-9:
In today's ocean, Cd is highly correlated with PO4-3 (figure 28.2. The reason for such a tight correlation is unknown, but appears to be robust over
a wide range of oceanic conditions.
Figure NA-2. Relation of Cd to PO4-3 in waters throughout the world ocean.

In addition, the Cd/Ca ratio in benthic forams correlated well with the estimated ratio in the overlying bottom water (Figure NA-3), yielding an
operational distribution coefficient of 2.8.
Figure NA-3. Correlation between benthic foram (four species) and bottom water Cd/Ca values.
Thus, based on its behaviour in today's ocean, Cd, as recorded in foram Cd/Ca ratios, is a powerful tool to record ocean nutrient concentrations.
Carbon-13
In an earlier lecture, it was shown that non-preformed
13
C and PO4-3 covary according to the relation:
Eq NA-10:
Eq NA-11:
Thus, phosphate and
13
C profiles should be mirror images, as shown in Figure NA-4.
Figure NA-4. PO4-3 and

Thus
13
C profiles from the northwestern Pacific Ocean.
13
C is a second potential proxy for nutrient concentrations that is well preserved in the geologic record.
Present Atlantic-Pacific contrasts
Before assessing core records of earlier distributions, it is worth reviewing the current vertical distributions of nutrients and nutrient-like
elements in the Atlantic and Pacific Oceans (Figure NA-5).
Figure NA-5. Vertical profiles of P, Si, and Ba in the modern North Atlantic and North Pacific Oceans.
Because the deep waters of the North Atlantic are predominantly NADW created only recently from nutrient-depleted surface waters, nutrients
are depleted throughout the water column. Deep waters in the Pacific, in contrast, are about half "aged" NADW, and about half Antarctic water.
These waters enter as bottom waters, but eventually upwell to the surface where primary producers strip them of their nutrients. The "oldest"
waters (in terms of time since last at the surface) are those just below the mixed layer. The nutrient distributions show this age effect, as well as
the relative regeneration rates of the three elements from sinking Corg.
Ice age patterns

Profiles of Cd/Ca in benthic forams down cores from the North Atlantic show a striking change across the boundary between the last glacial and
the current interglacial (Holocene; Figure NA-6). Atlantic values increase by more than 50%, whereas Pacific values decrease slightly (although
barely above the scatter in the data)
Figure NA-6. Profiles of Cd/Ca in benthic forams from the North Atlantic and eastern equatorial Pacific deposited during the past 25 ky.
Similar data from a number of North Atlantic cores from depths between 1.5 and 4.5 km (Figure NA-7) show that the increase is limited to deep
waters. The inferred nutrient content of glacial intermediate waters is the same as or even lower than today's value.
Figure NA-7. Glacial (green) and interglacial (red) Cd/Ca ratios in Atlantic benthic forams as a function of water depth.
13
C profiles from the two oceans are broadly consistent with the Cd/Ca findings. Boyle and Keigwin analyzed a number of cores. Their results
(assuming the global ocean is made up of 3 parts Pacific water and 1 part Atlantic) are shown in the following table.
Summary of Boyle and Keigwin Results
Atlantic
13
Cd/Ca
Pacific
13
Global Average
Cd/Ca
13
Cd/Ca
Interglacial
1.01
0.69
0.16
.174
.37
.148
Glacial
0.25
.110
-0.25
.200
-.13
.178
This summary leads to several conclusions:
in terms of the overall oceanic inventory of phosphate the Cd/Ca ratio suggests that the glacial ocean contained 20%
more phosphate (0.178/0.148=1.20); if the present mean is 2.25 then this is a change in the mean concentration of
0.45 M which would correspond to a change in carbon isotopic composition of -0.39--the observed shift is -0.50
the Redfield connection between phosphate and inorganic carbon would suggest that the glacial ocean also contained
20% more inorganic carbon
the contrast between the Atlantic and Pacific was much smaller in glacial times; a ratio of 2.5 vs 1.8 based on Cd/Ca, a
ocean to ocean diference of 0.98 M vs 0.58 M based on the carbon isotopes
corrected for the overall increase in inventory, the Pacific is about the same in phosphate (~-5% based on Cd/Ca, ~-1%
based on 13C), while the Atlantic is much higher (33%,53%)
The internal distribution at present represents a mixture of NADW (low in Cd, high 13C) and AABW (high Cd, lower 13C). How can the sharp
increase of phosphate in deep water of Atlantic be rationalized? The southward moving mass of NADW is influenced by two sources, a northern
source produced in northern seas with low phosphate content and two southern sources, AABW and AAIW, with very high phosphate contents.
Mixtures of these components satisfy the mass balance:
Eq NA-10:
FsCs+FnCn=FdCd
If Cn is zero, then:
Eq NA-11:
Cd=(Fs/Fd) * Cs
i.e., the increase in the phosphate content in glacial times can be rationalized as a decrease in the intensity of the northern source relative to the
southern source or as an increase in the phosphate content of the southern source.
The present day salinity and temperature characteristics of Pacific deep waters suggest about equal proportions of southern (Fs) and northern (Fn)
source waters now. Taking the Boyle and Keigwin Atlantic core as being representative of this mixing then the observed phosphate content of
1.19 would imply a southern source with 2.38, slightly above the observed 2.25.
Suppose that Cs were unchanged in glacial times. This would imply that Fs/Fd would be 0.70 (or the sources were then in the proportion 2.3:1).
Alternatively, suppose that Fs/Fd were unchanged. This would imply a southern source of 3.33 which is not reasonable as the central Pacific value
in the glacial is only 2.8. Taking that 2.8 as a reasonable upper limit Fs/Fd was at least 0.60 or the sources were in the ratio 1.5:1, still a
substantial change from the present day.
In fact, direct measurements in the southern ocean suggests little glacial-interglacial change in the Cd/Ca ratio for this region. Thus, the PO4-3
proxy data support a glacial Fs/Fd ratio of about 0.7 - a substantial change from the present day.
There seem to be, a problem in the southern ocean, however. Whereas the Cd/Ca (PO4-3) data show no change in the deep waters, the 13C record
(Figure NA-8) suggests that deep waters were substantially lower in 13C (lower than anywhere else in the oceans), with somewhat more
fractionation between surface and deep waters than is observed today, implying significantly higher nutrients and higher productivity in this area.
Because such a finding has been used to explain, at least in part, the glacial decline in atmospheric CO2, it is important to understand the
discrepancy between the Cd/Ca and 13C records.
Figure NA-8. Downcore variations in the

from the Antarctic.
13
C content of calcite benthic (green - Cibicides sp.) and planktonic (red - N. pachyderma) foram tests
Kohfeld et al. (Paleoceanography, v. 15, p. 53; 2000) have recently evaluated the factors that affect the reconstruction of dissolved
values measured in foram calcite. The principal factors are:
Inorganic equilibrium:
13
Caq =
13
CT + 1.0
13
C from
Anthropogenic efects:
Eq NA-12:
13
Eq NA-13:
13
C = 0.0 /oo - (0.012 /oo mol-1kg-1) Canth for C < 35mol.kg-1 and
C = 0.3 /oo - (0.020 /oo mol-1kg-1) Canth for C > 35mol.kg-1
Disequilibrium efects:
Eq NA-14:
Eq NA-15:
13
Cshell =
13
Ceq +
Where
disequ
Eq NA-16:
disequ
diet
carb
size
= (MX)diet + (MX)carb + (MX)T + (MX)size + Btotal

13
and where Mx is the change in
C per unit change of disequilibrium variable, x. The values for M compiled by Kohfeld et al. are:
Diet effect
0.084 /oo per 1 /oo

change in Corg
Carbonate effect
-0.013 /oo per 1

mol/kg CO3-2
Temperature effect
-0.013 /oo per 1C
increase
Size effect
Ignored - use
uniform sized tests
B (total for all

effects)
2.8 (empirical)
Application of Eq NA-14 to core top data from the South Atlantic brings the foram
(Figure NA-9).
13
C data into good agreement with analyses of dissolved CT
Figure NA-9. Effect on uncorrected 13C values from N. pachyderma (purple) of applying corrections for carbonate ion (blue), carbonate ion plus
diet (green) and carbonate ion plus diet plus temperature (red). The final values agree well with modern measurements for the upper ocean (<
200m) corrected for anthropogenic input.
Uncorrected glacial values are lower than surface values (Figure NA-10A) as discussed earlier. However, when corrections based on the best
estimates of glacial diet, carbonate, and temperature are applied, the corrected glacial values are higher than the modern values(Figure NA-10B),
in agreement with the Cd/Ca data. Because the corrections are 4 to 8 times the glacial-interglacial differences, PO4-3 values derived from 13C data
are more uncertain than those based on Cd/Ca ratios.
Figure NA-10. A. 13C data for core top (orange) and last glacial maximum (blue) foram calcite (N. pachyderma) compared to predicted
preindustrial and modern upper water column (< 200m) values for the South Atlantic. B. Last glacial maximum foram 13C values before (blue)
and after (green and yellow) corrections for diet, temperature, and dissolved carbonate changes (two models), compared to upper water column
values for the South Atlantic.
The apparent discrepancy between Cd/Ca and 13C estimates of Southern Ocean glacial nutrient values led to a number of imaginative
explanations involving changes in air-sea exchange and deep water circulation. While these are no longer necessary, there is additional evidence
of changes in deep circulation during glacial times (e.g. Figure NA-11). Because of the lack of biogenic carbonate (and, hence, Cd/Ca or 13C
data) in the deep Pacific, hypotheses about the changes that involve the formation of Pacific deep or bottom water are currently untestable.
Figure NA-11. Estimated bottom water PO4-3 values between 2,400 and 3,800 m for the modern and glacial oceans. Red - greater than 1.9
mol/kg; green - less than 1.5 mol/kg. Note the sparse sample coverage for most of the Pacific.
PLEISTOCENE GLACIAL CYCLES
What do we know about Pleistocene climate?

1. Numerous proxies show strong periodicities with periods of 23, 41, and 100 ky (Figure GC-3)
2. The amplitude of proxy peaks at periods of 23 and 41 ky can be explained by linear forcing by precession and tilt of the earth's axis (Figure
GC-1, 2) as represented by the level of incoming radiation at 65 N.
Figure GC-1. Variations in the amplitude of
18
O and radiation cycles
Figure GC-2. Gain in the precession and frequency bands as a function of time for the past 350-400 ky. The amplitude of incoming radiation at
65 N is used as forcing and the amplitude of the 18O signal in the same bands is used as response.
3. The significance and magnitude of phase relations between tilt (41 ky period) and proxies are ambiguous because both hemispheres are
impacted identically. Because the same lags (due, presumably, to the same processes) also occur in the 23 ky (precession) band in which the
radiation effects on the hemispheres are out of phase, however, the initial effects must originate in the northern hemisphere
Table 29-1. Response times relative to northern and southern forcing
41 ky
20-94
Relative to northern deg
forcing
2-11
ky
20-94
Relative to southern deg
forcing
2-11
ky
23 ky
30-158 deg
2-10 ky
210-338 deg
13-22 ky
4. Phase lags between the spectra of different proxies and inferred radiative forcing yield consistent and physically reasonable sequences of
oceanographic processes for each glacial-interglacial cycle (Figure 22-7).
An examination of 14 proxies (Figure GC-3) shows that 11 have similar phase relations in all three frequency bands. They can be separated into a
group that leads ice volume ( 18O) and a group that lags ice volume.
Figure GC-3. Mean lag times and phases for 14 climatic proxies relative to
Proxies that lead
18
O at the Milankovitch orbital frequencies.
18
O:
13
Antarctic - Pacific
Surface - Deep Pacific
Surface - Deep Atlantic
Cd/Ca (i.e. PO4-3) at 31 S in the South Atlantic
Sea surface temperature at 44 S
13
13
All these proxies are related to the magnitude of the flux of NADW. All would be expected to respond rapidly to events in the Nordic Sea that
increase or decrease the rate of formation of NADW.
Proxies that lag
18
O:
CaCO3 dissolution at intermediate depths in the Caribbean
Flux of dust to the Arabian Sea
Rate of accumulation of Chinese loess
13
Global
C (changes in the mass of the terrestrial biosphere)
Sea surface temperature at 50 N in the North Atlantic
Proportion of NADW at 41 N in the eastern North Atlantic
All these proxies are responses to the build-up of significant northern hemisphere ice masses.
In terms of years, the average lag times for both the lead and lag groups are indistinguishable in the 23 ky and 41 ky bands, suggesting that the
response mechanisms of the earth's climatic system to forcing are the same at these two frequencies. For the 100 ky band, the average lag is the
same, whereas the lead time is about four times as long as for the other two frequencies.
5. Given the striking variations in the flux of NADW, a likely area for initial responses is the Nordic Sea, as defined by Imbrie et al. Because
long proxy records suitable for spectral analysis do not exist for this area, the issue must be assessed in the time domain.
Figure GC-4 shows the relations of a number of processes during Termination 1, the conclusion of the last ice age. These events are well dated by
14
C.
The earliest event is the increase in early June incoming radiation at 65 N.
This is followed by the initial retreat of the margins of the Laurentide ice sheet which appears to be in phase with the increase in 16O (i.e.
decline in 18O) in the world ocean.
This is closely followed by a spike of
northern hemisphere ice sheets
18
O depleted water in the Nordic Sea, which is attributed to the initiation of massive melting of the
This is followed by accelerated retreat of the ice margin and rise in sea level.
Figure GC-4. Evidence that northern hemisphere deglaciation at the end of the last ice age was a response to orbital forcing and that the earliest
responses were in the northern Atlantic (Nordic Sea) region.
The frequency and time domain observations can be simulated by a simple mathematical model. For example, suppose there is a single forcing
(incoming radiation) and a single response ( 18O) variable. Then for a system with a delay time d and a response time T, the phase delay at
frequency f is given by:
= 2 fd+arctan(2 fT)
Since there are analogous phase delays in both the 23 and 41 ky bands, the two equations in two unknowns can be solved to yield d of about 1 ky
and T of about 17 ky. Both are plausible, given the response times of the ocean and a large continental ice mass.
As discussed earlier, Imbrie et al. conclude that four climatic states are involved in the Pleistocene climatic cycles. They modify the simple
mathematical model accordingly (Figure GC-5) to include a single delay time of 1 ky and three response times of 3, 5, and 0.5 ky.
Figure GC-5. Mathematical model of 23 and 41ky responses to radiative forcing at 65 N. Si are linear subsystems (components listed at top of
figure) having either delays (d) or time constants (T).
Figure GC-6. The phase clocks show the goodness of fit between the model of Figure GC-5 and data. Note that R1 is not sampled by the
available time series, but is inferred from the time domain relationships.
After iteration, the fit between models and data is reasonable, particularly if the date of maximum 65 N radiation is shifted from mid-June to
early June or late May.
The 100 ky problem

The simple model developed for the linearly forced precession and tilt frequency bands cannot be applied directly to the eccentricity (100 ky)
band. Some of the issues:
1. The forcing generated by eccentricity variations is too weak (Figure GC-7) to produce the responses observed in the terrestrial climate proxies.
Figure GC-7. Partitioning of incoming 65 N June radiation and 18O (ice volume) record into precession, tilt, and eccentricity bands. Note the
striking discrepancy between the amplitudes of the radiation and ice volume time series in the 100 ky band.
2. The eccentricity record has power at periods of both 100 and 400 ky, whereas the ice volume record only shows 100 ky power. At 100 ky, there
is fair agreement between the two signals (Figure GC-8).
Figure GC-8. Eccentricity showing 100 ky and 400 ky cycles during the past 500 ky. After removing the 400 ky cycle, the amplitudes of the
eccentricity and 100 ky 18O records are quite similar.
3. The time evolution of the spectra of the two parameters over the past 2 my (Figure GC-9) is very different. The amplitude of the 100 ky peak
in the eccentricity spectrum is decreasing whereas the amplitude of the same peak in the ice volume record is increasing.
Figure GC-9. Benthic 18O record for the past 2 my showing the increasing importance of the 100 ky frequency band. In contrast, the amplitude
of the 100 ky eccentricity signal declines over the same period.
4. The sequence of proxy phase leads and lags relative to ice volume is the same for the 100 ky band as for the other two, but the time delay
between the early response group and ice volume ( 18O) is 15 ky rather than about 3 ky (Figure GC-3).
5. Subtraction of the 23 and 41 ky contributions to the total
pass filtered record.
18
O record (Figure GC-10) leaves a residual that is very similar to the 100 ky band-
Figure GC-10. Residual
18
O signal after stripping the linear responses to precession and tilt.
6. Models of icecap development without 100 ky forcing nevertheless develop free oscillations with periods in the 70-130 ky range.
In a recent paper, Shackleton (Science , v. 289, p. 1897, 2000) has compared 18O and CO2 in trapped air in the Vostok ice core with deep-sea 18O
records. By independently tuning the ice and oceanic records to the Milankovitch tilt and precession cycles, he concludes that in the100 ky band,
maximum NH summer radiation, deep water temperature, antarctic ice surface temperature (based on D/H ratio), and atmospheric CO2 are in
phase (with 95% confidence), whereas ice volume lags by about 15 ky (Figure GC-11). This result throws no extra light on the link between
eccentricity and the correlated proxy changes, however.
Figure GC-11. Amplitude, coherence, and phase spectra for deep-sea temperature (A), and ice volume (B) from deep sea sediments and for D/H
(C) and CO2 (D) from the Vostok ice core relative to an eccentricity(e)-tilt(t)-precession(p) record that assigns roughly equal variance to each
component and with the phase of midsummer Northern Hemisphere insolation. The solid line is the amplitude spectrum, the shaded area the
coherence spectrum, and the phase spectrum (with 95% confidence limits) uses the convention that a positive angle is a phase lag (from
Shackleton, op cit.)
From their observations, Imbrie et al. drew the following conclusions

1. Linear forcing by variations in incoming radiation due to eccentricity is not possible.
2. The similarity of the phase lag groupings of proxies in all three frequency bands suggests that the climate system operates similarly at all three
periods, although with a longer delay between early responses and ice volume at 100 ky.
3. The similarity of the ice volume and eccentricity records in the 100 ky band is compatible with a resonance phenomenon in which "pacing" of
the ice volume record by eccentricity phase locks the two records.
The model developed for the 23 and 41 ky responses (Figure GC-5) works for the 100 ky record (Figure GC-12), as long as the ice volume time
constant is increased from 5 ky to 15 ky.
Figure GC-12. Imbrie et al. model for the 100 ky climate cycle compared to the 23/41 ky model (Figure GC-5). Note the longer response time for
S3.
This led Imbrie et al. to suggest that the 100 ky climatic signal results from non-linear amplification of the incoming radiation signal (Figure GC13) once an internal critical threshold is exceeded.
Figure GC-13. Conceptual model for climate where the response is linear until a threshold is crossed, after which the response is
disproportionately amplified.
They suggest that this threshold is set by feedbacks (albedo, atmospheric circulation) that develop when northern hemisphere icecaps exceed a
critical size which interferes with the re-establishment of the Nordic heat pump by radiation increases at the higher frequencies.
The system is paced by eccentricity because of its influence on the amplitude of the incoming radiation signal in the 23 ky band. The key appears
to lie in the extension of the icecaps on to isostatically depressed continental shelves as the ice masses approach maximum size. Once this
happens, even a small sea level rise (due to peak radiation at the precession frequency, for example) results in destabilization of the ice caps with
rapid melting that leads to further destabilization. Such rapid collapses produce the "Terminations" which are such a prominent feature of the late
Pleistocene ice volume record.
The detailed sequence of atmospheric, ice cap, and oceanographic events during the last two terminations (Figure GC-14, 15) is compatible with
this concept.
Figure GC-14. Sequence of events in the Nordic Sea - North Atlantic region during Termination 1 (the transition from the last ice age to the
present interglacial). ND=major Nordic deglaciation event, IE=increased exchange of Atlantic water, T=maximum temperature rise,
ME=maximum exchenge of Atlantic water, DE=decreased exchange of Atlantic water, meltwater = Greenland Sea - global 18O.
Figure GC-15. Sequence of events in the Nordic Sea - North Atlantic region during Termination 2 (the transition from the penultimate ice age to
the last interglacial). Abbreviations as Figure GC-14. Note the age uncertainties (in CO2, ND, IE, purple sections in Nordic Sea forams) relative
to Termination 1, which is well dated by 14C.
Recently, Maureen Raymo and her colleagues (e.g. Ridgwell et al ., 1999, Paleoceanography, v. 14, no. 4, pp.437-440) have suggested that low
65 N summer radiation peaks that occur every fourth or fifth precessional cycle lead to a build-up of "excess" ice on the icecaps, followed by
rapid deglaciation during the next precessional peak. Time series analysis of a mixed 4+5 precessional cycle yields a 100ky spectral peak that
resembles the SPECMAP 18O peak. Because the low summer radiation precession peaks result from modulation by eccentricity, such a 100 ky
spectral peak will be in phase with eccentricity.
Still unexplained is the geologically rapid evolution of the 100 ky spectral peak in 18O records compared to the 23 and particularly the 41 ky
records which have been relatively stable for the past two million years (Figure GC-9).
Hypotheses to explain the increasing importance of the 100 ky cycle focus on modifications to the ice covered regions by a succession of
glaciations. Examples include:
the gradual stripping of the regolith (surface soil and weathered rock layer) which is thought to act as a lubricant that
prevents the build-up of very thick ice, and
downcutting of the continental shelf by ice streams so that the toes of the streams are well below modern sea level and
are, therefore, vulnerable to collapse as a result of small rises in sea level due to radiation increases at 65 N resulting
from Milankovitch forcing.
DANSGAARD-OESCHGER CYCLES AND HEINRICH LAYERS
The earlier part of this section of the course has emphasized the Milankovitch portion of the frequency spectrum because it is a region where
there is well defined forcing that appears to generate responses in many component of the climate system. Likewise, although we have not
emphasized high frequency climate change, El Nino and higher frequency signals are a current focus of paleoceanographic research. In between
is a frequency band (corresponding roughly to periods of 10 to 20,000 years) where nothing is known about possible forcing mechanisms, yet
where there is considerable evidence of climate variability.
The first time series to provide a continuous record of this variability were 18O profiles from Greenland ice cores. Figure DO-1 shows such an
oxygen isotope record spanning about 200 ky in which the current interglacial displays little variability, whereas the preceding glacial record is
very "spiky".
Figure DO-1. Variation of 18O with depth in a Greenland ice core. An approximate age scale is given in the middle of the figure. The record
below about 2,750 m is suspect due to internal deformation of the ice.
Figure DO-2 shows the 18O record for the last 100 ky in a second, better dated, Greenland core, this time with the depth axis converted to time.
The high frequency spikes (Dansgaard-Oeschger (D-O) cycles) between the current and prior interglacials have amplitudes of 4-5 per mil, which
implies a warming of the air temperature of 6-7 degrees C, or roughly half the glacial-interglacial range. The onset of these brief warming
periods was very rapid (decades or less) in each case, whereas cooling was more gradual. The pattern is reminiscent of glacial-interglacial 18O
patterns in deep sea cores.
Figure DO-2. 18O versus age in a recent well-dated ice core from Greenland. Note the amplitude of the D-O cycles relative to the glacialinterglacial range.
More recent work (Figure DO-3) has shown that these D-O events coincide with lower dust concentrations in the ice, and with increases of
about 50 microatm in the CO2 content of bubbles of atmospheric gas trapped in the ice.
Figure DO-3. Variations in dust content and the CO2 concentration in gas bubbles in Greenland ice relative to the D-O cycles.
Both the magnitude and rapidity of onset of the D-O cycles are significant on human time scales, but the forcing mechanism is unknown.
Speculation centers on changes in either the global water vapor budget or the oceanic heat pumps of the North Atlantic, but neither hypothesis
has yet been tested rigorously.
Shortly after the ice core records were recognized, sediment layers enriched in ice-rafted debris (called Heinrich layers after their discoverer)
were identified in cores in a band across the North Atlantic (figure DO-4). The distribution pattern and types of rock fragments (which include
very old rocks and carbonates) suggest that the primary source is the Hudson Bay area of Canada. Subsequent more detailed studies have also
identified a component from Iceland and another from the St. Lawrence.
Figure DO-4. Locations of sediment cores containing Heinrich layers and inferred path of the icebergs that produced them.
The following table summarizes the distinctive characteristics of the Heinrich layers. They do not simply represent an increase in the background
flux of glacial ice-rafted debris, but have characteristics that point to both a different source and contemporaneous changes in the oceanographic
conditions in the area of deposition.
Property
Ambien
t
glacial
sedime
nt
Heinrich layers
Reason for difference
FORAMINIFERA
SHELLS (>150
microns)
Abundance
High
Low
Low productivity and/or high sedimentation rate
18
High
Low
Melt water
Species
Variable
All N. Pachyderma (l)
Cold ocean
~80%
~80%
---
O / 16O
LITHIC
FRAGMENTS
(>150 microns)
Quartz
Detrital CaCO3
<2%
~20%
Icebergs from eastern Canada
Smectite
present
Smectite absent
No debris from Iceland
FINE-GRAINED
MATERIAL
Clays
K - Ar age of clays 0.44 billion 0.90 billion
Icebergs from eastern Canada
K - Ar age of
individual
amphibole grains
Icebergs from Churchill province northeastern Canada
Large range 1.7 billion
Pb isotopes of
individual feldspar Large range 2.7 billion crust, 1.7 billion metamorphics
grains
Icebergs from Churchill province northeaastern Canada
Table DO-1. Comparison of Heinrich layers with the intervening ambient glacial sediment in North Atlantic cores.
Careful dating and correlation of the ice core and marine records (Figure DO-5) shows that the Heinrich events correlate with the cool intervals
immediately before D-O warm spikes (note the reversed 18O scale relative to the normal convention).
Figure DO-5. Correlation of Heinrich and interspersed layers of ice rafted debris with D-O cycles in Greenland ice cores.
Figure DO-5 shows that smaller peaks in ice-rafted debris occur between the four major Heinrich layers, suggesting that the subarctic climate
fluctuated with a period in the 2 ky range.
Similar fluctuations have been subsequently been identified in numerous other records as diverse as vegetation (pollen) indices in Florida Lakes
and laminated sediments in the Santa Barbara Basin off southern California (Figure DO-6), the Cariaco Trench off Venezuela (Figure DO-7),
and the upwelling region off southwest Africa (Figure DO-8).
Figure DO-6. Summer (red) and winter (blue) temperature fluctuations in Santa Barbara Basin (based on foram assemblage transfer functions)
compared to D-O cycles (yellow numbered intervals). Horizontal lines at 10C and 15C are coolest and warmest modern temperatures observed
at 25 m.
Figure DO-7. Correlation of fluctuations in the color of Cariaco Trench sediments with DO cycles in the GISP II ice core. Dark sediment bands
are laminated and indicative of low bottom oxygen, high surface productivity, and high runoff from South America.
Figure DO-8. Correlation of fluctuations in the abundance of left coiling N. pachyderma beneath the Benguela current, west of South Africa
(indicative of incursions of water from south of the subtropical convergence) with DO cycles in the GISP II ice core.
The limited geographic coverage of cores showing DO-type fluctuations (Figure DO-9) does not define the global pattern of such fluctuations, or
seriously constrain theories as to their origin.
Figure DO-9, Sites with paleoclimatic records spanning the last 40-80 ky with resolution better than 200 years.
A recent paper by Shackleton and Hall (Paleoceanography v. 15, p. 565, 2000) does suggest that the effect is primarily Northern Hemisphere and
upper water column; Antarctic air temperature (from D/H ratios) and deep-water ice volume/temperature time series do not show D-O
fluctuations (Figure DO-10).
Figure DO-10. Fluctuations in the oxygen isotopic composition of planktonic and benthic forams from a core off the Iberian peninsula compared
to the oxygen isotope record of the GRIP (greenland) and the D/H ratio in the Vostok (Antarctica) ice cores.
The frequency spectra of these records show barely significant peaks corresponding to periods of a few hundred to a few thousand years (e.g
Figure DO-11). Wunch suggests (Paleoceanography v.15, p. 417, 2000), however, that the 1,460 peak that is prominent in many of the records
results form aliasing of the seasonal cycle and, in any case, that the power in peaks is a small fraction of the total power under the spectral curves.
Figure DO-11. Examples of spectra of proxies that show D-O type fluctuations. Numbers at top show some of the peak periods in years (after van
Kreveld et al., Paleoceanography v. 15, p. 425, 2000).
There is little question that these are global signals (both the temperature and CO2 changes exceed reasonable local effects), but whether the
cause is internal forcing (natural oscillations in the icecaps, hydrological cycle, or oceans, for example) or external forcing (by variations in the
solar constant, for example) is still an open question.
Sub-Milankovitch cycles, particularly those associated with rapid temperature changes, are of interest to climatologists because they provide a
lower limit on the rapidity with which climate can change, and because they can be combined with models of the impact of greenhouse gases to
make more plausible empirical projections of future climate change (e.g. Figure DO-12).
Figure DO-12. Effect of including high frequency temperature variations inferred from Camp Century ice core record on projections of
greenhouse gas warming of the earth.
Particle Size Distributions
The physical behavior of particles is controlled principally by their size and density and to a lesser extent their shape. Because
the density of many common minerals is similar, of order 2600-2700 m 3kg-1, sediment and suspended particulates are first
generaly characterized by a particle size distribution, expressing the number or mass of particles as a function of their size.
We begin with some common nomenclature. Particle size was first characterized by sieving material. As apparatus became standardized, particle
size began to be expressed on a scale due to Wentworth (1922) called the phi-scale:
Class Name
Size
Range
(mm)
Size
Range
Units
( m)
Boulders Very
Large
40962048
-11
Large
20481024
-10
Medium
1024512
-9
Small
512256
-8
Large
256128
-7
Cobbles
Gravel
Sand
Small
128-64
-6
Very
Coarse
64-32
-5
Coarse
32-16
-4
Medium
16-8
-3
Fine
8-4
-2
Very
Fine
4-2
-1
Very
Coarse
2-1
20001000
Coarse
1-0.5
1000500
Medium
0.50.25
500250
Silt
Clay
Fine
250125
Very
Fine
125-62
Coarse
62-31
Medium
31-16
Fine
16-8
Very
Fine
8-4
Coarse
4-2
Medium
2-1
10
1-0.5
11
Fine
Very
Fine
0.50.24
12
A standard sieve for a particular phi-value would collect all material larger than the smallest size in the class, e.g., a =-4 sieve would collect
material larger than 16 mm.
There are now many techniques for determining particle size. Classical techniques include:
direct observation, aided by microscope
sieving through standard sieves (generally applied to very fine sands and above, >=0.6 mm)
sedimentation rate through a fluid (generally applied to silts and clays, <=0.6 mm)
Instrumental methods are rapidly replacing manual techniques. These include:
laser difraction
acoustic difraction
field flow fractionation
microscopy coupled with computerized image analysis
Results of these analyses are generally expressed either as a histogram showing weight or number in size class, or
equivalently as a cumulative frequency versus size:
Figure 32-1
There are a number of statistical descriptors to quickly summarize the character of these distributions. Two keys ones are the
width of the distribution and the skewness of the distribution. When the distribution is narrow the sediment is said to be wellsorted while when it is broad sediment is said to be poorly-sorted. Similarly the distribution may be skewed toward finer or
coarse size fractions or may be symmetrical, all possible indicators of the origin of the sediment.
In the deep-sea, particles form a benthic boundary layer
Figure 32-2, (35)
The parameter c is the light attenuation which is related to the particle density. Note the approximately 50 m thick layer of high particle
concentration grading to background conditions approximately 100 m off the seafloor. Size distributions change considerably with small particles
dominating the bottom mixed layer and large particles dominating the background ocean. What controls the distribution of sizes?
Sedimentation of Particles
Stokes developed a analysis of the settling of spherical particles involving the balance between the net gravity force acting on the particle and the
viscous drag exerted by the fluid.
The viscous drag force will depend on the size, r, velocity, w, and viscosity, , of the fluid medium:
Eq 32-2:
Analyzing dimensionally:
Eq 32-2:
By inspection a=b=c=1. Without developing the hydrodynamics, the viscous drag force acting on a particle is:
Eq 32-3:
The net gravitational force (gravity less buoyancy) will be depend on the volume, density, , and the acceleration of gravity, g:
Eq 32-4:
At terminal velocity, these two forces are balanced:
Eq 32-5:
Solving for w:
Eq 32-6:
Typically
is 1.5 g cm and for seawater is .01 g-cm sec so:
1000 meter
descent
1 mm
87 cm s-1
1.15 x 103 s ~ 19
minutes
0.1 mm
0.87 cm s-1
1.15 x 105 s ~ 31
hours
0.01 mm
0.0087 cm s-1
1.15 x 107 s ~ 1/3

year
0.001 mm = 1 m
0.000087 cm s-1
1.15 x 109 s ~ 35
years
If this were all that were happening, once in suspension small particles will be transported long distances and large ones small distances.
Settling of Aggregates
While discrete particles generally exhibit Stokes settling behavior, particles have attractive forces that cause them to form aggregates or flocs.
These aggregates may be held together by relative weak forces (e.g., electrostatic forces) or by coatings of organic material. These aggregates
will often have lower density than the constituent grains, due both to the inclusion of water in the interstices of the aggregates and to the lower
density of associated organic material. Because many of the forces that lead to aggregation are weak, turbulence can cause disaggregation and the
size distribution of flocs will represent some kind of balance between the rates of aggregation and disaggregation.
Figure 32-3 shows an example contrasting the size distribution of the discrete particles making up a collection of natural aggregates (as
determined by Coulter sizing) and the size distribution of the aggregates themselves (as determined by image analysis of photographic images).
The constituent grains are poorly sorted (spanning greater than 4 phi units) and a factor of 100-1000 smaller than the aggregates.
Figure 32-3
These aggregates settle much more slowly than the Stokes settling velocity:
Figure 32-4
The Stokes settling velocity of a 1 mm grain is 870 mm/s. For these aggregates the settling velocity at 1 mm size is ~1.8 mm/s, ~480x lower. This
could be expressed as an excess density; instead of 1600 kg m-3 about 3.3 kg m-3.
Particle Aggregation
In the last lecture we saw that small particles settle very slowly such that the time scale associated with settling of 1m
particles is measured in tens of years. Particles however have a tendency to aggregate, creating larger particles which settle
more rapidly. To aggregate, a collision must occur between the particles and when that collision occurs the particles must
adhere to one another. We first consider why particles adhere and then consider the question of what processes lead to
particle collisions and how frequently do they occur?
Particle-Particle Interactions
All particles have a van der Waals attraction brought about by a variety of forces, principal among these the interaction of dipoles of neighboring
atoms and molecules. This attractive force is modelled (the Lennard-Jones model) with a r decay with separation:
Eq 33-2:
where is a scaling factor for the energy of interaction and is a scaling length for the molecular size. To a good approximation the force exerted
by individual atoms or molecules is additive so that the force exerted by a particle can be obtained by integration and the force exerted by a
spherical particle can be described by a constant Q:
Eq 33-3:
where v is the molecular volume. Q ranges from a few kT for strongly polar molecules to of order 0.1 kT for crystalline silicates and oxides.
For interaction of two spherical particles of radius r and r , with centers separated a distance d the energy of interaction is:
Eq 33-4:
When the particles are much larger than their distance of separation, the energy of interaction decays by approximately the square of the distance
between particle edges at closest approach, as marked in red:
Figure 33-1a
At very close approach the Born repulsion of adjacent electron clouds (marked in blue) begins to play a role, so in the absence of any other forces
there is a strong energy minimum with distances of approach of order 1 nm.
Figure 33-1b
There is an additional important term, the electrostatic interaction of particles. Particles in water generally have a surface charge, acquired by
ionization of atoms at the surface of the particle. Here Si represents any cation at the surface of a crystalline oxide or aluminosilicate while the
carboxyl group is representative of a functional group that would be found at the surface of particulate organic material (POM). These surfaces
exhibit acid-base behavior with surfaces becoming more negatively charged with increasing pH.
Eq 33-5:
This acid-base behavior leads to positive surface charge at low pH and negative surface charge at intermediate and high pH for a variety of
materials. At moderate pH most surfaces are negatively charged.
Figure 33-2
How do these charged surfaces interact? First we consider the nature of an ionic fluid near a charged surface. The simplest model aligns an
excess of positive ions along a negatively charged surface, generating a potential across a thin layer (the capacitor or Helmholtz model of a
surface). However experiment shows this model to be too simple. A more realistic model is that proposed by Guoy and independently by
Chapman in the early 1910s. This model holds that ions redistribute themselves in solution to accomodate the gradient of potential brought about
by a charged surface
Figure 33-3
At all points the chemical potential:
Eq 33-6:
is a constant. For positively charged ions:
Eq 33-7:
and for negatively charged ions:
Eq 33-8:
The charge density at any point is:
Eq 33-9:
Potential and charge q are related by the Poisson equation:
Eq 33-10:
where the product in the denominator on the right hand side is related to the dielectric characteristics of the fluid. This differential equation can
be solved subject to appropriate boundary conditions to find that (for low potentials):
Eq 33-11:
where
Eq 33-12:
and I is the ionic strength. This characteristic length then determines how the fields of two adjacent particles interact. For aqueous solutions:
Eq 33-13:
so that in seawater (ionic strength 0.7) the scale length is 0.3 nm and in river water with ionic strength of 10
the scale length is 9 nm.
The electrostatic repulsion (marked in cyan) and the van der Waals attraction combine to give the overall energy of interaction:
Figure 33-1c
As a function of ionic strength, the overall interaction undergoes a sharp transition from repulsive to attractive:
Figure 33-4
This is manifested in the behavior of particles, e.g., as shown in these data from experiments in which suspensions of the clay mineral kaolinite
were monitored at different ionic strengths.
Figure 33-5. A, ionic strength .036; B, ionic strength .087; C, ionic strength .343. From (37)
We have emphasized the tradeoff of repulsive electrostatic forces with van der Waals attraction. There are other way of promoting adhesion
between particles. In particular coatings of biopolymers, in addition to having electrostatic interactions, can bridge between particles by forming
chemical bonds with a range of energies (hydrogen bonding, covalent bonding).
Particle Collisions
Particles move simply because of their thermal energy, i.e., they undergo Brownian motion. Consider a disperse suspension of one size of
particles that have some tendency to aggregate. The rate of change of the number density of particles will be described by:
Eq 33-7:
where
Eq 33-8:
In this expression n is the volume density of particles, t is time, is the fraction of collisions leading to aggregation (=1 all adhere, =0 none
adhere), D is the thermal diffusion coefficient of the particles (which is kT/(6 r)), and r is the particle radius. Simplifying:
Eq 33-9:
Eq 33-10:
At 25C, the value of k is 5.5 x 10
cm s .
Suppose is one and consider as an example 1 m particles in a 1 g/l suspension. Each particle has a mass of about 10 g so the volume density
of particles is about 10 cm . What is the time to halve the density of particles? Solving equation 24-5 for the initial condition n(t=0)=n :
Eq 33-11:
At this particle density, the characteristic time is about 30 minutes. If the particles were a factor of ten larger (10 m), then the characteristic time
would increase to about 21 days. If we compare to the Stokes settling times, we see that the effect is only important for the smallest particles.
The discussion can be made more general by introducing the concept of a coagulation kernel which unifies the treatment of different physical
mechanisms for bringing particles together:
Eq 33-12:
In these expressions, f is the frequency of collision of particles. The subscripts i and j denote size classes of particles of volume v. The time rate
of change of particle concentration in the k size class is given by two terms. The first represents the particular combinations of smaller particles
that can aggregate to form a particle exactly in size class k. The second represents the disappearance of particles from the k size class due to
collision and aggregation with all size classes. The frequency of collision is the product of a coagulation kernel and the particle density in two
size classes, e.g., for Brownian motion:
Eq 33-13:
where the kernel is given by:
Eq 33-14:
We can compare this expression to our earlier discussion of Brownian motion if we consider a monodisperse (single particle size) suspension:
Eq 33-15:
(For a monodisperse suspension there are no smaller particles from which to make particles of the initial size so that the first term in equation 247 is zero.)
A useful way of visualizing the magnitude of the kernel is to consider it size for a particular particle size as a function of the size of the other
particle:
Figure 33-2, (35)
This figure shows (among other things) the Brownian motion kernel for three different values of d . There is a minimum at the point d =d with a
1/d dependence at smaller d and a d dependence at larger d .
Brownian motion is not the only mechanism which can bring particles together. In the presence of shear, particles at one level are traveling at a
velocity different from those on another level leading to collisions.
The coagulation kernel due to shear is given by:
Eq 33-16:
If the flow is laminar:
Eq 33-17:
If the flow is turbulent:
Eq 33-18:
where is the turbulent dissipation rate and the kinematic viscosity.
We can compare the relative sizes of the kernels for Brownian motion and shear:
Eq 33-19:
In the lower 100 m of the water column, ~0.1 s , so for a monodisperse suspension the boundary between dominance of the two mechanisms
falls at a particle size given by:
Eq 33-20:
If the suspension is monodisperse, then Brownian motion is more important for particles smaller than 3.5 m and shear more important is
bringing about collisions of particles larger than 3.5 m.
Figure 33-2 also shows the behavior of the kernel for collisions due to shear for two different values of the turbulent dissipation rate. When d is
small, the value of the kernel approaches a constant value controlled by d : K ~ d /6. When d is large, the value of the kernel follows the cubic
dependence on d : K ~ d /6.
Particles also come together by differential settling with larger particles sweeping out smaller particles.
The kernel for this process is given by:
Eq 33-21:
where w is the settling velocity (denoted by the black arrows), E is the collision cross-section, and E is the enhancement of the collision cross
section by Brownian motion (denoted by the red arrows). E and E are complicated functions of the sizes of the particles and thus so is the
kernel (Figure 33-2).
The relative sizes of the coagulation kernels summarized in Figure 33-2 indicates that different mechanisms are important for different sized
particles. For very small particles, Brownian motion and shear are important mechanisms. For intermediate sized particles, all three mechanisms
can play a role. For the large particles, differential settling is dominant except for interactions with much larger and smaller particles.
It is important to keep in mind that Figure 33-2 is somewhat misleading in that it contains no information on particle size distribution, while the
frequency of collisions depend both on the magnitude of the particle densities and the kernel. n(d) is strongly size dependent:
Figure 33-3, (35)
By combining the information in these two figures, we can estimate the residence time of particles in various size classes and compare these to
the residence time due to Stokes settling through the lower 100 meters of the water column (these residence times are in years):
Coagulation through Brownian motion moves particles to larger size classes where they are removed by Stokes settling. Note that there is likely
to be a maximum in the residence time for particles somewhere between 0.5 and 5 m.
While interparticle forces acting in the ocean are generally attractive, experimental studies made on natural particle
populations suggest that the value of is about 0.1. Some factors include:
shape. In order for the van der Waals attraction to be strong, the particles need to be aligned. A rough surface
geometry can prevent this.
shear. Shear at the surface of a aggregated particle can dislodge individual particles. While progress has been made in
representing this process with a disaggregation kernel, the net efect is often included within an efective value for
GRKalpha|.
For more information on this topic see references (35), (36) and (37).
Marine Benthic Boundary Layers

Shear Stress
Physical and Geological Perspective
Sediment transport in the marine environment is driven by fluid motion. Some important transport mechanisms are depicted in Figure 1 in the
context of the continental shelf:
Wind-driven alongshore (driving upwelling and downwelling) and across-shore
Surface gravity waves
Tidal currents (not depicted in figure)
Internal waves
Infragravity oscillations
Buoyant Plumes (with both positive and negative buoyancy depending on sediment concentration)
Wave driven surf zone processes (e.g., waves breaking, rip currents)
Figure 34-1
Different regimes exist across the continental shelf:
Figure 34-2
Internal waves can act in any of these regimes, though their importance increases in the outer zones. The important processes in deeper water are
the same as on the outer shelf, i.e., dominated by geostrophic flow and secondarily internal waves.
Context for Study of Boundary Layers
A boundary layer exhibits a velocity profile connecting a background oceanographic flow and a no-slip, zero velocity condition at a solid
boundary.
Figure 34-3
Transfer of momentum toward the boundary imposes a shear stress; when it is sufficient (i.e., exceeds a critical value), it can cause transport of
sediment either as bedload or suspended load.
We begin by considering how a moving fluid interacts with a solid boundary and the nature of the velocity profile.
Shear Stress
Stress, denoted , is a force applied to a plane surface per unit area. Dimensionally:
Eq 34-1
Consider a volume element. Each of the three surfaces can have three forces acting on it, so there are nine components in all. They are denoted
where the i subscript describes the surface (e.g., z would indicate the surface perpendicular to the z-axis) and the j subscript describes the
direction of the force (e.g., x would indicate the force acting in the direction of the x-axis).
Figure 34-4
The three components acting perpendicular to the surfaces represent the pressure force acting on the volume:
Eq 34-2
Of the remaining six components, we are interested in the two acting in the x and y directions on the plane perpendicular to the z-axis :
Eq 34-3
At the seabed these stresses are denoted
and
For simplicity we will consider flow in the x-direction that is steady (

the stress at the boundary
as
) and uniform (
) so henceforth we denote
as and
Consider the case of laminar flow (no turbulence) between a fixed surface and a plate of area A, separated by a height h, and moving at a steady
velocity u (such that the applied force is opposed equally by the drag of the solid surface).
Figure 34-5
In this situation the fluid velocity is proportional to the plate separation and the force per unit area:
Eq 34-4:
For laminar flow (of a Newtonian fluid, see below), the velocity profile would be linear:
Eq 34-5:
The gradient of velocity which appears in this equation is called the shear. Since the stress is the force per unit area, we use Eq 34-4 and 34-5 to
write:
Eq 34-6:
The proportionality constant is the absolute viscosity and is a measure of the resistance to deformation of a fluid. Dimensionally the stress has
units of force per unit area (g cm/s )/(cm ) and the velocity gradient units of (s ) so the viscosity has units of g/(cm s). One g/(cm s) is one poise.
In seawater the viscosity depends on temperature and salinity:
Fresh Water
Seawater, S=35 o/oo
Tempera Dens
ture
ity
Absol Kinem Dens Absol Kinem

ute
atic
ity
ute
atic
Viscos Viscos
Viscos Viscosi
ity
ity
ity
ty
g/cm centip
3
oise
cm2/s
g/cm centip
3
oise
(x 10-2)
cm2/s
(x 10-2)
1.00
00
1.52
1.52
1.02
8
1.61
1.57
10
.
9997
1.31
1.31
1.02
7
1.39
1.35
15
.
9991
1.14
1.14
1.02
6
1.22
1.19
20
1.005
1.00
1.02
1.07
1.05
9982
30
.
9956
0.801
0.804
1.02
2
0.87
0.85
An alternative way of thinking about the stress due to shear is as a transfer of momentum in a direction perpendicular to the surface on which the
stress is acting. Multiplying the numerator and denominator by :
Eq 34-7
The stress is the product of the gradient of fluid momentum and a constant , the kinematic viscosity.
The discussion to this point pertains to a laminar flow and is for a Newtonian fluid. We say a fluid is Newtonian when the viscosity as defined by
the proportionality in Eq 34-6 is a constant.
Figure 34-6
Other materials may resist deformation until a critical stress is reached, for example a so-called Bingham plastic which does not deform until a
yield strength is exceeded after which it deforms linearly with increasing stress. Pseudo-plastic materials follow a path intermediate to a
Newtonian fluid and a Bingham plastic. Water and dilute suspensions exhibit Newtonian behavior, while dense clay suspensions (say > 10 g/l)
exhibit pseudo-plastic behavior (flocs are broken down to the constituent units before deformation takes place).
Turbulent Shear Stress

Friction Velocity
At the boundary, fluid velocity slows to zero. By transport of momentum, velocity in the interior must match this condition through some
adjustment mechanism that will determine the thickness of the boundary layer. Typically friction is thought to be the adjustment mechanism.
Recalling that the viscous stress is:
Eq 1
We can divide both sides by the density to yield:
Eq 2
The dimensions of the left hand side are
Eq 3
or units of velocity squared. We define a term called the friction velocity or shear velocity:
Eq 4
Thus
Eq 5
Integrating
Eq 6
Since u=0 at z=0, C=0 and:
Eq 7
Velocity Fluctuations
A record of velocity as a function of time can be characterized as fluctuations about some mean value:
Eq 8
Turbulence will give rise to velocity fluctuations in both the horizontal (velocity u) and vertical (velocity w) directions characterized by u' and w'.
The associated momentum fluctuation is called the turbulent stress or Reynolds stress. Analogous to the viscous case, we can
express the turbulent stress as being the product of eddy viscosity, Av and the shear:
Eq 9
Thus the friction velocity can be written:
Eq 10
Prandtl hypothesized that turbulent fluctuations should act over some correlation scale l so that
Eq 11
Thus
Eq 12
von Karman further hypothesized that the correlation scale should be proportional to the distance from the boundary:
Eq 13
where kappa is von Karman's constant, the turbulent momentum exchange coefficient. By experiment kappa has been found to be 0.41.
Eq 14
Integrating:
Eq 15
Let u be 0 at z=z0. Then
Eq 16
This is the von Karman-Prandtl equation, the Law of the Wall.
Structure of the Boundary Layer
We can now develop the structure of the boundary layer. Very near to the boundary where viscous forces dominate there may be a viscous sublayer where equation 1 applies. Whether this viscous sub-layer exists depends on whether the boundary is smooth or rough. If it is rough it will
generate turbulence at the boundary so that turbulent forces become more important than viscous forces.
The roughness is characterized by a dimensionless number
Eq 17
where ks is a length scale of the roughness elements (which may be due to grain roughness of the surface, ripple patterns, or fluid stratification)
When R* is less than 5, the flow is said to be hydrodynamically smooth (HSF), when R* is greater than 70 the flow is said to be
hydrodynamically rough (HRF), when between 5 and 70 the flow is transitional.
The value of z0 for HSF and HRF have been established by experiment. For HSF:
Eq 18
For HRF:
Eq 19
To summarize:
The most common way in which u* and z0 are measured is by determing the velocity profile above a boundary. The profile is fit to the Law of the
Wall and the parameters from the fit used to decide whether the flow is HSF or HRF.
Eq 20
Threshold of Motion
Context
Grains forming the boundary between a fluid and a sediment possess a finite weight and finite coefficient of friction. When the applied shear
stress is low they are not brought into motion. As applied shear stress is increased, a critical shear stress is reached at which grains will begin to
move. The value of the critical stress will depend primarily on the size and density of the particles and secondarily on their shape and packing
and the cohesive forces acting between particles.
One the critical stress is just exceeded, particles will advance in the direction of flow due to irregular jumps or less commonly rolls. This mode of
tranport is termed the bedload and conceptually can be thought of as being deterministic, that is the behavior of a particle once in motion is
dominanted by the gravity force. As the stress is further increased, particles will also begin to be suspended in solution and subject to turbulent
forces. This mode of transport is termed the suspended load. Due to these two modes of transport there will be a flux of material across a plane
perpendicular to the flow. Our ultimate goal is to determine this mass flux by integrating the product of the velocity profile and concentration
profile.
The Critical Stress
The motion of sediment can be parameterized in a number of ways. The oldest of these is due to Hjulstrom who summarized observational data
in terms of fluid velocity and grain size:
Figure 36-1
There are a number of variants of the Hjulstrom diagram, using grain diameter as one parameter and some measure of the stress as the other (via
the quadratic stress law: u, u100 or stress itself: ustar):
Figure 36-2
In several of these figures there is a envelope of values for small particles, contrasting unconsolidated and consolidated/cohesive sediment. This
reflects the importance of interparticle forces because of the higher ratio of surface area to volume.
Shields first suggested a more general scaling, based on an analysis of the forces acting on particles. His analysis suggested an organization of the
data based on two factors.
The first is the entrainment factor, , which scales the boundary stress to the density contrast between the particles and the fluid, - , and
the particle size, D and thus compares the applied stress to the gravitational force acting on an individual particle:
Eq 36-2:
g is the gravitational acceleration.
The other is a Reynolds number, relevant to an individual particle, comparing turbulent and viscous forces:
Eq 36-3:
For non-cohesive sediments, under a wide range of experimental conditions, a boundary is found in this parameter space (entrainment factor,
Reynolds number) between no motion of the sediment and active erosion of the sediment:
Figure 36-3
As an example consider the situation for Re=10 where (reading from Figure 36-3) the critical is 3 x 10 :
Eq 36-4:
Solving each of the two equations for D:
Eq 36-5:
Equating these expressions for D:
Eq 36-6:
and solving for u*, we find:
Eq 36-7:
For grains of density 2.7 g/cm in water of viscosity 10
g cm s
(1 cp):
Eq 36-8:
That is 1.6 cm/s is the scale of the rms velocity fluctuations necessary to just begin to transport a 0.6 mm grain.
Sediment Transport
Sediment transport is expressed in terms of a flux of material through a plane surface perpendicular to the flow (dimensions M L-2 T-1):
FIGURE
Restricting attention on the two dimensional case, the transport can be integrated vertically:
The quantity q has the dimensions M L-1 T-1 so that one would obtain the transport within a channel in terms of mass per time by multiplying by
the channel width.
Ultimately the interest in evaluating sediment transport is to assess the effects of transport on erosion or deposition of the bed:
FIGURE
Let the porosity, p, be the volume percent of the bed that is fluid. In the absence of a vertical sediment flux the depth of the bed relative to some
datum will be related to the downstream change in sediment transport:
When dq/dx is positive, material is progressively being added to the transported load and h at a point increases with time (i.e., material is eroded).
(Note: this equation needs to be fixed should be 1/rho)
Bedload and suspended load are distinguished by the dominant forces acting on the particles being transported. Bedload transport is dominated
by a balance of applied stress acting to put particles into motion and gravitational forces acting to keep them in place. Suspended load is
dominated by a balance of turbulent forces acting to keep particles in suspension and gravitational settling. Besides this separation by physical
mechanisms involved, the two modes of transport are also distinguished by being measured with quite different techniques.
Bedload Transport
The physics of bedload transport involve many complicated mechanisms and feedbacks and so becomes difficult to express in simple terms.
Many equations for bedload include empirically derived terms which may or may not be of general utility. Historically three approaches have
been taken in developing expressions for the bedload:
shear stress relationships (DuBoys approach)
force balances (Einstein, Bagnold approaches)
discharge relationships (Schoklitsch approach)
The last of these three is relevant to analysis of channels and is not considered here. The first two will ultimately yield expressions that may be
comparable, but the starting point is quite different.
DuBoys approach posits that the sediment transport is a function of the boundary stress. As one example, assume that this function can be
represented by the power series:
When there is no stress, there is no transport and so k1 must be zero. Just at the critical condition q is zero as well:
Thus
The characteristic constant is then determined empirically. Variants exist parameterizing sediment transport with similar functions, e.g.,
While effective in parameterizing transport, the physical basis for this class of equations is not always obvious and it would be preferable to
begin with some consideration of the forces that are acting. The approach of Bagnold is the one most commonly applied in the marine
environment. Bagnold posits that the transport q, as measured by the weight of particles being transported should reflect the power expened per
area of bed in order to initiate transport.
Recall that the work is the integral of a force acting over some distance and the power is the work accomplished over some time. On a per area
basis then, the dimensions of force are ML-1T-2, of work MT-2 and of power MT-3.
Applying these ideas:
The dimensionless constant K can be loosely thought of as an efficiency of transport, i.e., what fraction of the work is going to motion of
sediment. (Loosely because in some situations K will exceed one somewhat.)
In this form we see an immediate problem. The transport does not approach zero at the critical stress. This has been solved in two ways
(empiricism returns), either through the parameterization of K or through introducing a modified term in the stress related term. Some examples:
Suspended Load
Analysis of the suspended load focuses on establishing the concentration profile above the bed in order to be able to directly compute the integral
Suspended load is operationally defined as those particles affected by a balance between turbulent diffusion and gravitational settling of particles.
Recall that the turbulent stress can be expressed in the form
By the law of the wall
So that
We will assume that this eddy diffusivity is closely related to the turbulent diffusivity relevant to suspension of sediment:
where is a coefficient close to one. The coefficient would be exactly one if the tracer were a passive one. That particles settle on their own in
spite of turbulence and thus will not exactly trace the patterns of turbulence suggest that the coefficient should be less than one. Conversely, the
centrifugal forces acting on the particles due to turbulent stress will give them inertia allowing better mixing suggesting the coefficient may be
greater than one. Empirical observation show
and so it is reasonable to assume that the value is one.
The balance we seek is between the settling of particles downward and their diffusive flux upward (in the direction of lower concentration) due to
turbulence:
In order to integrate this expression we need a boundary condition, but there are no likely candidates. The theories for bedload transport are not
expressed in terms of sediment concentration but rather integrated transport in a layer of uncertain dimension. Generally then we consider only
relative concentration, that is a comparison of concentration with that at some reference level. Integrating:
This is called the Rouse equation. The ratio in the exponent is called the Rouse parameter and is a measure of the relative size of the settling
velocity and turbulent stress. In practice, the reference level is established by measuring sediment concentration and size distribution at some
height off the bed in order to estimate the profile at all levels off the bed.
Because w depends on particle size, the Rouse parameter does as well. The vertical distribution of sediment reflects this, with fine material
distributed more uniformly that coarse material. Increased shear also results in a more uniform distribution.
Figure 38-1. In the left hand panel, the grains are very fine sand; in the right hand panel the shear velocity is 5 cm/sec.
We can superimpose the kind of information in this diagram onto a Hjulstrom-type diagram:
Figure 38-2
The contours depict the ratio of concentration at 100 cm to that at 2 cm. Two features bear noting:
Toward the right side of the diagram, the load is concentrated near the bed. Thus, the Rouse parameter is often used as an operation definition for
distinguishing bed load from suspended load. Various specific values have been suggested for this boundary, for example 0.8 and 2.0.
Biological Efects on Sediment Transport
Jumars et al. (38) summarize some the many possible mechanisms by which the presence of organisms can afect sediment
transport:
The mechanisms include:
the wake zone of protrusions can create zone of increased and decreased stress and concomitant changes in the
stability of the bed
burrowing organisms can expose particles to the flow (and secondarily give them momentum)
fecal pellets repackage sediment in forms with increased adhesion. These pellets also afect the roughness of the bed.
some organisms directly efect particle momentum
In order to summarize these efects in a semi-quanitative fashion recall two basic relationships. The first is the Law of the Wall
relates the shear velocity to the bulk flow:
Eq 41-1:
The parameter z is related to the roughness of the bed. From the shear velocity, the stress can be calculated.
The other relevant relationship is not as well constrain but expresses the flux of sediment in the direction of flow as a function of the excess
stress.
Eq 41-2:
Figure 41-2
This figure depicts the relationship of erosion to applied stress in the absence of organisms and the ways in which the transport is modified by the
four processes listed above. Implicit in these plots is that as applied stress becomes very large, biological effects play a relatively less important
role. An obvious limitation of this approach is that it is static. For example one can imagine a dynamic balance between the rate of track-making
by organisms and the rate of track destruction due to applied stress--effects that can not be incorporated in the structure of this figure.
Within this general framework, we can consider some specific examples from HEBBLE-sponsored experimental work conducted in the Friday
Harbor flume. A flume is an experimental appartus in which the flow of fluid through a channel can be very precisely controlled. In it
instrumented so as to characterize the velocity profile (thus u* and z are determined). During experimental runs, the motion of the bed on the
floor of the flume is visually monitored.
The Nowell et al. paper (34) describes three kinds of experiments:
1. entrainment of tracked sediments. The sediment is of 118 m grain size, wet sieved to remove macrofauna and animal
tubes. The relief is smoothed to less than 0.05 cm relief. Fluid flow through the flume is increased until motion occurs to
establish the control. Fluid flow is reduced to zero, a small motile bivalve (transenella tantilla) is added to the sediment
and the organism modifies the bed for 15 hours
Figure 41-3, from (34). The individual tracks are ~2 mm wide and have vertical relief of ~1 mm.
The critical shear stress is then again determined. Comparing the two runs:
These are the two velocity profiles, the control run and the run post-modification, just at the onset of motion.
From the data for the untracked run: z is about 0.6 m) and u* is about 1.7 cm/sec. The value of z is consistent with the bed being
hydrodynamically smooth (HSF).
From the run post-tracking: the bed is somewhat rougher but the flow is still HSF. The shear velocity required to initiate motion is about
1.4 cm/sec. Note that the roughness is not uniform--the visual protrusions are about 1 mm high but widely spaced; the effect on the
roughness scale is to increase it from .018 cm to about .036 cm.
2. fecal material from/on a coarse bed. As an example of these experiments, one run was conducted with Hobsonia
florida, a polychaete, in two diferent sediments, a fine sand and a sandy silt (both non-cohesive in the absence of the
polychaete). In these experiments the bed is smoothed, then the polychaete is introduced and allowed to feed and
defecate for 8 hours. Some of the typical fecal mounds produced:
Figure 41-5. The mound is ~4 mm high, the individual fecal pellets making up the mound, ~100 m.
The conditions of initial motion are then determined. In this situation "motion" is an imprecise term; Nowell et al. distinguish rocking of
the mound, collapse of the mounds and removal of mounds in terms of required shear velocity:
Shear Velocity Required to Induce Mound Movement

64 m silt
121 m sand
rocking
1.20 cm/sec
1.52 cm/sec
collapse
1.60 cm/sec
1.93 cm/sec
removal
1.87 cm/sec
2.18 cm/sec
3. fecal pellets ejected on a cohesive sediment (18.2 m grain size) of high shear strength. A polychaete Amphicteis
scaphobranchiata ejects pellets approximately 1 mm diameter x 9 mm long:
Figure 41-6, from (34). The tracks are approximately 1 mm wide; the pellet at the right hand end of the track is ~9 mm long.
At 1 cm/sec shear velocity these travel about 10 cm "on the fly". Pellets lying still on the bed go into motion at a shear velocity of 1.8
cm/sec; however at lower shear velocity newly ejected particles will remain in motion. The net effect is to create tracks which are more
easily eroded than for the untracked sediment.
Another approach to understanding biological efects is illustrated in Eckman et al. (39) in which some of the hydrodynamic
aspects of a tube exposed to flow are considered, as an analog to a protruding organism. They map the shear velocity around
a cylindrical protrusion. Representative results are shown in the following figures illustrating the flow, stress map and fate of
tracer particles, respectively:
Three regions are seen in the stress map:
A: an area of enhanced erosion due to applied stress, extended 1-5 tube diameters around the tube
B: upstream of the tube, an area of reduced erosion in the vicinity of the divergence in flow
C: downstream of the tube in two parallel bands, depositional ridges

Oceanography 540

Transféré par

Informations du document

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Oceanography 540

Transféré par

Droits d'auteur :

Formats disponibles

Oceanography_540.

Lecture Schedule (Autumn 2002)

Old Notes Index (2001)

Problem Set #1 Due Monday, October 14

Problem Set #2 Due Monday, October 21

Problem Set #3 Due Monday, October 28

Problem Set #4 Due Monday, November 4

Final Exam Information

Course Reserve Materials

GeoRef Database Search (UW access only)

Oceanography 540--Marine Geological Processes--Autumn Quarter 2002

Ocean 540 (i.e., Ocean 549b):

(and occasionally in Spatial Analysis Lab, 111 Ocean Sciences Building)

Course Web http://www2.ocean.washington.edu/oc540/

Problem Sets (75%)

There will be a one-hour individual oral final.

Oceanography 540--Marine Geological Processes--Autumn Quarter 2002

Figure 1-1. (from (41))

This layering is inferred to reflect a chemical layering:

crust: average composition grandiorite on continents, basalt in ocean

An Aside on Igneous Rocks In the Mantle and Crust

The remaining variations are termed the Bouguer anomaly:

when positive, indicates an excess of more dense material.

Figure 1-4. from (30)

From (2), Figure 5-3, pg 135.

Figure 1-8. From (5).

Figure 1-9a,b. Production of magnetic anomaly patterns. From (41)

Figure 1-10. From (41).

Figure 1-11. From (41)

Topography of the Ocean Basins

Fast Spreading--Southern East Pacific Rise from 13 to 23S

Medium Spreading--Juan de Fuca Ridge near 48N

Slow Spreading--Mid Atlantic Ridge near 22N

The Oceanic Crust

An Aside on Igneous Rocks In the Mantle and Crust

pillow and sheet lavas are volcanic products (rapidly cooled)

gabbros are plutonic products (slowly cooled)

dikes feeder zones for magma to rise to surface.

Figure 1-14. From (2), Figure 7-4, pg 211.

Overview of Marine Sedimentation

chemistry and mineralogy (reflecting origin of particles)

grain size(s) (reflecting energetics of transport, deposition)

suspended load: 180 x 10 g/y

dissolved load: 40 x 10 g/y (corrected for cyclic salt component)

atmospheric dust: 5 x 10 g/y

organism: forams, a protozoan

most common source is genus Globigerina

mineral form: calcite

organism: algae coccolithophoridae

mineral form: calcite

organism: planktonic molluscs

mineral form: aragonite

organism: algal diatoms

mineral form: opal

organism: radiolaria (protozoans)

mineral form: opal

Figure 13-4, from (2)

Figure 13-5, from (20)

establish markers of a sequence of events and thus relative ages of events,

correlate these markers of events either regionally or globally, and,

Earth's history is divided into ages:

Indicators of Changes in Ocean Conditions

Background Information on Sediment Classification

CaMg(CO Fe silicate oxide