Apendice Geodynamics

An Appendix to the Classic Textbook:
Geodynamics by Turcotte and Schubert
David T. Sandwell
December 26, 2018

c 2018 by David T. Sandwell
Copyright
Contents
1 Observations Related to Plate Tectonics 1
1.1 Global Maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2 Fourier Transform Methods in Geophysics 10
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2 Definitions of Fourier Transforms . . . . . . . . . . . . . . . . . . . 11
2.3 Fourier Sine and Cosine Transforms . . . . . . . . . . . . . . . . . . 12
2.4 Examples of Fourier Transforms . . . . . . . . . . . . . . . . . . . . 13
2.5 Properties of Fourier transforms . . . . . . . . . . . . . . . . . . . . 17
2.6 Solving a Linear PDE Using Fourier Methods and the Cauchy Residue
Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.7 Fourier series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3 Plate Kinematics 26
3.1 Plate Motions on a Flat Earth . . . . . . . . . . . . . . . . . . . . . . 26
3.2 Triple Junction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
ii
CONTENTS iii
3.3 Plate Motions on a Sphere . . . . . . . . . . . . . . . . . . . . . . . 31
3.4 Velocity Azimuth . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.5 Recipe for Computing Velocity Magnitude . . . . . . . . . . . . . . . 35
3.6 Triple Junctions on a Sphere . . . . . . . . . . . . . . . . . . . . . . 36
3.7 Hot Spots and Absolute Plate Motions . . . . . . . . . . . . . . . . . 36
3.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4 Marine Magnetic Anomalies 38
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.2 Crustal Magnetization at a Spreading Ridge . . . . . . . . . . . . . . 38
4.3 Uniformly Magnetized Block . . . . . . . . . . . . . . . . . . . . . . 43
4.4 Anomalies in the Earth’s Magnetic Field . . . . . . . . . . . . . . . . 43
4.5 Magnetic Anomalies Due to Seafloor Spreading . . . . . . . . . . . . 44
4.6 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
5 Cooling of the Oceanic Lithosphere 52
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5.2 Temperature versus Depth and Age . . . . . . . . . . . . . . . . . . . 56
5.3 Heat Flow versus Age . . . . . . . . . . . . . . . . . . . . . . . . . . 58
5.4 Thermal Subsidence . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
5.5 Buoyancy of the Cooling Lithosphere . . . . . . . . . . . . . . . . . 63
5.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
6 Crustal Structure, Isostasy, and Rheology 68

CONTENTS iv
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
6.2 Oceanic Crustal Structure . . . . . . . . . . . . . . . . . . . . . . . . 69
6.3 Continental Crustal Structure . . . . . . . . . . . . . . . . . . . . . . 71
6.4 Vertical Force Balance - Isostasy . . . . . . . . . . . . . . . . . . . . 72
6.5 Horizontal Force Balance - Swell Push Force . . . . . . . . . . . . . 75
6.6 Rheology of the Lithosphere . . . . . . . . . . . . . . . . . . . . . . 78
6.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
7 Brief Review of Elasticity 80
7.1 Stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
7.2 Strain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
7.3 Stress vs. Strain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
7.4 Invariants and Principal Stress . . . . . . . . . . . . . . . . . . . . . 82
7.5 Principal Stress and Strain . . . . . . . . . . . . . . . . . . . . . . . 83
7.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
8 Flexure of the Lithosphere 86
8.1 Constant Flexural Rigidity, Line Load, No End Load . . . . . . . . . 88
8.2 Variable Flexural Rigidity,

Arbitrary Line Load, No End Load . . . . . . . . . . . . . . . . . . . 93
8.3 Stability of Thin Elastic Plate Under End Load . . . . . . . . . . . . 95
8.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
9 Flexure Examples 98
9.1 Large Seamount . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99

CONTENTS v
9.2 Seamount Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
9.3 Seamount Flexure vs. Age . . . . . . . . . . . . . . . . . . . . . . . 101
9.4 Trench Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
9.5 Trench Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
9.6 Outer Rise Normal Faults . . . . . . . . . . . . . . . . . . . . . . . . 104
9.7 Brittle and Ductile Deformation . . . . . . . . . . . . . . . . . . . . 105
9.8 Trench Flexure and Earthquakes . . . . . . . . . . . . . . . . . . . . 106
9.9 Continental Yield Strength and Earthquakes . . . . . . . . . . . . . . 107
9.10 Fracture Zone Model . . . . . . . . . . . . . . . . . . . . . . . . . . 108
9.11 Fracture Zone Data . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
9.12 Normal Fault Model . . . . . . . . . . . . . . . . . . . . . . . . . . 110
9.13 Normal Fault Data . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
9.14 Thrust Fault . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
9.15 Continental Rift and Basin Evolution . . . . . . . . . . . . . . . . . . 113
9.16 Passive Margin Model . . . . . . . . . . . . . . . . . . . . . . . . . 114
9.17 Passive Margin Data . . . . . . . . . . . . . . . . . . . . . . . . . . 115
9.18 Venus Trench Data . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
9.19 Venus Trench Model . . . . . . . . . . . . . . . . . . . . . . . . . . 117
10 Elastic Solutions for Strike-Slip Faulting 118
10.1 Interseismic Strain Buildup . . . . . . . . . . . . . . . . . . . . . . . 118
10.1.1 Force Balance . . . . . . . . . . . . . . . . . . . . . . . . . . 119
10.1.2 Line-Source Green’s Function . . . . . . . . . . . . . . . . . 120
10.1.3 Screw Dislocation for Line Source Green’s Function . . . . . 122

CONTENTS vi
10.1.4 Surface Boundary Condition: Method of Images . . . . . . . 123
10.1.5 Vertical Integration of Line Source to Create a Fault Plane . . 123
10.2 Geodetic Moment Accumulation Rate . . . . . . . . . . . . . . . . . 128
10.3 Inclined Fault Plane . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
10.4 MATLAB Examples . . . . . . . . . . . . . . . . . . . . . . . . . . 133
11 Heat Flow Paradox 136
11.1 Paradox . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
11.1.1 MATLAB Example . . . . . . . . . . . . . . . . . . . . . . . 139
11.2 Moment Paradox: Seismic Moment versus Tectonic Saturation Moment 141
11.2.1 Seismic Moment Released During an Earthquake . . . . . . . 141
11.3 Tectonic Saturation Moment . . . . . . . . . . . . . . . . . . . . . . 143
11.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
12 The Gravity Field of the Earth, Part 1 145
12.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
12.1.1 Elevation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
12.1.2 Gravity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
12.2 Global Gravity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
12.2.1 Spherical Earth Model . . . . . . . . . . . . . . . . . . . . . 148
12.2.2 Ellipsoidal Earth Model . . . . . . . . . . . . . . . . . . . . 150
12.2.3 Flattening of the Earth by Rotation . . . . . . . . . . . . . . . 152
12.2.4 Measurement of J2 . . . . . . . . . . . . . . . . . . . . . . . 153
12.2.5 Hydrostatic Flattening . . . . . . . . . . . . . . . . . . . . . 154

CONTENTS vii
12.3 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
13 Reference Earth Model: WGS84 157
13.1 Some Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
13.2 Disturbing Potential and Geoid Height . . . . . . . . . . . . . . . . . 159
13.3 Reference Gravity and Gravity Anomaly . . . . . . . . . . . . . . . . 160
13.4 Free-Air Gravity Anomaly . . . . . . . . . . . . . . . . . . . . . . . 161
13.5 Summary of Anomalies . . . . . . . . . . . . . . . . . . . . . . . . . 161
14 Laplace’s Equation in Spherical Coordinates 163
14.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
14.2 Spherical Harmonics . . . . . . . . . . . . . . . . . . . . . . . . . . 164
14.3 Laplace’s Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
14.4 Earth’s Gravity Field . . . . . . . . . . . . . . . . . . . . . . . . . . 168
14.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
15 Laplace’s Equation in Cartesian Coordinates and Satellite Altimetry 171
15.1 Solution to Laplace’s Equation . . . . . . . . . . . . . . . . . . . . . 171
15.2 Derivatives of the gravitational potential . . . . . . . . . . . . . . . . 175
15.3 Conversion of Geoid Height to Vertical Deflection, Gravity Anomaly,

and Vertical Gravity Gradient from Satellite Altimeter Profiles . . . . 178
15.4 Vertical Deflections from Along-Track Slopes . . . . . . . . . . . . . 184
15.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
16 Poisson’s Equation in Cartesian Coordinates 187
16.1 Solution to Poisson’s Equation . . . . . . . . . . . . . . . . . . . . . 187

CONTENTS viii
16.2 Gravity Due to Seafloor Topography: Approximate Formula . . . . . 190
16.3 Gravity Anomaly from a 3-D Density Model . . . . . . . . . . . . . . 190
16.4 Computation of Geoid Height and Gravity Anomaly . . . . . . . . . . 191
16.5 Gravity Anomaly for a Slab of Thickness H and a Density of ρo . . . 192
16.6 Bouguer Gravity Anomaly . . . . . . . . . . . . . . . . . . . . . . . 193
16.7 Gravity Anomaly from Topography: Parker’s Exact Formula . . . . . 193
16.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
17 Gravity/Topography Transfer Function and Isostatic Geoid Anomalies 197
17.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
17.2 Flexure Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
17.3 Gravity/Topography Transfer Function . . . . . . . . . . . . . . . . . 200
17.4 Geoid/Topography Transfer Function . . . . . . . . . . . . . . . . . . 201
17.5 Isostatic Geoid Anomalies . . . . . . . . . . . . . . . . . . . . . . . 203
17.6 Isostatic Geoid and the Swell-Push Force . . . . . . . . . . . . . . . 204
17.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
18 Postglacial Rebound 206
18.1 Introduction and Dimensional Analysis . . . . . . . . . . . . . . . . 206
18.2 Exact Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
18.3 Elastic plate over a viscous half space . . . . . . . . . . . . . . . . . 210
18.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214

Introduction
Geodynamics by Turcotte and Schubert (T&S) [2014] [81] provides a deterministic,

physics-based exposition of solid-Earth processes at a mathematical level assessible
to most students. This classic textbook begins with a clear and concise overview of
plate tectonics which is followed by stress and strain in solids, elasticity and flexure,
heat transfer, gravity, fluid mechanics, rock rheology, faulting, flows in porous media,
chemical geodynamics, and the latest edition has sections on numerical modeling. I
have used this textbook in a graduate level class for the past 28 years to prepare stu-
dents in quantitative modeling of earth processes. The book uses a minimum of math-
ematical complexity so it can be understood by a wide range of students in a variety of
fields. However, this more limited mathematical approach does not provide the grad-
uate student with the tools to develop more advanced models having 3-dimensional
geometries and time dependence.
This appendix to Geodynamics was developed to augment the book with more complex
and foundational mathematical methods and approaches. The main new tool is the
introduction of multi-dimensional Fourier analysis for solving linear partial differential
equations.
Chapter 1 provides an observational discussion of plate tectonics and introduces the

latest seafloor maps of gravity and bathymetry developed in our lab.
Chapter 2 provides a brief overview of Fourier transforms and their properties, the
use of the derivative property to reduce a multi-dimensional, linear, partial differential
equation to an algebraic equation that is easily solved. The Cauchy integral theorem is
used as the standard method to perform the inverse Fourier transform of the solution to
obtain a Green’s function.
Chapter 3 uses vector calculus for rigorous plate kinematic modeling.
Chapter 4 solves the classic problem of the scalar magnetic field that is observed by a
ship as it crosses over magnetic reversals in the oceanic crust.
ix
CONTENTS x
Chapter 5 provides an enhanced derivation of the cooling and subsidence of the oceanic
lithosphere.
Chapters 6-9 introduce stress and strain in 3-dimensional using tensors and apply the
tool to the flexure and plastic deformation of the lithosphere.
Chapters 10 and 11 provide a comprehensive analysis of a continental strike-slip fault

and discuss the heat-flow paradox.
Chapters 12-15 discuss the geodetic reference frame as well as the expansion of poten-
tial/gravity anomalies in spherical harmonics and Fourier series.
Chapters 16-17 use Fourier methods to develop isostatic gravity and topography mod-
els of the oceanic lithosphere.
Chapters 18 uses Fourier methods to solve classic postglacial rebound problem.
The mathematical developments refers back to the section or equation in T&S where
the solutions are provided. These connections are highlighted in bold in this appendix.
Note that T&S contains much more information than is provided in this brief appendix
so both will be needed for a graduate-level geodynamics course. Each chapter has a set
of homewok problems that make use of the mathematical and numerical tools. These
are intended to augment the already excellent homework problems provide in T&S.
Acknowledgements The individual chapters in this book were first developed as

lecture notes using numerous versions of Microsoft Word and lovingly converted to
LATEXby the people at Dangerous Curve (typesetting@dangerouscurve.org), who have
been doing typography since 1979.
Chapter 1
Observations Related
to Plate Tectonics
(Modified from [19])
1.1 Global Maps
The plate tectonic model states that the outer shell (lithosphere) of the Earth is divided
into a small number of nearly rigid plates which slide over the weak asthenosphere.
The plates are the surface thermal boundary layer (TBL) of mantle convection and
descending slabs are the primary active components of the convective system. Plate
boundaries are generally narrow and are characterized by earthquakes and volcanoes.
It is useful to assess the global data sets that are most relevant to plate tectonics. Below
are a series of global maps that help to confirm various aspects of plate tectonic theory.
Plate boundaries are classified as ridges, transform faults, or subduction zones based
on basic observations of topography (Figure 1.1 and Figure 1.2) and seismicity (Fig-
ure 1.3). Remarkably, nearly all seafloor spreading ridges lie at a depth of 2500–3000
m below sea level, which is the level of isostasy for hot thin lithosphere. Depths grad-
ually increase away from the ridges because of cooling and thermal contraction, so old
ocean basins are commonly 4500–5000 m deep. Fracture zones and aseismic ridges
also show up on these maps. Global seismicity (magnitude >5.1, Figure 1.3) high-
lights the plate boundaries and reveals their tectonic style. Shallow normal-faulting
earthquakes (<30 km deep) are common along slow-spreading ridges, but largely ab-
sent along faster-spreading ridges where the plates are too thin and weak to retain suf-
ficient elastic energy to generate large earthquakes. Transform faults are characterized
1
CHAPTER 1. OBSERVATIONS RELATED TO PLATE TECTONICS 2
by relatively shallow (<30 km) strike-slip earthquakes and they are common along both
fast and slow-spreading ridges. The deeper earthquakes (green and blue dots in Fig-
ure 1.3) occur only in subduction zones where sheets of seismicity (i.e., Benioff zones)
are critical evidence that relatively cold lithosphere is subducting back into the mantle.
But even convergent boundaries are characterized by shallow extensional earthquakes
on the ocean side of the trenches. Some regions (e.g., Africa. Asia, Western North
America, the Indian Ocean) have distributed earthquake activity, indicating broad de-
formational zones. Topography and seismicity provide strong evidence for tectonic
activity but little or no information on the rate of plate motion.
Marine magnetic anomalies, combined with relative plate motion directions based on
satellite altimeter measurements of fracture-zone trends, have been used to construct a
global age map (Figure 1.4) of the relatively young (<180 Myr) oceanic lithosphere.
Finally, the distribution of off-ridge volcanoes that have been active during the Quater-
nary mainly occur directly behind trenches where wet subducting slabs reach astheno-
spheric depths and trigger back-arc volcanism (Figure 1.5). A few active volcanoes
occur within the interiors of the plates and in diffuse extensional plate boundaries.
The geoid (Figure 1.6) shows little correlation at long wavelengths with surface tec-
tonics, and primarily reflects mass anomalies deep in the mantle. It is expected that
the dynamic topography, the topography not due to crustal and near-surface variations,
and the stress-state of the lithosphere at long wavelengths will also reflect deep density
differences. Insofar as volcanoes correlate with high surface elevations and extensional
stress one expects correlation of volcanoes with deep mantle structure, even if there
is no material transfer. These maps are available for viewing in Google Earth at the
following site http://topex.ucsd.edu/geodynamics/tectonics.kmz .
Acknowledgements The data provided in these figures represent decades of data

collection by thousands of scientists. Figures were constructed using Generic Mapping
Tools [93].
1.2 Exercises
Exercise 1.1. Install Google Earth on your computer and download the global tec-
tonic maps (http://topex.ucsd.edu/geodynamics/tectonics.kmz). Identify the following
triple junctions and use the overlays to determine the type of deformation (R-ridge,
F-transform fault, T-trench) for each of the three boundaries. Mendocino, Galapagos,
Easter, Chile, Bouvet, Azores, and Indian Ocean triple junctions.
Exercise 1.2. Sketch a topographic profile across the Atlantic Ocean following a tec-
tonic flow line (i.e. perpendicular to isochrons) . The profile should extend from the
east coast of North America to the west coast of Africa. Label the major topographic
features. Provide approximate depths for the major topographic features. Sketch a
second profile that extends from the ridge axis to the coast of North America and also
intersects the Island of Bermuda. What are some major differences between this profile
and the first profile?
Exercise 1.3. Where is the youngest ocean floor? Where is the oldest ocean floor?
What is its approximate age?
Exercise 1.4. What types of earthquake focal mechanisms occur on the three main
types of plate boundaries?
Exercise 1.5. Use the book Geodynamics [81] to complete the following table. Devise
a thought experiment to measure each quantity. The experiments should be physically
realistic but not necessarily practical.
Example temperature - One could use a thermometer to measure temperature but that
depends on knowing the coefficient of thermal expansion. One could use the definitions
of the freezing/boiling point of water to define temperatures of 0◦C and 100◦C. One
could use the Stefan-Boltzmann law to measure temperature by measuring radiation L
from a black body at temperature T .
L = σT 4 (1.1)
where σ is the Stefan Boltzmann constant (5.67x10 Wm K ).
8 2◦ 4
parameter symbol units typical value and quantity

◦
temperature T C or ◦ K 100◦C is the boiling point of wa-
ter at 1 atm. of pressure
thermal conduc-
tivity
heat capacity
density
coefficient
of thermal
expansion
(volumetric)
acceleration of
gravity
gravitational
constant
Young’s modu-
lus
Poisson’s ratio
shear modulus
bulk modulus
dynamic viscos-
ity
CHAPTER 1. OBSERVATIONS RELATED TO PLATE TECTONICS
Figure 1.1: Topography of the Earth based on a global compilation of land data (SRTM and other sources), and ocean data from a
4
combination of sparse ship soundings and marine gravity anomalies derived from satellite altimetry [76, 2].
maximum (Mt. Everest 8848 m)
8000
6000
elevation (m)
4000
2000
mean height (743 m)

0
−2000
mean depth (-3734 m)

−4000 median depth (-4093 m)
depth (m)
−6000
−8000
−10000
maximum (Mariana Trench -11034 m)
0 1 2 3 4
area %
Figure 1.2: Histogram of the percentage of surface area (100 m bins) as a function of
elevation above sea level and depth below sea level based on a global compilation of
topography and bathymetry [2]. The earth has a bimodal histogram with the largest
peak representing the land and sumerged continental shelf. The second largest peak
represents the median ocean depth of about 4000 m. This ocean peak has two subpeaks
of unknown origin but perhaps representing the relatively uniform depth of hot-spot
swells.
6
Figure 1.3: Well-located earthquakes with magnitude >5.1 reveal the global plate boundaries [21]. Shallow earthquakes (0–70 km: red)
provide clear definition of the boundaries of the oceanic plates but have a more diffuse distribution on the continents. Intermediate (70–300
km: green) and deep (300–700 km: blue) earthquakes occur in subduction zones and are the primary evidence for lithospheric subduction
to at least 670 km.
0 30 60 90 120 150 180 -150 -120 -90 -60 -30 0
60
180
154
30 148
140
132
0 127
120 A
g
84 e
-30 68
M
56 a
48
40
-60 33
20
11
0
0 30 60 90 120 150 180 -150 -120 -90 -60 -30 0
7
Figure 1.4: Seafloor age [49] based on identified magnetic anomalies and relative plate reconstructions along trends identified in satellite
altimeter measurements of marine gravity. Ages in the Cretaceous quiet zone (64–127 Ma) have poor control.
0˚ 30˚ 60˚ 90˚ 120˚ 150˚ 180˚ -150˚ -120˚ -90˚ -60˚ -30˚ 0˚
60˚ 60˚
30˚ 30˚
0˚ 0˚
-30˚ -30˚
-60˚ -60˚
0˚ 30˚ 60˚ 90˚ 120˚ 150˚ 180˚ -150˚ -120˚ -90˚ -60˚ -30˚ 0˚
8
Figure 1.5: Plates, plate boundaries, and subaerial Quaternary volcanoes. The dozen or so tectonic plates are separated by spreading cen-
ters, transform faults, and subduction/thrust faults. The majority of active or recently active volcanoes [73] are associated with convergent
plate boundaries.
0˚ 30˚ 60˚ 90˚ 120˚ 150˚ 180˚ -150˚ -120˚ -90˚ -60˚ -30˚ 0˚
0
20
-20
-20
-20
60˚ 60
60˚
0
-40
0 0
-40
20
20
-20
20
0
30˚ 30˚
-40
-40
0
-40
604 -40 -20
-60
0˚ 60 0˚
-20
-80
40
0
-20
0 40
20
0
0
-30˚ -30˚
0
20
-20
40
-60˚ -60˚
20
20 -40
20
-6
0
0˚ 30˚ 60˚ 90˚ 120˚ 150˚ 180˚ -150˚ -120˚ -90˚ -60˚ -30˚ 0˚
9
Figure 1.6: Geoid height (EGM96) above reference ellipsoid WGS84 [39] based mostly on satellite tracking data and some terrestrial
gravity anomaly measurements. Unlike topography, seismicity, and age shown in the other maps, the geoid is poorly correlated with
surface tectonics, except in areas where mature lithosphere has subducted.
Chapter 2
Fourier Transform Methods

in Geophysics
2.1 Introduction
Fourier transforms are used in many areas of geophysics such as image processing,
time series analysis, and antenna design. Here we focus on the use of Fourier trans-
forms for solving linear partial differential equations (PDE). Some examples include:
Poisson’s equation for problems in gravity and magnetics; the biharmonic equation for
problems in linear viscoelasticity; and the diffusion equation for problems in heat con-
duction. We do not treat the wave equation in this chapter because there are already
many excellent books on seismology. For each of these problems we search for the
Green’s function that represents the response of the model to a point source. There
are two approaches to solving this class of problem. In some cases one can derive a
fully analytic solution, or Green’s function, to the point-source problem. Then a more
general model can be constructed by convolving the actual distribution of sources with
the Green’s function. A familiar example is the case of constructing a gravity anomaly
model given a 3-D density anomaly structure. The second semi-analytic approach can
be used to solve more complicated problems where the development of a fully ana-
lytic Green’s function is impossible. This involves using the derivative property of the
Fourier transform to reduce the PDE and boundary conditions to algebraic equations
that can be solved exactly in the transform domain. A more general model can be con-
structed by taking the Fourier transform of the source, multiplying by the transform
domain solution, and performing the inverse transform numerically. Indeed, the only
difference between the two methods is that in the first case the final model is generated
by direct convolution, while in the second case, the convolution theorem is used for
model generation. When dealing with spatially complex models, the second approach
10
CHAPTER 2. FOURIER TRANSFORM METHODS IN GEOPHYSICS 11
can be sometimes orders of magnitude more computationally efficient because of the

efficiency of the fast Fourier transform algorithm.
This chapter introduces the minimum amount of Fourier analysis needed to understand
the solutions to the PDE’s provided in the following chapters. If the reader is not
familiar with Fourier transforms and complex analysis, they should first study any of
the excellent books on the topic. We recommend the first six chapters of the book by
Bracewell [5] for a more complete discussion of the material presented here.
2.2 Definitions of Fourier Transforms
The 1-dimensional forward and inverse Fourier transforms are defined as:
Z∞
F (k) = f (x)e−i2πkx dx or F (k) = = f (x)

(2.1)
−∞
Z∞
f (x) = F (k) ei2πkx dk f (x) = =−1 F (k)

or (2.2)
−∞
where x is distance and k is wavenumber, where k = 1/λ and λ is wavelength. We also

use the shorthand notation introduced by [5]. The 2-dimensional forward and inverse
Fourier transforms are defined as:
Z∞ Z∞
F (k) = f (x) e−i2πk•x d2 x or F (k) = =2 f (x)

(2.3)
−∞ −∞
Z∞ Z∞
f (x) = F (k) ei2πk•x d2 k f (x) = =−1

or 2 F (k) (2.4)
−∞ −∞
where x = (x, y) is the position vector, k = (k x , ky ) is the wavenumber vector, and

k • x = k x x + ky y. The magnitude of the spatial wavenumber
1/2
β = k2x + ky2
is used in later chapters. For several of the derivations, we’ll also take the Fourier
transform in the z-direction (i.e., 3-D transform) using the following notation:
Z∞
F (kz ) = f (z) e−i2πkz z dz (2.5)
−∞
Z∞
f (z) = F (kz ) ei2πkz z dkz (2.6)
−∞
Fourier transformation with respect to time is also sometimes used to form a 4-D
transform.
Z∞
F (ν) = f (t) e−i2πνt dt (2.7)
−∞
Z∞
f (t) = F (ν) ei2πνt dν (2.8)
−∞
While algebraic manipulation of equations in 4-D is sometimes challenging and error

prone, we’ll use computers to help us in two ways. First we’ll use the tools of computer
algebra to solve the most challenging algebraic manipulations associated with the 3-D
and 4-D problems. This will result in a closed-form solution on the Fourier domain
called a transfer function. Then we’ll use the Fast Fourier Transform (FFT) algorithm
to forward transform a complicated source and inverse transform the transfer function
times this source to arrive at the final result.
2.3 Fourier Sine and Cosine Transforms
Here we introduce the sine and cosine transforms to illustrate the transforms of odd and
even functions. Also, in later chapters we’ll use sine and cosine transforms to match
asymmetric and symmetric boundary conditions for particular models.
Any function f (x) can be decomposed into odd O (x) and even E (x) functions such that
f (x) = E (x) + O (x) (2.9)
where E (x) = 12 f (x) + f (−x) and O (x) = 1

2 f (x) − f (−x) . Note the complex
exponential function can be written as
eiθ = cos (θ) + i sin (θ) (2.10)
Exercise 2.1. Use equation (2.10) to show that

1 1
cos (θ) = 2 eiθ + e−iθ and sin (θ) = 2i eiθ − e−iθ
Using this expression (2.10) we can write the forward 1-D transform as the sum of
two parts
Z∞ Z∞
F (k) = f (x) cos (2πkx) dx − i f (x) sin (2πkx) dx (2.11)
−∞ −∞
After inserting equation (2.9) into this expression and noting that the integral of an odd
function times an even function is zero, we arrive at the expressions for the cosine and
sine transforms.
Z∞ Z∞
F (k) = 2 E (x) cos (2πkx) dx − 2i O (x) sin (2πkx) dx (2.12)
0 0
Throughout this book we’ll be dealing with real-valued functions. From equation (2.12)
it is evident that the cosine transform of a real, even function is also real and even. Also
the sine transform of a real odd function is imaginary and odd. In other words, when
a function in the space domain is real valued, its Fourier transform F (k) has a spe-
cial Hermitian property F (k) = F(−k), where the overbar signifies complex conjugate.
Therefore one can reconstruct the transform of the function with negative wavenum-
bers from the transform with positive wavenumbers. Later when we perform numerical
examples using real valued functions such as topography, we can use this Hermitian
property to reduce the memory allocation for the Fourier transformed array by a factor
of 2. This is important for large 2-D and 3-D transforms.
2.4 Examples of Fourier Transforms
Throughout the book we will work with only linear partial differential equations, so all
the problems are separable and the order of differentiation and integration is irrelevant.
For example the 2-D Fourier transform is given by
∞
 ∞ 
Z  Z 
F k x , ky = −i2πk x  −i2πky y

 f (x, y) e x
dx  e dy
−∞ −∞
(2.13)
Z∞  Z∞
 
 
= f (x, y) e−i2πky y dy e−i2πkx x dx
 
 
−∞ −∞
Note that this 2-D transform consists of a sequence of 1-D transforms. This property
can be extended to 3-D, 4-D and even N-D; each transform can be performed separately
and independently of the transforms in the other dimensions. In the following analysis
we’ll only show examples of 1-D transforms, but the extension to higher dimensions
is trivial.
Delta Function By definition the delta function has the following property:
Z∞
f (x) δ (x − a) dx ≡ f (a) (2.14)
−∞
Under integration it extracts the value of f (x) at the position x = a. One can describe
the delta function as having infinite height at zero argument and zero height elsewhere.
The area under the delta function is 1. In terms of pure mathematics the delta function
is not a function and only has meaning when integrated against another function. In
this book we use the delta function as a powerful tool provided to us by the mathemati-
cians, so we trust all the mathematical theory behind it. What is the Fourier transform
of a delta function? By definition if one performs a forward transform of a function
followed by an inverse transform, or vice versa, one will arrive back with the original
function. Let’s try this using the delta function. By definition, the inverse transform of
a delta function is
Z∞
δ (k − ko )ei2πkx d k = ei2πko x (2.15)
−∞
Next let’s take the forward transform of equation (2.15). The left hand side will be the
delta function, because we have performed an inverse and forward transform. The right
hand side is given by
Z∞ Z∞
δ (k − ko ) = i2πko x −i2πkx
e e dx = e−i2π(k−ko )x dx (2.16)
−∞ −∞
This result shows that the Fourier basis functions are orthonormal. If we consider the
special case of ko = 0, we see that the inverse Fourier transform of a delta function is
=−1 [δ (k)] = 1. Since Fourier transformation is reciprocal in distance x and wavenum-
ber k, it is also true that = [δ (x)] = 1. The delta function and its Fourier transform
provide an amazingly powerful tool for solving linear PDEs.
Cosine and Sine Functions Let’s use the delta function tool and the expressions
from Exercise 2.1 to calculate the Fourier transform of a cosine function having a
single wavenumber cos (2πk0 x).
Z∞ Z∞
1
−i2πkx
cos (2πko x) e dx = 2
ei2πko x + e−i2πko x e−i2πkx dx
−∞ −∞
(2.17)
1
= 2 [δ (k − ko ) + δ (k + ko )]
So the Fourier transform of a cosine function is simply two delta functions located
at ±ko .
Exercise 2.2. Show that the Fourier transform of sin (2πk0 x) is
1
δ (k − ko ) − δ (k + ko ) .

2i
2
Gaussian Function The Gaussian e−πx function also plays a fundamental role in
solutions to several types of PDEs. Its Fourier transform is
Z∞ Z∞
+i2kx)
−πx2 −i2πkx
e−π( x
2
F (k) = e e dx = dx (2.18)
−∞ −∞

Note that (x + ik)2 = x2 + i2kx − k2 . Using this we can rewrite equation (2.18) as
Z∞ Z∞
−πk2 −π(x+ik)2 −πk2 2 2
F(k) = e e dx = e e−π(x+ik) d (x + ik) = e−πk (2.19)
−∞ −∞
2
where we have used the result that the infinite integral of e−πx is 1. This is a remarkable
and powerful result that the Fourier transform of a Gaussian is simply a Gaussian. See
Figure 2.1.
Figure 2.1: Schematic plots of 1-D Fourier transform pairs. Solid line is real-valued
function while dashed line is imaginary valued function (figure from [5]).
2.5 Properties of Fourier transforms
There are several properties of Fourier transforms that can be used as tools for solving
PDEs.
Similarity Property The first property called the similarity property says that if you
scale a function by a factor of a along the x-axis, its Fourier transform will scaled by
a−1 along the k-axis and the amplitude will be scaled by |a|−1 .
Exercise 2.3. Use the definition of the Fourier transform equation (2.1) and a change
of variable to show the following. Try positive and negative values of a to understand
why the absolute value is needed in the amplitude scaling.
1 k
= f (ax) = F a

(2.20)
|a|
Shift Property The shift property says that the Fourier transform of a function that
is shifted by a along the x-axis equals the original Fourier transform scaled by a phase
factor. This property is especially useful for numerically shifting a function a non-
integer amount of the data spacing along the axis.
Exercise 2.4. Use the definition of the Fourier transform and a change of variable to
show the following
= f (x − a) = e−i2πka F(k)

(2.21)
Derivative Property The derivative property of the Fourier transform is the essential
tool used in this book to transform linear PDEs into algebraic equations that are easily
solved. It states that the Fourier transform of the derivative of a function is the Fourier
transform of the original function scaled by the imaginary wavenumber.
∂f
" #
= = i2πk F(k) (2.22)
∂x
To show this is true we start with the inverse transform of equation (2.22)
Z∞
∂f
= i2πk F(k) ei2πkx dk (2.23)
∂x
−∞
Next take the forward transform of equation (2.23) and rearrange terms
# Z∞ Z∞
∂f
"
= = i2πko F(ko ) ei2πko x dko e−i2πkx dx
∂x
−∞ −∞
(2.24)
Z∞
 ∞ 

 Z 

= −i2π(k−ko )x
 
i2πko F(ko )  e dx dko
 

 


 
−∞ −∞
The term in the curly brackets is the delta function δ (k − ko ) given in equation (2.16).
The result is
" # Z∞
∂f
= = i2πko F(ko ) δ (k − ko ) dko = i2πk F(k) (2.25)
∂x
−∞
Convolution Theorem The final property considered here is the convolution theo-
rem, which states that the Fourier transform of the convolution of two functions is
equal to the product of the Fourier transforms of the original functions.
 ∞ 
Z 
=  f (u) g (x − u) du = F(k) G(k) (2.26)
 
 
−∞
To show this is true one can perform the Fourier integration on the left side of equa-
tion (2.26) and rearrange the order of the integrations
 ∞
 Z∞  Z∞
  
 Z 
=  f g(x − u) du =
(u) f g(x − u) du e−i2πkx dx
(u)
   
   
−∞ −∞ −∞
(2.27)
Z∞
 ∞ 

 Z 

= g(x − u) e−i2πkx dx
 
f (u)  du
 


 


−∞ −∞
Next use the shift property of the Fourier transform to note that the function in the curly
brackets on the right side of equation (2.27) is e−i2πkuG (k). The result becomes
 ∞
Z∞

 Z 
=  f (u) g(x − u) du = G(k) f (u) e−i2πku du = F(k) G(k) (2.28)
 
 
−∞ −∞
Note that these four properties are equally valid in 2-dimensions or even N-dimensions.
The properties also apply to discrete data. See Chapter 18 in [5].
Cauchy Residue Theorem The Cauchy Residue Theorem is an additional tool that
we will use many times in the book to perform inverse transforms for cases where the
function is analytic and has poles in the complex plane. Let f (z) be an analytic function
in the complex plane z = x + iy.
An analytic function has the special property that a path integral of the function about
any closed loop in the complex plane is zero.
I
f (z) dz = 0 (2.29)
As an example, the gravitational potential associated with the topography of the earth
represents an analytic function. A cyclist riding along any closed path will gain and
lose potential energy along the path, but no matter how the path is traversed, he will
have the same potential at the end of the circuit as he started with. Next suppose the
same integration is performed with a pole (zero) in the denominator at a complex point
zo . The Cauchy Residue Theorem states
I
f (z)
dz = i2π f (zo ) (2.30)
z − zo
The function in the numerator could be very complicated, but as long as it is analytic,
the path integral can be evaluated.
Exercise 2.5. Without using the Cauchy Residue Theorem show that the following
is true. I
1
z
dz = i2π (2.31)
2.6 Solving a Linear PDE Using Fourier Methods

and the Cauchy Residue Theorem
In the following chapters we’ll derive the Green’s function and/or its Fourier transform,
starting from a PDE and boundary conditions. The approach will follow the same
format, so here we have selected a simple example to illustrate the general method.
Heat Flow for a Line Source of Heat at Depth
Consider a line source of heat at a depth of −a buried in a conductive half-space, as

shown in Figure 2.2. Calculate the temperature in the half-space. The differential
equation and three boundary conditions are
∂2 T ∂2 T
+ 2 = δ (x) δ (z + a)
∂x2 ∂z
T (x, 0) = 0
(2.32)
lim T (x, z) = 0
|x|→∞
lim T (x, z) = 0
z→∞
The first step in the solution is to take the 2-D Fourier transform of the PDE. Each
a y
-a
Q(x,z)=(x)(z+a)
Figure 2.2: Line source of heat at z = −a. The surface boundary condition can be met
by placing an equal but opposite line sink of heat at z = a.
derivative on the left hand side is replaced with an i2πk according to the derivative
property of the Fourier transform, and the 2-D transform of the right hand side is done
using the definition of the Delta function. The result is

−4π2 k2x + kz2 T (k x , kz ) = ei2πkz a
ei2πkz a (2.33)
T (k x , kz ) =
−4π2 k2x + kz2
Now that we have solved the algebraic problem, we’ll start by taking the inverse Fourier
transform in the z-direction.
Z∞
−1 ei2πkz (z+a)
T (k x , z) = dkz (2.34)
4π2

kz2 + k2x
−∞
The denominator of equation (2.34) can be factored as (kz + ik x ) (kz − ik x ), which rep-
resent two poles in the complex plane (Figure 2.3.)
Z∞
−1 ei2πkz (z+a)
T (k x , z) = 2 dkz (2.35)
4π (kz + ik x ) (kz − ik x )
−∞
Im kz
ikx
Re kz
-ikx
Figure 2.3: Closed integration path about a pole in the complex plane.
To perform this integration we’ll integrate around one of these poles using the Cauchy
Residue Theorem. First consider the case k x > 0, z > −a. We would like to integrate
along the kz axis from−∞ to ∞, as shown in Figure 2.3. If we close the integration path
in the upper hemisphere of the complex plane, the numerator will become vanishingly
small because it is a decaying exponential function. Therefore the integration along
the kz axis will be equivalent to the full integration around the pole kz = ik x . Using
equation (2.30), the result is
−i2π ei2πikx (z+a) −e−2πkx (z+a)

T (k x , z) = = (2.36)
4π2 2ik x 4πk x
Note this is a decaying exponential function of z that satisfies the last boundary con-
dition in equation (2.32). Next consider the case k x < 0, z > −a. This time we close
the integration path in the lower hemisphere to also achieve a decaying exponential
function. Integration about the pole in the clockwise direction reverses the sign in the
Cauchy Residue Theorem so the result is
−i2π e−i2πikx (z+a) e2πkx (z+a)

T (k x , z) = = (2.37)
4π2 −2ik x 4πk x
Equations (2.36) and (2.37) can be combined by using the absolute value of k x .
−e−2π|kx |(z+a)
T (k x , z) = (2.38)
4π |k x |
The next step is to take the inverse transform with respect to k x

Z∞
−1 e−2π|kx |(z+a) i2πkx x
T (x, z) = e dk x (2.39)
4π |k x |
−∞
One way to perform this final integration is to use the derivative property of the Fourier
transform to find the solution for ∂T/∂z and then integrate over z to get the desired result.
Z∞
∂T (x, z) 1
= e−2π|kx |(z+a) ei2πkx x dk x (2.40)
∂z 2
−∞
One can look up this definite integral

∂T (x, z) − (z + a)
= i, z > −a (2.41)
∂z
h
4π x2 + (z + a)2
After integrating over z, we find

−1 h i1/2
T (x, z) = 4π ln x2 + (z + a)2 (2.42)
Finally, we have not yet met the surface boundary condition T (x, 0) = 0. This can be
achieved by placing a line heat sink at z = a. The sum of the source and sink satisfy
the differential equation and the four boundary conditions.
h i1/2 i1/2
−1 h
T (x, z) = 4π ln x2 + (z + a)2 − ln x2 + (z − a)2 (2.43)
2.7 Fourier series
Many geophysical problems are concerned with a small area on the surface of the Earth
having a width of W and length of L, as shown in Figure 2.4. The coefficients of the
2-dimensional Fourier series are computed by the following integration.
ZL ZW
1 m n
Fnm = f (x, y) exp −i2π x + y dy dx (2.44)
LW L W
o o
The function is reconstructed by the following summations over the Fourier coeffi-
cients.
∞ ∞ h m n i
X X
f (x, y) = Fnm exp i2π L x + W y (2.45)
n=−∞ m=−∞
The finite size of the area leads to a discrete set of wavenumbers k x = m/L, ky =
n/W and a discrete set of Fourier coefficients Fnm . In addition to the finite size of the
y
W
∆y
∆x
L x
Figure 2.4: Cartesian coordinate system used throughout the book with z positive up.
The z = 0 plane is the surface of the earth. Fourier transforms deal with infinite do-
mains while the Fourier series has finite domains. For our numerical examples we
will select an area of length L and with W consisting of uniform cells of size ∆x and
∆y. This can be represented as a 2-D array of numbers with J = L/∆x columns and
I = W/∆y rows.
area, geophysical data commonly have a characteristic sampling interval ∆x and ∆y.
Note that I = L/∆x is the number of points in the x-direction and J = W/∆y is the
number of points in the y-direction. The Nyquist wavenumbers are k x = 1/ (2∆x) and
k x = 1/ (2∆x), so there is a finite set of Fourier coefficients −I/2 < m < I/2 and
−J/2 < n < J/2. Recall the trapezoidal rule of integration.
ZL I−1
X
f (x) dx f (xi )∆x where xi = i∆x
0 i=0
(2.46)
ZL I−1
L X
f (x) dx f (xi )
I i=0
0
The discrete forward and inverse Fourier transform are:

I−1 X
J−1
1 m n i
X h
j
Fnm = I J fi exp −i2π I i + J j (2.47)
i=0 j=0
I/2−1 J/2−1
i j
X X
j
fi = Fnm exp i2π I m + J n (2.48)
n=−I/2 m=−J/2
The summations for the forward and inverse discrete Fourier transforms are similar so
one can use the same computer code for both transforms. Sorry for the dual use of
√
the symbol ‘i.’ The italic ‘i’ in front of the 2π is −1, while the other, non-italic ‘i’s
are integers.
2.8 Exercises
Exercise 2.6. What is the Fourier transform of the following function? Show your
work and simplify the result.




1 |x| < 1/2
Π(x) =  |x| = 1/2

(2.49)

1/2


0
 otherwise
Exercise 2.7. Use the convolution theorem to calculate the Fourier transform of the
following. Show your work.

1 − |x| |x| ≤ 1

Λ(x) = 

(2.50)
0
 |x| > 1
Note Λ = Π ∗ Π.
Exercise 2.8. Perform the following path integral on |z| = 2.
I
z+2
dz (2.51)
2 (z + 1)
Exercise 2.9.
%
% 1) Write a program to generate a cosine function
% using 2048 points. Generate exactly 32, or 64 cycles
% of the function. Plot the results and add labels.
%
figure(1)
clf
nx=2048;
kc=64/nx;
x=0:nx-1;
%
% generate the function
%
y=cos(2*pi*x*kc);
%
figure(1)
plot(x,y);
xlabel(’x’)
ylabel(’cos(x)’)
pause
%
% 2) Take the fourier transform of the function that you made in problem 1.
% Use fftshift to shift the zero frequency to the center of the spectrum.
% Generate wavenumbers for the horizontal axis.
% Take the inverse FFT. Do you get what you started with? (don’t
% forget to undo the fftshift.)

%
figure(2)
subplot(5,1,1),plot(x,y);
xlabel(’x’)
%
% generate the wavenumbers
%
k=-nx/2:nx/2-1;
%
cy=fftshift(fft(y));
subplot(5,1,2),plot(k,real(cy));
xlabel(’k’)
subplot(5,1,3),plot(k,imag(cy));
%
% do the inverse FFT
%
yo=ifft(fftshift(cy));
subplot(5,1,4),plot(x,real(yo));
xlabel(’x’)
subplot(5,1,5),plot(x,real(y-yo));
xlabel(’x’)
ylabel(’difference’)
pause
%
% 3) Do problem 2 over using a sine function instead of a cosine function.
%
% 4) Show that the Fourier transform of a Gaussian function is a Gaussian function.
% Plot the difference between the fft result and the exact function.
% When you do this problem, it is best to make the Gaussian function an even function
% of x just prior to computing the fft(). If you do this then the transformed
% Gaussian will be real and even. Also you will need to scale the transform by
% the point spacing dx = L/nx.
%
clear
figure(3)
nx=2048;
L=20;
dx=L/nx;
a=1.;
x=a*(-nx/2:nx/2-1)*dx;
g=exp(-pi*x.*x);
subplot(4,1,1),plot(x,g);
axis([-4,4,-.5,1.1])
xlabel(’x’)
ylabel(’Gaussian’)
%
% generate the wavenumbers
%
k=(-nx/2:nx/2-1)/L;
%
cg=fftshift(fft(fftshift(g)))*dx;
%
% 5) Use this Gaussian example to demonstrate the stretch property of Fourier transform. The
% results should be compared in the wavenumber domain.
%
%
% 6) Use this Gaussian function to illustrate the shift property of the Fourier transform.
% The results should be displayed as a shifted Gaussian in the space domain.
%
%
% 7) Use the Gaussian function to demonstrate the derivative property of the Fourier
% transform. The analytic derivative of the Gaussian should be compared with the
% Fourier derivative in the space domain.
%
Chapter 3
Plate Kinematics
(Reference: [25, Chapter 2])
3.1 Plate Motions on a Flat Earth
Plate tectonic theory describes the motions of rigid plates on a spherical Earth. How-
ever, when considering the relative motions very close to the plate boundary, or at a
triple junction, it is appropriate to use a flat-Earth approximation. We’ll begin with
the flat-Earth case and then move on to the spherical case. Consider the two plates A
and B, which have a subduction zone boundary between them, such as the Nazca and
South American plates. In this analysis all plate motions are relative, so one can ei-
ther consider plate B as fixed or plate A as fixed, and draw the relative vector velocity
between them—as shown in Figure 3.1.
26
CHAPTER 3. PLATE KINEMATICS 27
A B A B
VAB fixed fixed VBA
VAB velocity vector of plate A relative to plate B.

VBA velocity vector of plate B relative to plate A.
VAB = −VBA
VBA = V x i + Vy j
Figure 3.1
3.2 Triple Junction
A triple junction is the intersection of three plate boundaries. The most common types
of triple junctions are ridge-ridge-ridge (R-R-R), ridge-fault-fault (R-F-F), and ridge-
trench-trench (R-T-T); see Figure 3.2.
Each type of plate boundary has rules about relative velocities:
Ridge relative velocity must be divergent and is usually perpendicular to the ridge.
Transform Fault relative velocity must be parallel to the fault.
Trench relative velocity must be convergent, but no direction is preferred.
All triple junctions must satisfy a velocity condition such that the vector sum around
the plate circuit is zero:
VBA + VCB + VAC = 0 (3.1)
In the real world we usually can map the geometry of the spreading ridges, transform
faults, and trenches, but cannot always measure the relative velocities. The triple junc-
A
A B A B
(fixed) (fixed)
B
C C C
R-R-R R-F-F R-T-T
Figure 3.2: Three of the most common triple junctions. R-ridge, F-fault, T-trench
tion closure equation (3.1) can be used to solve for spreading velocities given the triple
junction geometry, the rules, and at least one relative plate velocity.
EPR
o
A 260
o
95
Cocos
EPR
Figure 3.3: Galapagos Triple Junction: R-R-R
Example 3.1. Given the geometry in Figure 3.3 and one spreading velocity |VAC | =
120 mm/yr, what are the other two spreading rates? Also see Figure 3.4.
3
Example: Given the above geometry and one spreading velocity VAC = 120 mm/yr,
what are the other two spreading rates?
VBA
85 sum of interior angles = 180˚
15
VCB Sum of interior angles = 180◦
VAC 80
Solving for the lengths of the vectorsFigure

we find
3.4VBA = 118.45 mm/yr and VCB = 31.12
mm/yr. One of these flat-earth triple junction closure problems will be on the quiz. The
map on the next page shows the other triple junctions. As an exercise, use a bathymetric
mapfor
Solving (e.g.,
theGoogle Earth
lengths ofand
thethis we find |VBA | = 118.45 mm/yr and |VCB | =
KMZ-file
vectors
31.12 mm/yr. z. The map in Figure 3.5 shows the other triple) junctions.
ftp://topex.ucsd.edu/pub/srtm30_plus/SRTM30_PLUS_V7.kmz to determine
As anthe
ex-
geometry of another triple junction and use Table 1 (below) to calculate the spreading
ercise, use a bathymetric map (e.g., Google Earth and this KMZ-file ftp://topex.
rate at one of the ridges. The next section develops the mathematics for calculation of
ucsd.edu/pub/srtm30_plus/SRTM30_PLUS.kmz)
plate motions on a spherical earth. to determine the geometry of an-
other triple junction, and then use Table 3.1 in the next section to calculate the spreading
rate at one of the ridges. The next section develops the mathematics for calculation of
plate motions on a spherical Earth.
CHAPTER 3. PLATE KINEMATICS
30
Figure 3.5: Names of major plates and plate boundaries from Fowler [25].
3.3 Plate Motions on a Sphere

5
(Reference: [47, 15, 16])
Plate Motions on a Sphere
(Minster, J-B, T.H. Jordan, Present-day ω

plate motions, J. Geophys. Res., 85, p.
5331-5354, 1978.
Given: DeMets et al., Current plate motions,
Geophys. J. R. Astr. Soc., 101, p.425- v
478, 1990. rad
ω angular
DeMets, velocity vector
C., R. G. Gordon, and D. F.
s
r position on Earthcurrent
Argus, Geologically (m) plate r Δ
motions, Geophys. J. Int., 181, 1-80,
2010)
Calculate:Given:

m
v velocity vector at position r
# rads&
ω − angular velocity vector % (
$ s '
r − position on earth (m)
Calculate:
Of course, the velocity of the plate must be tangent to the surface of the Earth, so the
" m%
velocity isv the
− velocity
crossvector at position
product of ther $position
# s '& vector and the angular velocity vector.
Of course, the velocity of the plate vmust= ωbe×tangent

r to the surface of the earth so the (3.2)
velocity is the cross product of the position vector and the angular velocity vector.
or
v = ı̂ ωy z − ωz y − ̂ (ω x z − ωz x) + k̂ ω x y − ωy x (3.3)
v=ω×r (2)
where i, j, and k are unit vectors. The magnitude of the velocity is given by
or
|v| = |ω| |r| sin (∆) (3.4)
( ) (
v = iˆ ω y z − ω z y − ĵ (ω x z − ω z x ) + k̂ ω x y − ω y x ) (3)
where ∆ is the angle between the position vector and the angular velocity vector. It is
given by the following
where formula:
i, j, and k are unit vectors. The magnitude of the velocity is given by
ω•r
v = ω r sin(Δ) cos (∆) = (4) (3.5)
|ω| |r|
where Δ is the angle between the position vector and the angular velocity vector. It is
given by the following formula.
The formulas above assume that the angular velocity vector and the position vector
ω•r
are provided in =Cartesian
cos(Δ) coordinates. However,
(5) usually they are specified in terms
ω r
of latitude and longitude. Thus one must transform both vectors. The usual case is
to calculate the relative velocity between two plates somewhere along their common
boundary. Table 3.1 lists the pole position and rates of rotation for relative motion
between plate pairs shown in Figure 3.6. The Cartesian position of a point along the
plate boundary is
x = a cos θ cos φ
y = a cos θ sin φ (3.6)
z = a sin θ
where θ is latitude, φ is longitude, and a is the mean earth radius. It is helpful to

memorize the conversion from latitude-longitude to the Cartesian coordinate system
where the x-axis runs from the center of the Earth to a point at 0◦ latitude and 0◦
longitude (i.e., the Greenwich meridian), the y-axis runs through a point at 0◦ latitude
and 90◦ east longitude, and the z-axis runs along the spin-axis to the North Pole.
Similarly the pole positions must be converted from geographic coordinates (θ p , φ p )

into the Cartesian system
ω x = |ω| cos θ p cos φ p

ωy = |ω| cos θ p sin φ p (3.7)
ωz = |ω| sin θ p
where |ω| is the magnitude of the rotation vector provided in Table 3.1. There are two
ways to compute the magnitude of the velocity. One could compute the cross product
of the rotation vector and the position vector (equation (3.2)). Then the magnitude of
the velocity is
1/2
|v| = v2x + v2y + v2z (3.8)
One could also calculate the angle ∆ between the position vector and the angular veloc-
ity vector using equation (3.5) and then use that value in equation (3.4) to calculate the
magnitude of the velocity. Indeed, both Fowler [25] and Turcotte and Schubert [81] use
this second approach. However, they use the rather cumbersome spherical trigonom-
etry to calculate the angle ∆. Since I don’t remember the spherical trigonometry for-
mulas, I prefer to use equation (3.4) above after converting everything to Cartesian
coordinates.
MORVEL: [16]
Table 3.1: Best-fitting angular velocities and covariances.

Figure 3.6: (A) Epicentres for Earthquakes with magnitudes equal to or larger than
3.5 (black) and 5.5 (red), and depths shallower than 40 km, for the period 1967–2007.
Hypocentral information is from U.S. Geological Survey National Earthquake Infor-
mation Center files. (B) Plate boundaries and geometries employed for MORVEL.
Plate name abreviations are as follows: AM, Amur; An, Antarctic; AR Arabia; AU,
Australia; AZ, Azores; BE, Bering; CA, Caribbean; CO, Cocos; CP, Capricorn; CR,
Caroline; EU, Eurasia; IN, India; JF, Juan de Fuca; LW, Lwandle; MQ Macquarie; NA,
North America; NB, Nubia; NZ, Nazca; OK, Okhotsk; PA, Pacific; PS, Philppine Sea;
RI, Rivera; SA, South America; SC, Scotia; SM, Somalia; SU, Sundaland; SW, Sand-
wich; YZ, Yangtze. Blue labels indicate plates not included in MORVEL. Patterned
red areas show diffuse plate boundaries, Figure from [16].
3.4 Velocity Azimuth
We know that the velocity vector is tangent to the sphere. Given the Cartesian velocity
components from equation (3.3), we would like to compute the latitude vθ and longi-
tude vφ components of velocity. Begin by taking the time derivative of equation (3.6):

v x = a − cos φ sin θ vθ − cos θ sin φ vφ

vy = a − sin φ sin θ vθ + cos θ cos φ vφ (3.9)
vz = a (cos θ vθ )
From the last equation in 3.9, we can solve for the latitude velocity component.
vz
vθ = (3.10)
a cos θ
Now plug vθ into either the v x or vy equation and solve for vφ .
vy + vz sin φ tan θ
vφ = (3.11)
a cos θ cos φ
If this equation turns out to be singular, then use the v x equation.
v x + vz cos φ tan θ
vφ = − (3.12)
a cos θ sin φ
3.5 Recipe for Computing Velocity Magnitude
In summary, to calculate the magnitude of the velocity:
1. Transform lat and lon into the unit vector x = (x, y, z) using equation (3.6)
2. Transform pole lat and lon into the unit vector x p = (x p , y p , z p ) using equa-
tion (3.7)
3. cos ∆ = x • x p
4. v = ωa sin ∆
Example 3.2. Given the rotation pole between the Pacific and Nazca plates, calculate
the spreading rate at −20◦ 113.5◦ W.
Pole Point
52.7 −88.6 1.326 × 10−6 deg/yr 20.0◦ S 113.5◦ W

52.7 271.4 2.314 × 10−8 rad/yr −20.0 246.5
x p = 0.0148 x = −0.375
y p = −0.606 y = −0.862
z p = 0.795 z = −0.342
cos ∆ = x • x p = (−0.0056 + .522 − .272) = .244

∆ = 75.8 v = ω a sin ∆ = 142.9 mm/yr
3.6 Triple Junctions on a Sphere
Triple junction closure on a sphere is similar to triple junction closure on a flat-Earth

except that the sum of the rotation vectors must be zero:
ωBA + ωCB + ωAC = 0 (3.13)
Example 3.3. Galapagos Triple Junction Given the rotation vectors of the Cocos
plate relative to the Pacific plate and the Pacific plate relative to the Nazca plate, calcu-
late the spreading rate at 2◦ N, 260◦ E.
ωCP + ωNC + ωPN = 0
ωNC = −ωCP − ωPN
vNC = ωNC × r(θ, φ)
|v| - magnitude of spreading rate
3.7 Hot Spots and Absolute Plate Motions
So far we have only considered relative plate motions because there was no way to tie
the positions of the plates to the mantle. One method of making this connection and
thus determining absolute plate motions is to assume that “hot spots” remain fixed with
respect to the lower mantle.
A hot spot is an area of concentrated volcanic activity. There is a subset of hot spots
that have the following characteristics:
1. They produce linear volcanic chains in the interiors of the plates.

2. The youngest volcanoes occur at one end of the volcanic chain, and there is a
linear increase in age away from that end.
3. The chemistry of the erupted lavas is significantly different from lava erupted at
mid-ocean ridges or island arcs.
4. Some hotspots are surrounded by a broad topographic swell about 1000 m above
the surrounding ocean basin.
These features are consistent with a model where the plates are moving over a relatively
fixed mantle plume. After identifying the linear volcanic chains associated with the
mantle plumes, it has been shown that the relative motions among hot spots is about 10
times less than the relative plate motions.
3.8 Exercises
Exercise 3.1. Calculate the spreading rate at a point on the northern Mid-Atlantic
Ridge (latitude 30, longitude 319).
Exercise 3.2. Calculate the slip rate along the San Andreas Fault in San Francisco
(latitude 38, longitude -122.7). You will need to use tables from [16].
Exercise 3.3. Where is the fastest seafloor spreading ridge on the Earth? Use Table
3.1 to calculate the spreading rate at that location.
Exercise 3.4. The vector sum of relative plate velocities around a triple junction is
zero.
vBA + vCB + vAC = 0 (3.14)
Show that the following is also true at a position ro .
ωBA + ωCB + ωAC = 0 (3.15)
where the ω’s are the relative rotation poles on a sphere and ro is not parallel to any of
the ω’s.
Exercise 3.5. Use the Google Earth overlays of vertical gravity gradient and earth-
quake epicenters to sketch the geometry of the ridges and transform faults around the
Indian Ocean triple junction. Given the spreading rate along the Southeast Indian ridge
is 50 mm/yr, calculate the spreading rates on the Southwest Indian and Central Indian
ridges.
Chapter 4
Marine Magnetic Anomalies
(Copyright, 2001, David T. Sandwell)
4.1 Introduction
This chapter develops the equations needed to compute the scalar magnetic field that
would be recorded by a magnetometer towed behind a ship, given a magnetic timescale,
a spreading rate, and a skewness. A number of assumptions are made to simplify the
mathematics. The intent is to first review the origin of natural remnant magnetism
(NRM), to illustrate that the magnetized layer is thin compared with its horizontal
dimension. Then the relevant differential equations are developed and solved under
the ideal case of seafloor spreading at the north magnetic pole. This development
highlights the Fourier approach to the solution to linear partial differential equations.
The same approach will be used to develop the Green’s functions for heat flow, flexure,
gravity, and elastic dislocation. For a more general development of the geomagnetic
solution, see the reference by Parker [52].
4.2 Crustal Magnetization at a Spreading Ridge
As magma is extruded at the ridge axis, its temperature falls below the Curie point and
the uppermost part of the crust becomes magnetized in the direction of the present-day
magnetic field. Figure 4.1,
38
CHAPTER 4. MARINE MAGNETIC ANOMALIES 39
Figure 4.1: Model of crustal structure derived from reflection and refraction seismol-
ogy [36].
from Kent et al. [36], illustrates the current model of crustal generation. Partial melt
that forms by pressure-release in the uppermost mantle (∼40 km depth) percolates to a
depth of about 2000 m beneath the ridge, where it accumulates to form a thin magma
lens. Beneath the lens a mush-zone develops into a 3500-m thick gabbro layer by
some complicated ductile flow. Above the lens, sheeted dikes (∼1400 m thick) are
injected into the widening crack at the ridge axis. Part of this volcanism is extruded
into the seafloor as pillow basalts. The pillow basalts and sheeted dikes cool rapidly
below the Curie temperature as cool seawater percolates to a depth of at least 2000 m.
This process forms the basic crustal layers seen by reflection and refraction seismology
methods.
The highest magnetization occurs in the extrusive layer 2A (Figure 4.2, Table 4.1),
although the
Figure 4.2: NRM values (in A/m) from Hole 504B. Depths measured from the seabot-
tom and include 274.5 m of sediment. The horizontal lines separate the upper units,
the transition zone, and the dike complex. (From [74])
Layers Thermoremnant
Thickness/Velocity Description
Magnetism (TRM)
Layer 1 variable
sediment N/A
<2.5 km/s
Layer 2A 400–600 m extrusive,

5–10 A/m
2.2–5.5 km/s pillow basalts
Magma Lens
Layer 2B 1400 m intrusive,

∼1 A/m
5.5–6.5 km/s sheeted dikes
Layer 3 3500 m
intrusive, gabbro ∼1 A/m
6.8–7.6 km/s
Table 4.1
dikes and gabbro layers provide some contribution to the magnetic anomaly measured
on the ocean surface. Note that the reversals recorded in the gabbro layer are do not
have sharp vertical boundaries (Figure 4.3). The tilting reflects the time delay when
the temperature of the gabbro falls below the Curie point. The sea-surface magnetic-
anomaly model shown in Figure 4.3 [28] includes the thickness and precise geometry
of the magnetization of all three layers. For the calculation below, we assume all of the
magnetic field comes from the thin extrusive layer.
The other assumptions are that the ridge axis is 2-D, so there are no along-strike varia-
tions in magnetization, and that the spreading rate is uniform with time. Before going
into the calculation, we briefly review the magnetic field generated by a uniformly
magnetized block.
Figure 4.3: Magnetic anomalies generated by realistic model of crustal magnetization.

The primary magnetization signature comes from the thin layer of extrusives. Dipping
magnetization in the Gabbros reflects the position of the Curie isotherm at depth away
from the ridge axis. (Jeff Gee, personal communication.)
5.5 - 6.5 km/s dikes
layer 3 3500 m intrusive, gabbro ~1 A/m
6.8 - 7.6km/s
The other assumptions are the ridge axis is 2-D so there are no along-strike variations in
magnetization and the spreading rate is uniform with time. Before going into the
calculation, we briefly review the magnetic field generated by a uniformly magnetized
block.
4.3 Uniformly Magnetized

Uniformly Magnetized Block Block
ΔB
M - magnetization vector
−1 (Am-1)
M magnetization
ΔB - vector (Am )
magnetic anomaly vector (T)
µo - magnetic permeability (4π x 10-7 TA-1m) M
∆B magnetic anomaly vector (T)
µo magnetic permeability (4π × 10−7 TA−1 m)
A magnetized rock contains minerals of magnetite and haematite that can be preferen-
tially aligned in some direction. For a body with a uniform magnetization direction,
the magnetic anomaly vector will be parallel to that direction. The amplitude of the
external magnetic field will have some complicated form:
∆B(r) = µo M f (r) (4.1)
where f (r) is a function of position that depends on geometry. The total magnetization
of a rock has two components: thermoremnant magnetism (TRM) MT RM , and magne-
tization that is induced by the present-day dipole field MI :
M = MT RM + MI MI = χH (4.2)
where χ is the magnetic susceptibility and H is the applied dipole field of the Earth.
The Koenigberger ratio Q is the ratio of the remnant field to the induced field. This
value should be much greater than 1 to be able to detect the crustal anomaly. Like the
magnetization, the value of Q is between 5 and 10 in Layer 2A but falls to about 1
deeper in the crust.
4.4 Anomalies in the Earth’s Magnetic Field
When a magnetometer is towed behind a ship, one measures the total magnetic field B,
and must subtract out the reference Earth magnetic Be field to establish the magnetic
anomaly ∆B:
B = Be + ∆B (4.3)
Most marine magnetometers measure the scalar magnetic field. This is an easier mea-
surement because the orientation of the magnetometer does not need to be known. The
total scalar magnetic field is
1/2
|B| = |Be |2 + 2Be • ∆B + |∆B|2 (4.4)
The dipolar field of the Earth is typically 50,000 nT, while the crustal anomalies are
only about 300 nT. Thus |∆B|2 is small relative to the other terms and we can develop
an approximate formula for the total scalar field.

6
2∆B • Be 1/2 ∆B • Be

|B| |Be | 1 + |Be | 1 + (4.5)
|Be |2 2
|Be |
The dipolar
Equation fieldbeofrearranged
4.5 can the earth is to
typically
relate 50,000 nT whilescalar
the measured the crustal anomalies
anomaly are only
A to the vector
anomaly ∆B
about 300 nT. Thus |ΔB| is small relative to the other terms and we can developBan
given an independent
2 measurement of the dipolar field of the Earth e.
approximate formula for the total scalar field. ∆B • Be

A= |B| − |Be | = (4.6)
|Be |
⎛ 2ΔB • B ⎞ 1 / 2 ⎛ ΔB• B ⎞
B ≅ Be ⎜⎜ 1+ e⎟
⎟ ≅ Be ⎜⎜1 +
(5) e⎟
4.5 Magnetic Anomalies Due to Seafloor Spreading Be ⎟⎠
2 2
⎝ Be ⎠ ⎝
Equation (5) can be re-arranged to relate the measured scalar anomaly A to the vector
To calculate the anomalous scalar field on the sea surface due to thin magnetic stripes
anomaly
on the ΔB given
seafloor, we goanback
independent measurement
Poisson’s of the dipolar
equation relating field field
magnetic of thetoearth Be.
magnetization.
The model is shown in Figure 4.4. We have an xyz coordinate system with z pointed
upward. The z = 0 level corresponds to sea level and there is a thin magnetized layer
at a depth of zo . ΔB • B
A = B − Be = e
(6)
Be
We define a scalar potential U and a magnetization vector M. The magnetic anomaly ∆B
is the negative
Magnetic gradient
Anomalies dueof
to the potential.
Seafloor The potential satisfies Laplace’s equation
Spreading
above the source layer and it satisfies Poisson’s equation within the source layer.
To calculate the anomalous scalar field on the sea surface due to thin magnetic stripes
∆B = equation
on the seafloor, we go back Poisson's −∇U relating magnetic field to magnetization.
(4.7)
The model is shown in the following diagram. We have an xyz co-ordinate system with z
2
∇ U=0 z , zo (4.8)
pointed upward. The z=0 level corresponds to sea level and there is a thin magnetized
layer at a depth of zo. ∇2 U = µo ∇ • M z = zo (4.9)
y
x
zo
p(x)
Figure 4.4
U(x, y, z) magnetic potential Tm
µo magnetic permeability 4π × 10−7 TA−1 m
M magnetization vector Am−1
In addition to assuming the layer is infinitesimally thin, we assume that the magneti-
zation direction is constant but that the magnetization varies in strength and polarity as
specified by the reversal function p(x). The approach to the solution is:
1. Solve the differential equation and calculate the magnetic potential U at z = 0.

2. Calculate the magnetic anomaly vector ∆B.
3. Calculate the scalar magnetic field A = (∆B • Be )/|Be | .
Let the magnetization be of the following general form.

M(x, y, z) = M x ı̂ + My ̂ + Mz k̂ p (x) δ(z − zo ) (4.10)
The differential equation 4.9 becomes

∂2 U ∂2 U ∂2 U ∂ ∂ XXX ∂
$ X %
+ + = µ o M x p (x) δ(z − z o ) + M y p (x) δ(z − zo ) + M z p (x) δ(z − zo )
∂x2 ∂y2 ∂z2 ∂x ∂y X ∂z
X XXX

(4.11)
The y-source term vanishes because the source does not vary in the y-direction (i.e., 8
the y derivative is zero). Thus the component of magnetization that is parallel to the
ridge axis does not produce any external magnetic potential or external magnetic field.
Consider a N-S
N-Soriented
orientedspreading
spreadingridge
ridgeat the magnetic
at the magneticequator. In this
equator. casecase
In this the TRM of the crust
the TRM
of the crusthas
hasa acomponent
componentparallel to the
parallel to dipole field field,
the dipole whichwhich
happens to be parallel
happens to the ridge axis
to be parallel
to the ridgesoaxis, so there will be no external magnetic
there will be no external magnetic field anomaly. field anomaly. See Figure 4.5.
This explains why the global map of magnetic anomaly picks [Cande et al., 1989] has no
Figure 4.5
data in the equatorial Atlantic or the equatorial Pacific where ridges are oriented N-S.
Now with the ridge-parallel component of magnetization gone, the differential equation
This explains why the global map of magnetic anomaly picks [7] has no data in the
reduces to.
equatorial Atlantic or the equatorial Pacific, where ridges are oriented N-S. Now with
∂ 2 U ∂ 2U ⎡∂ ∂ ⎤
+ = µo ⎢ Mx p(x)δ (z − zo ) + M z p(x)δ (z − zo )⎥ (12)
∂x 2 ∂z 2 ⎣ ∂x ∂z ⎦
This is a second order differential equation in 2 dimensions so 4 boundary conditions are

needed for a unique solution.
lim U(x ) = 0 and lim U( x) = 0 (13)

the ridge-parallel component of magnetization gone, the differential equation reduces to
∂2 U ∂2 U
∂ ∂

+ = µ o ∂x
M x p (x) δ(z − z o ) + ∂z
M z p (x) δ(z − zo ) (4.12)
∂x2 ∂z2
This is a second-order differential equation in two dimensions, so three boundary con-
ditions are needed for a unique solution:
lim U(x) = 0 and lim U(x) = 0 (4.13)

|x|→∞ |z|→∞
Take the 2-dimensional Fourier transform of the differential equation where the forward
and inverse transforms are defined as
Z∞ Z∞
F(k) = f (x)e−i2π(k·x) d2 x F(k) = =2 f (x)

−∞ −∞
(4.14)
Z∞ Z∞
f (x) = F(k)ei2π(k·x) d2 k f (x) = =−1
2 [F(k)]
−∞ −∞
where x = (x, z) is the position vector, k = (k x , kz ) is the wave number vector, and
(k • x) = k x x + kz z. The derivative property is =2 [dU/ dx] = i2πk x =2 [U]. The Fourier
transform of the differential equation is
h i
− (2πk x )2 + (2πkz )2 U(k x , kz ) = µo p(k x )e−i2πkz zo (i2πk • M) (4.15)
The Fourier transform in the z-direction was done using the following identity:
Z∞
δ (z − zo )e−i2πkz z dz ≡ e−i2πkz zo (4.16)
−∞
Now we can solve for U(k)

−iµ e−i2πkz zo
U(k) = 2πo p(k x )(k • M) 2 2 (4.17)
k x + kz
Next, take the inverse Fourier transform with respect to kz , using the Cauchy Residue
Theorem.
Z∞
µo (k • M)ei2πkz (z−zo )
U(k x , z) = 2πi p (k x ) dkz (4.18)
2 k x + kz
2
−∞
The poles of the integrand are found by factoring the denominator.
k2x + kz2 = (kz + ik x )(kz − ik x ) (4.19)

We see that U(k) is an analytic function with poles at ±ik x . The integral of this function
about any closed path in the complex kz plane is zero unless the contour includes a pole,
in which case the integral is i2π times the residue at the pole.
I 10
f (z)
dz = i2π f (zo ) (4.20)
z − zo
One possible path integral is shown in the following diagram.

One possible path integral is shown in Figure 4.6.
Im kz
i kx
Re kz
-ikx
There are two ways to close the path at infinity. The selection of the proper path, and
Figure 4.6
thus the residue, depends on the boundary condition, equation (13). First consider the
case where kx > 0. If we close the path of integration in the upper imaginary plane, then
There arewill
the pole twobeways
ikx. to The close the path
residue at infinity.
will have The selection
an exponential termof thevanished
that proper path, and to
as z goes
thus the residue, depends on the boundary condition, equation (4.13). First consider the
plus where
case infinity.
k x This
> 0. If is wewhat we the
close need since
path the observation
of integration in theplane
upperisimaginary
above the plane,
source.then
the pole will be-2πik . The residue will have an exponential term that vanished as z goes
e k x (xz- z0 )
∫ ()dk z = 2ikx (k x Mx + ikx Mz )
to plus infinity. This is what we need, since the observation plane is above(20) source.
the
e−2πkx (z−z0 )
I
() dkz = (k x M x + ik x Mz ) (4.21)
Next consider the case where kx < 0. To 2iksatisfy
x the boundary condition as z goes to plus
infinity, the -ikx pole should be used and the integration path will be clockwise instead of
Next consider the case where k x < 0. To satisfy the boundary condition as z goes to
counterclockwise as in (20).
plus infinity, the −ik x pole should be used, and the integration path will be clockwise
instead of counterclockwise, as in equation (4.20).
e +2π kx (z -z0 ) I e+2πkx (z−z0 )
∫ ()dk z = 2ikx ( kx M()x dk − ik
z =x Mz )
2ik x
(k x M x − ik x Mz ) (21) (4.22)
One can combine the two cases by using the absolute value of k x :
One can combine the two cases by using the absolute
value of kx
µ k
U(k, z) = 2o p (k) e−2π|k|(z−z0 ) Mz − i M x (4.23)
|k|
µ ⎛ k ⎞
U(k,z ) = o p(k)e -2π k (z- z0 ) ⎜⎜ Mz − i Mx ⎟⎟ (22)
2 ⎝ k ⎠
where we have dropped the subscript on the x-wavenumber.
This is the general case of an infinitely-long ridge. To further simplify the problem,
lets assume that this spreading ridge is located at the magnetic pole of the earth so the
dipolar field lines will be parallel to the z-axis and there will be no x-component of
magnetization.
CHAPTER 4. The result MAGNETIC
MARINE is. ANOMALIES 48
where we have dropped the subscript on the x-wavenumber.

µ M
U(k,z ) = o z p(k)e -2π k (z- z0 ) (23)
2
This is the general case of an infinitely long ridge. To further simplify the problem,
let’s assume that this spreading ridge is located at the magnetic pole of the Earth, so
the dipolar field lines will be parallel to the z-axis and there will be no x-component of
Next calculate the magnetic anomaly ΔB = -∇U.
magnetization. The result is
µo Mz
U(k, z) = p (k)e−2π|k|(z−zo ) (4.24)
ΔB = (-i2πk, 2π k )U(k,z )
2
(24)
Next calculate the magnetic anomaly ∆B = −∇U.

The scalar magnetic field is given by equation (6) and since only the z-component of the
∆B = (−i2πk, 2π |k|) U(k, z) (4.25)
earth's field is non-zero, the anomaly simplifies to.
The scalar magnetic field is given by equation (4.6). Since only the z-component of the
Earth’s field isµnon-zero, the anomaly simplifies to
o Mz
A(k, z) = p(k) 2π k e -2π k (z- z0 ) (25)
2 A(k, z) = µo Mz p (k) × 2π |k| e−2π|k|(z−zo ) (4.26)
observed = reversal x earth 2
anomaly pattern
observed filter Earth
reversal
anomaly pattern filter
The reversal pattern is a sequence of positive and negative polarities. To generate the
The reversal
model anomaly pattern is a sequence
one would take theoffourier
positive and negative
transform of thepolarities. To generate
reversal pattern, the by
multiply
model anomaly, one would take the Fourier transform of the reversal pattern, multiply
the
byearth filter filter,
the Earth and take
and the
takeinverse transform
the inverse of theofresult.
transform An examination
the result. of theofearth
An examination
the Earth
filter filterwhy
illustrates in Figure 4.7 illustrates
a square-wave whypattern
reversal a square-wave
becomes reversal
distorted.pattern becomes
distorted.
e-2π|k|z - upward
continuation
gain 2π|k| - derivative
|k|
Figure 4.7
This Earth filter attenuates both long and short wavelengths, so it acts like a band-pass
filter. In the space domain it modifies the shape of the square-wave reversal pattern, as
sketched in Figure 4.7.
12
This earth filter attenuates both long and short wavelengths so it acts like a band-pass
filter. In the space domain it modifies the shape of the square-wave reversal pattern as
sketched in4.theMARINE
CHAPTER following diagram.
MAGNETIC ANOMALIES 49
output (filtered)
A
x
input magnetization
When the seafloor spreading ridge is notFigure 4.8

at the magnetic pole, both the magnetization and
the earth's magnetic field will have an x-component. This introduces a phase shift or
skewness
When Θ, in thespreading
the seafloor output magnetic
ridge anomaly. At the
is not at the ocean surface
magnetic pole, the skewed
both magnetic
the magnetization
andanomaly
the Earth’s
is.
magnetic field will have an x-component. This introduces a phase
shift, or skewness Θ, in the output magnetic anomaly. At the ocean surface the skewed
magnetic anomaly is
k µ M
A(k) = o z p (k)eiΘ |k| 2π |k| e+2π|k|zo
k
µ M iΘ
(4.27)
A(k) = o z p(k)e k 2π k e +2π k z 2 0
(26)
2
The skewness depends on both the geomagnetic latitude and the orientation of the
The skewness depends on both the geomagnetic latitude and the orientation of the
spreading ridge when the crust was magnetized. Moreover, this parameter will vary
spreading
over time. Ifridge
onewhen
knowsthethe
crust was magnetized.
skewness, then theMoreover, this parameter
model profile can be will vary over
skewed to match
thetime. If one
observed knows Alternatively,
profile. the skewness then
the the model magnetic
observed profile cananomaly
be skewed
canto be
match the
de-skewed.
This
observed profile. Alternatively, the observed magnetic anomaly can be de-skewed. This have
is called reduction to the pole because it synthesizes the anomaly that would
formed on the magnetic pole.
is called reduction to the pole because it synthesizes the anomaly that would have formed
on the magnetic pole. k
A pole (k) = Aobserved (k)e−iΘ |k| (4.28)
k
− iΘ
k
4.6 Apole (k) = Aobserved (k )e
Discussion (27)
The ability to observe magnetic reversals from a magnetometer towed behind a ship
relies on some rather incredible coincidences related to reversal rate, spreading rate,
ocean depth, and Earth temperatures (mantle, seafloor, and Curie). In the case of ma-
rine magnetic anomalies, four scales must match.
First the temperature of the mantle (1200◦ C), the seafloor (0◦ C), and the Curie tem-
perature of basalt (∼500◦ C) must be just right for recording the direction of the Earth’s
magnetic field at the seafloor spreading ridge axis. Most of the thermoremnant mag-
netism (TRM) is recorded in the upper 1000 meters of the oceanic crust. If the thick-
ness of the TRM layer was too great, then as the plate cooled while it moved off the
spreading ridge axis, the positive and negative reversals would be juxtaposed in dipping
vertical layers (Figure 4.3). This superposition would smear the pattern observed by a
ship. If the seafloor temperature was above the Curie temperature, as it is on Venus,
then no recording would be possible.
The second scale is related to ocean depth and thus the Earth filter. The external mag-
netic field is the derivative of the magnetization which, as shown above, acts as a high-
pass filter applied to the reversal pattern recorded in the crust. The magnetic field
measured at the ocean surface will be naturally smooth (upward continuation) due to
the distance from the seafloor to the sea surface; this is a low-pass filter. This smooth-
ing depends exponentially on ocean depth so, for a wavelength of 2π times the mean
ocean depth, the field amplitude will be attenuated by 1/e, or 0.37, with respect to the
value measured at the seafloor. The combined result of the derivative and the upward
continuation is a band-pass filter with a peak response at a wavelength of 2π times the
mean ocean depth, or about 25 km. Wavelengths that are shorter (<10 km) or much
longer (>500 km) than this value will be undetectable at the ocean surface.
The third and fourth scales that must match are the reversal rate and the seafloor-
spreading rate. Half-spreading rates on the Earth vary from 10 to 80 km per million
years. Thus for the magnetic anomalies to be most visible on the ocean surface, the
reversal rate should be between 2.5 and 0.3 million years. It is astonishing that this
is the typical reversal rate observed in sequences of lava flows on land!! While most
ocean basins display clear reversal patterns, there was a period between 85 and 120
million years ago when the magnetic field polarity of the Earth remained positive, so
the ocean surface anomaly is too far from the reversal boundaries to provide timing
information. This area of seafloor is called the Cretaceous Quiet Zone; it is a problem
area for accurate plate reconstructions.
The lucky convergence of length and time scales makes it very unlikely that magnetic
anomalies due to crustal spreading will ever be observed on another planet.
4.7 Exercises
Exercise 4.1. Explain why magnetic lineations cannot be observed from a spacacraft
orbiting the Earth at an altitude of 400 km?
Exercise 4.2. Explain why scalar magnetic anomalies are not observed at a N-S ori-
ented spreading ridge located at the magnetic equator.
Exercise 4.3. Write a Matlab program to generate marine magnetic anomaly versus
distance from a spreading ridge axis. Use equation 4.27 relating the Fourier transform
of the magnetic anomaly to the Fourier transform of the magnetic timescale. You will
need a magnetic timescale and the start of a Matlab program
(ftp://topex.ucsd.edu/pub/class/geodynamics/hw3). Assume symmetric spreading about
the ridge axis, constant spreading rate, and constant ocean depth.
Use the program and magnetic anomaly profiles across the Pacific-Antarctic Rise
(NBP9707.xydm) and the Mid-Atlantic Ridge (a9321.xydm) to estimate the half-spreading
rate at each of these ridges. You may need to vary the mean ocean depth and skewness
to obtain good fits.
Describe some of the problems that you had fitting the data. Provide some estimates
on the range of total spreading rate for each ridge.
Chapter 5
Cooling of the Oceanic

Lithosphere and Ocean Floor
Topography
5.1 Introduction
This chapter uses the Fourier transform method to solve for the temperature in the
cooling oceanic lithosphere. For researchers in the areas of marine geology, marine
geophysics, and geodynamics, this is the most important concept you can learn from
this book. As noted in the original paper on the topic by Turcotte and Oxburgh [80],
convection of the mantle is primarily controlled by thin thermal boundary layers. The
surface thermal boundary layer, or oceanic lithosphere, is the most important compo-
nent of the convecting system because it represents the greatest temperature gradient
in the earth. It also has a greater surface area than the second most important ther-
mal boundary layer, which is at the core-mantle boundary. As the lithosphere cools it
becomes denser, the seafloor depth increases, and ultimately the lithosphere founders
(subduction). This subduction process both drives the convective flow and efficiently
quenches the mantle.
This chapter covers the same material as sections 4-15 and 4-16 in Turcotte and Schu-
bert [81]. The main difference is the method of solution. Turcotte and Schubert solve
the half-space cooling problem by guessing a similarity variable and then using this
variable to reduce the time-dependent heat conduction equation from a partial differ-
ential equation to an ordinary differential equation that can be solved by integration.
These notes provide an alternate solution to the problem by using the tools of Fourier
52
CHAPTER 5. COOLING OF THE OCEANIC LITHOSPHERE 53
analysis. Basically, any type of heat conduction problem can be solved with the2 Fourier
approach [9]. This Fourier approach is more than just a new way to solve an old prob-
The basic
lem. model
Many 3-Dis heat
shown in the following
conduction problemsdiagram
with that representssources
complicated one halfandof boundary
a
conditions do not have complete analytic solutions, but do
seafloor spreading system. The model assumptions and consequences are: have solutions in the Fourier
transform domain. In these cases, the FFT algorithms, coupled with modern comput-
 lithospheric plates are rigid and move away from the spreading ridge axis at a
ers, can be used to compute accurate results in seconds. Resorting to finite difference
uniform rate of v;
or other
 hot, numerical
low-viscosity schemes is fills
asthenosphere error-prone and the results are more difficult to interpret,
the void (passive);
 internal heat generation is much smaller than thethe
since the analytic foundation is gone. Thus Fourier
other terms approach
in the heatisequation
worth learning.
so it
is neglected;
 there is a singular
The basic model point at x in
is shown = Figure
z = 0. 5.1,
(We'll let the
which "ridge scientists"
represents one halfdeal
of a with this spread-
seafloor
issue.)
ing system.
v To
x
lithosphere
asthenosphere
Tm
Figure
This is a 2-dimensional problem with no heat 5.1 so the heat equation has only
sources
diffusive and advective terms
The model assumptions and consequences are:

! T ! 2 T v !T
2
+ = (1)
!x 2 !•z2 lithospheric
" !x plates are rigid and move away from the spreading ridge axis at a
uniform rate of v;
where T is temperature and κ is the thermal diffusivity. The first term represents the
• hot, low-viscosity asthenosphere fills the void (passive);
lateral diffusion of heat, the second term represents the vertical diffusion of heat, and the
• (on
third term internal heatside)
the right generation is much smaller
is the advection of heat than
by thethe other of
motion terms in the heat
the plate. Away equation,
so it is neglected;
from the ridge axis (x >> 0), one can show that the lateral heat diffusion is much smaller
than the •vertical
there heat
is a singular
diffusion.point at x =thisz term
Dropping = 0. simplifies
(This heattheisdifferential
released by hydrothermal
equation
circulation.)
This is a 2-dimensional problem with no heat sources, so the heat equation has only
diffusive and advective terms:
∂2 T ∂2 T v ∂T
+ 2 = (5.1)
∂x 2 ∂z κ ∂x
where T is temperature and κ is the thermal diffusivity. The first term represents the
lateral diffusion of heat, the second term represents the vertical diffusion of heat, and
the third term (on the right side) is the advection of heat by the motion of the plate.
Away from the ridge axis (x 0), one can show that the lateral heat diffusion is much
smaller than the vertical heat diffusion. Dropping this term simplifies the differential
equation, although a solution can also be developed where the term is retained. Next
we move from an Eulerian coordinate system to a Lagrangian system:
∂x ∂T ∂x ∂T
v= → = (5.2)
∂t ∂x ∂t ∂t
This reduces the problem to the half-space cooling problem:

∂2 T 1 ∂T
= (5.3)
∂z 2 κ ∂t
The boundary and initial conditions are:

T (0, t) = T o
T (∞, t) = T m (5.4)
T (z, 0) = T m
The infinite half-space has constant thermal diffusivity and an initially constant tem-
perature T m . At times greater than zero, the surface temperature is reduced to T o . The
temperature will evolve with time. Note that for this problem, time also corresponds to
the age of the cooling oceanic lithosphere. Define a dimentionless temperature as
T −T
θ = T − To (5.5)
m o
Now the differential equation and boundary conditions become

∂2 θ 1 ∂θ
=
∂z2 κ ∂t
θ(0, t) = 0 (5.6)
θ(∞, t) = 1
θ(z, 0) = 1
Turcotte and Schubert [81, page 184] introduce the following dimensionless quantity
and use this to reduce this to an ordinary differential equation with two boundary con-
ditions. They then integrate the differential equation twice and match the boundary
conditions:
z
η= √ (5.7)
2 κt
Suppose one did not know this trick or the problem was more complicated. An ap-
proach called method of images is straightforward. The model is expanded to a full-
space with an initial step-function temperature distribution, so the 0-temperature bound-
ary condition is always matched. The problem becomes
∂2 θ 1 ∂θ
= κ
∂z2 ∂t
(5.8)
θ(∞, t) = 1
θ(z, 0) = 2H(z) − 1
where the definition of the step function is
Zz
H(z) ≡ δ(z) dz (5.9)
−∞
Now take the Fourier transform of equation (5.8) with respect to z. The differential
equation becomes
∂Θ
− κ(2πk)2 Θ(k, t) = ∂t (5.10)
The general solution is

2
Θ(k, t) = Co e−κ(2πk) t (5.11)
Now take the Fourier transform of the initial condition
= Θ(k,0) = = 2H(z) − = [1]

(5.12)
We know that
= [1] = δ(k). (5.13)
Also, using the derivative property we know that
∂H
" #
= = i2πk =[H(z)] (5.14)
∂z
Since the derivative of the step function is a delta function, the Fourier transform of the
initial condition is
1
Θ(k,t) = iπk − δ(k) (5.15)
The solution that satisfies the initial condition is

1 2
Θ(k, t) = iπk − δ(k) e−κ(2πk) t (5.16)
Now we take the inverse Fourier transform

Z∞ 2 Z∞
e−κ(2πk) t i2πkz 2
θ(z, t) = e dk − δ(k) e−κ(2πk) t ei2πkz dk (5.17)
iπk
−∞ −∞
The second integral on the right side of equation (5.17) is equal to 1 since the delta
function extracts the integrand at k = 0. The first integral on the right side of equa-
tion (5.17) is performed in two steps. First take the derivative with respect to z to
note that
Z∞
∂θ(z, t) 2
= 2 e−κ(2πk) t ei2πkz dk (5.18)
∂z
−∞
This is the Fourier transform of a Gaussian function. The following substitution puts
the integral in the form that appears in [5]:
√ z
k0 = k 4πκt and z0 = √ (5.19)
4πκt
The result is
∂θ(z, t) 2 −z2
= √ e 4κt (5.20)
∂z 4πκt
Next integrate equation (5.20) over z. The introduction of the similarity variable based
on equation (5.20) helps to identify the integral as the definition of the error function:
z √
η= √ so dz = 2 κt dη (5.21)
2 κt
The integral becomes
Zη
2 2
θ(z, t) = √ e−η dη − 1 (5.22)
π
∞
The right side of equation (5.22) is just the definition of the error function erf(η). The
final solution is
z
T (z, t) = (T m − T o )er f √ + T o (5.23)
2 κt
5.2 Temperature versus Depth and Age
The thermal parameters and temperatures appropriate to the Earth are given in Table 5.1.
Parameter Definition Value
To surface temperature 0◦ C
T1 temperature at base 1100◦ C
of thermal boundary
layer
Tm mantle temperature 1300◦ C
κ thermal diffusivity 8 × 10−7 m2 s−1
k thermal conductivity 3.3 W m−1 C−1
Table 5.1
If we define the base of the thermal boundary layer as some large fraction of the deep
mantle temperature, as in the table, one can calculate the thickness of the thermal
boundary layer versus the age of the lithosphere.

Tl − To z
= 0.84 = erf √ (5.24)
T −T
m o 2 κt
or √
z 2 κt
p
or z(km) 10 age(Ma) (5.25)
The isotherms for this model are displayed in Figure 5.2.
Figure 5.2: The solid lines are isotherms, T − T o (◦ K), in the oceanic lithosphere from
equation (5.4). The data points are the thickness of the oceanic lithosphere in the Pacific
determined from studies of Rayleigh wave dispersion data. (From [38].)
5.3 Heat Flow versus Age
The heat flow is the thermal conductivity times the temperature gradient:
∂T
q(z) = k ∂z (5.26)
To calculate the heat flow we take the derivative of the error function with respect to z:
∂ erf(η) ∂ erf(η) ∂ η 1 2
= = √ e−η (5.27)
∂z ∂η ∂z πκt
k(T m − T o ) −z2
q(z, t) = √ e 4κt (5.28)
πκt
In the limit as depth z goes to infinity, the heat flow is zero. So for this model, there is
no heat transport into the base of the lithosphere. Later we’ll compute seafloor depth
versus age for this model and show that there are large deviations at old age (i.e.,
>70 Ma). One way to flatten the depth-versus-age curve is to supply heat to the base
of the lithosphere. There are a variety of ways to accomplish this:
• Increasing basal heat flux with age corresponds to the plate cooling model of
[54]. The physical mechanism for this basal heat input is small-scale convective
rolls beneath the old lithosphere.
• A constant basal heat flux with age corresponds to the CHABLIS cooling model
of [17].
• Some papers (e.g., Crough, 1983 [13]) propose that mantle plumes reheat the old
lithosphere and eventually all old lithosphere encounters one or more plumes, so
reheating is pervasive.
The surface heat flow is just equation (5.28) evaluated at the surface of the earth.
k(T m − T o )
q(t) = √ (5.29)
πκt
The match to the observed heat flow is shown in Figure 5.3. For ages less than about
40 Ma, the surface heat flux is less than predicted by the model. This heat flow deficit
occurs because cold seawater circulates deep into the crust and advects the heat. So
the temperature gradient will be less than predicted by a purely conductive model. At
older ages, the heat flow is higher than expected. This could either be due to a non-zero
basal heat flux or an incorrect estimate of thermal conductivity of the crust.
Figure 5.3: Mean values and standard deviations of ocean floor heat flow measurements
as functions of age compared with equation (5.29). Data from [71] .
5.4 Thermal Subsidence
As the oceanic lithosphere cools by conductive heat loss, it contracts. This thermal
contraction causes the average density of the lithosphere to increase. The seafloor10
depth increases with age and eventually the lithosphere becomes so dense it founders
atThermal Subsidence
a subduction zone. To develop a linear relationship between density and tempera-
ture, consider a cube of volume V, mass m, and density ρ, at temperature T o under a
As the oceanic lithosphere cools by conductive heat loss, it contracts. This thermal
confining pressure Po . (See Figure 5.4.)
contraction causes the average density of the lithosphere to increase. The seafloor depth
Changes
increasesinwith
bothage
temperature and pressure
and eventually will produce
the lithosphere becomes changes in the
so dense volume at
it founders of a
the cube:
subduction zone. To develop a linear
∂V relationship∂V between density and temperature,
! !
dV = dT + dP (5.30)
∂T
consider a cube of volume V, mass m, andPo density∂Pρ, at
T 0 temperature To under a confining
pressure Po.
V=m/ρ
Changes in both temperature and pressure will produce changes in the volume of the
Figure 5.4
cube.
" !V % " !V %
dV = $ ' dT + $ ' dP (30)
# !T & Po # !P & To
The two terns in (30) are related to the volumetric coefficient of thermal
1 # "V & 1 $ #V '
expansion ! = % ( and the isothermal compressibility is ! = " & ) . Since
V $ "T ' Po V % #P ( To
!" !V
The two terms in equation (5.30) are related to the volumetric coefficient of thermal
expansion:
1 ∂V
!
α= V
∂T Po
and the isothermal compressibility is
∂V
!
1
β = −V .
∂P To
Since ρ = mV −1 it is easy to show that

∂ρ ∂V
=− ,
ρ V
so the coefficient of thermal expansion becomes
1 ∂ρ
!
α = −ρ .
∂T Po
We are considering the lithosphere that slides laterally across the surface of the earth,
so there are no significant pressure variations. Thus we need only the first term in
equation (5.30). If ρm is the density of the lithosphere at a temperature of T m , then a
reduction in temperature will cause an increase in density.
h i
ρ T = ρm 1 − α T − T m

(5.31)
The diagram in Figure 5.5 illustrates the thermal subsidence of the oceanic lithosphere
11
t=x/v
water ρw
lithosphere ρ= ρm[1-α(T-Tm)]
depth of compensation ρm
Figure 5.5 of the oceanic lithosphere as it spreads

The diagram illustrates the thermal subsidence
from the ridge axis at a velocity of v. There are three layers in the model. The ocean has
as it spreads from the ridge axis at a velocity of v. There are three layers in the model.
a density of ρw and a depth of d0 at the ridge axis. This depth increases with age/distance
from the ridge axis. We will use the principles of thermal contraction and isostasy to
determine the increase in seafloor depth with increasing age d(t). The density of the
lithosphere depends on temperature according to equation (31). The asthenosphere
behaves as a fluid on geological timescales so the lithosphere floats on the mantle.
The major assumptions are:
 The pressure at the depth of compensation is a constant value and depends only on
the weight of the rock and water directly above (i.e., isostatic equilibrium).
The ocean has a density of ρw and a depth of d0 at the ridge axis. This depth increases
with age/distance from the ridge axis. We will use the principles of thermal contrac-
tion and isostasy to determine the increase in seafloor depth with increasing age d(t).
The density of the lithosphere depends on temperature, according to equation (5.31).
The asthenosphere behaves as a fluid on geological timescales so the lithosphere floats
on the mantle.
The major assumptions are:
• The pressure at the depth of compensation is a constant value and depends only
on the weight of the rock and water directly above (i.e., isostatic equilibrium).
• The crust has uniform thickness, so it has no effect on the overall isostatic bal-
ance.
• The thermal diffusivity κ is isotropic and independent of P and T .
• The thermal expansion coefficient α is isotropic and independent of P and T .
• Heat is transferred by conduction, so hydrothermal circulation is not important.
This is a poor assumption at the ridge axis.
• Heat conducts only vertically. This is also a poor assumption at the ridge axis.
• There are no heat sources in the crust or lithosphere.
• No heat flows into the base of the lithosphere See [17] for a discussion of alter-
nate models with basal heat input.
An additional assumption is that the lithosphere is free to contract in all three dimen-
sions. Since the lithosphere is thin in relation to its horizontal dimension, free con-
traction in the vertical dimension is a good assumption. Contraction of the plate in
the direction perpendicular to the ridge axis is probably valid as well. However, con-
traction in the ridge-parallel direction will produce significant shear strain, which will
result in thermoelastic stress. We will neglect this for now but this is an interesting area
of research.
As the lithosphere cools and contracts, its vertically-integrated density increases which
will increase the pressure at its base. To maintain isostatic balance (i.e., constant pres-
sure at constant depth zl ), ocean depth must increase to replace high density rock with
lower density water. The increase in depth is determined by the following isostatic
balance between a ridge-axis column and an off-axis column. See Figure 5.6.
The mathematical statement of isostatic balance is

Zzl Zd Zzl h
i
g ρm dz = g ρw dz + g ρm 1 − α T − T m dz (5.32)
o o d
thermoelastic stress. We will neglect this for now but this is an interesting area of
research.
As the lithosphere cools and contracts, its vertically-integrated density increases which
will increase the pressure at its base. To maintain isostatic balance (i.e., constant pressure
at constant depth zl), ocean depth must increase to replace high density rock with lower
density water. The increase in depth is determined by the following isostatic balance
between a ridge-axis column and an off-axis column.
do
d(t) ρw
ρm ρm[1-α(T-Tm)]
zl
The mathematical statement of isostatic balance is Figure 5.6
zl d zl
g " where
! m dz = gg"is
! wthe
dz +acceleration
g" ! m [1 # $ (T #ofTmgravity.
)]dz (32)
o o d
where g is theAfter
acceleration of gravity.
subtracting the standard ridge-axis column from both sides and dividing through
by g we get:
After subtracting the standard ridge-axis column from both sides and dividing through by
Zd Zzl
g we get.
0= (ρw − ρm ) dz − ρm α T − T m dz.

(5.33)
d zl
o d
0 = # (we’ll
Now m )dz the
! w " !use m$ (T " Tmto
" # !solution )dz.the half-space cooling (33)
problem (equation (5.23)) to define
T (t,oz). Note thisd solution has temperature perturbations at infinite depth, so we must
extend the depth integration from the seafloor to infinity.
Z∞
z−d
d ρm − ρw = ρm α T m − T o

1 − erf √ dz (5.34)
2 κt
d
By setting z0 = z − d and solving for d(t), we find

Z∞
ρm α T m − T o
!
z
d(t) = erfc √ dz (5.35)
(ρm − ρw ) 2 κt
o
√ √
To integrate this function let η = z 2 κt , so dz = 2 κt dη:
Z∞
2ρm α T m − T o √

d(t) = κt erfc (η) dη (5.36)
(ρm − ρw )
o
After performing the definite integral of

Z∞
1
erfc (η) dη = √
π
o
and adding the ridge axis depth do , we find depth depends on material constants times
the square root of seafloor age:
2ρ α (T − T ) κt 1/2

dtot (t) = do + m(ρ −mρ ) o (5.37)
m w π
Now let’s plug in some numbers to get an estimate of how seafloor depth varies with
age (Table 5.2).
Parameter Definition Value
To surface temperature 0◦ C
Tm mantle temperature 1300◦ C
k thermal conductivity 3.3W m−1 C−1
α thermal expansion coefficient 3.1 × 10−5 C−1
ρw seawater density 1025 kg m−3
ρm mantle density 3300 kg m−3
do ridge axis depth 2500 m
Table 5.2
A good approximation for the depth-age relation is
d = 2500m + 350 age(Ma)

p
(5.38)
To test this model of the cooling oceanic lithosphere we need seafloor depth, seafloor
age, and sediment thickness [58].
5.5 Buoyancy of the Cooling Lithosphere
Cooling and contraction of the lithosphere causes seafloor depth to increase with age.
The average density of cooled mantle lithosphere is greater than the density of the un-
derlying mantle so it would founder and subduct if it could be decoupled from more
buoyant surrounding lithosphere (e.g., continents). Once subduction begins, the nega-
tive buoyancy of the subducted slab pulls the surface plate into the mantle driving plate
tectonics. The amount of slab pull is related to the average density of the lithosphere
relative to the mantle so older lithosphere will have greater slab pull. This simple re-
lationship needs some modification because the oceanic crust, that is bonded to the
lithosphere, is less dense (compositional buoyancy) than the underlying mantle so very
young lithosphere will be positively buoyant. An important question is how long does
the oceanic lithosphere need to cool before the negative thermal buoyancy exceeds the
compositional buoyancy. Oxburgh and Parmentier [50] have addressed this issue and
have estimated the total buoyancy of oceanic lithosphere.
The compositional buoyancy has two components. Mantle upwelling beneath ridges
undergoes decompression melting at a depth of about 40 km. This melt migrates to the
magma chamber at the ridge axis where it forms oceanic crust with a normal thickness
of 6-7 km and an average density of 2800 km m−3 . In addition, ultramafic residues
formed by partial melting during the generation of basalt are less dense that undepleted
mantle. This zone of depleted mantle has a thickness of about 8 km and an average
density ρd of 3140 kg m−3 , which is less than the normal density of peridotite mantle
of 3300 kg m−3 . So this adds an additional buoyance to the oceanic lithosphere.
Oxburgh and Parmentier [50] have made a quantitative assessment of the overall buoy-
ancy of the lithosphere and define a parameter called the density defect thickness δ as
Z∞ "
ρm − ρ (z)
#
δ= dz (5.39)
ρm
0
where ρ(z) is the density of the lithosphere including the crust, depleted mantle, and
cooled lithosphere and ρm is the normal mantle density. The density defect thickness
has units of length such that when δ < 0 the entire package is negatively buoyant and
can subduct while when δ > 0 then the package is positively buoyant and will resist
subduction. They further divide the density defect thickness into a compositional and
a thermal component
δtotal = δcomp + δthermal . (5.40)
The compositional part has a contribution from the crust 0.91km = 6km × (ρm − ρc )/ρm
and a contribution from the depleted mantle 0.39km = 8km × (ρm − ρd )/ρm ) for a total
of δcomp = 1.3km. They also point out that areas having thickened crust such as oceanic
plateaus will have proportionally larger buoyancy so that a 20 km thick crust will have
a buoyancy of δcomp = 4.3km
The thermal buoyancy depends on the density of the cooled lithosphere given by equa-
tion (5.31) ρ (T ) = ρm [1 − α (T − T m )]. Inserting this into (5.39) and simplifying one
finds
Z∞ !
z
δthermal = −2α (T m − T o ) er f c √ dz (5.41)
2 κt
0
We performed a similar integration of equation (5.35). The result is
κt
r
δthermal = −2α (T m − T o ) (5.42)
π
The total density defect thickness decreases with increasing age and eventually the
thermal buoyancy dominates. An example is shown in Figure 5.7. For a normal crustal
thickness of 6 km, the lithosphere is positively buoyant between 0 and 30 Ma and neg-
atively buoyant for greater ages. Lithosphere with thicker crust such as that associated
with oceanic plateaus remains positively buoyant longer time so that, for example, an
18-km thick oceanic plateau will resist subduction at even the oldest ages found in the
ocean basins of 200 Ma. In addition to crustal thickness, the buoyancy depends on the
temperature difference across the lithosphere or thermal boundary layer. Lithosphere
on the planet Venus will be more buoyant than comparable lithosphere on Earth be-
cause Venus has a higher surface temperature of 455◦ C. The calculation of buoyancy
versus cooling time for Venus is left as an exercise below.
no subduction
4
2 24 km
δ (km)
18 km
0
12 km
-2 6 km
subduction
-4
-6
0 100 200 300 400 500 600
cooling time (Ma)
Figure 5.7: Total density defect thickness versus the age (or cooling time) of the litho-
sphere. Normal oceanic crust is 6 km thick and the lithosphere becomes negatively
buoyant after cooling for 30 Ma. For a crustal thickness of 12 km the time until neg-
ative buoyancy is delayed until 120 Ma. Lithosphere having a 24 km thick crust will
remain buoyant for 500 Ma. Moreover, the lower lithosphere may be reheated or de-
laminate during this long time so it may never subduct.
5.6 Exercises
Exercise 5.1. (a) What two measurements must be made to determine the conductive
heat flow at the bottom of the ocean? (b) Why is it OK to measure heat flow in the
upper few meters of sediment on the seafloor while one needs a borehole hundreds of
meters deep obtain a reliable measure of heat flow on the continents.
Exercise 5.2. Assume the lithosphere of Venus has evolved to a steady-state tempera-
ture profile and there is no heat generated in the lithosphere. Given a current heat flow
of 4x10-2 W m-2 , a surface temperature of 450 C, a mantle temperature of 1500 C and
a thermal conductivity of 3.3 W m-1 C-1 calculate the thickness of the lithosphere.
Exercise 5.3. Solve for the temperature T as a function of time t and depth z in a cool-
ing half space. The differential equation for heat diffusion is and the boundary/initial
conditions are
∂2 T 1 ∂T
=
∂z 2 κ ∂t
T (0, t) = T o (5.43)
T (∞, t) = T m
T (z, 0) = T m .
Use the following similarity variable η = 2 √z κt to reduce the partial differential equation
to an ordinary differential equation where κ is the thermal diffusivity, T o is the surface
temperature, and T m is the initial temperature of the half space.
Exercise 5.4. Derive the following relationship between the rate of increase in seafloor
depth with age ∂d
∂t and the difference between the surface and basal heat flow
(q s − qL ).
∂d α
= (q s − qL ) (5.44)
∂t C p (ρm − ρw )
You will need Fourier’s law, energy conservation, and isostasy as follows:
∂T
q=k
∂z
∂T k ∂2 T
=
∂t ρmC p ∂z2 (5.45)
ZL
−αρm
d(t) = T dz
(ρm − ρw )
o
where:
L - asymptotic lithospheric thickness and also the depth of compensation (m)

d - seafloor depth (m)
q - heat flow (W m -2 )
α - coefficient of thermal expansion (C-1 )
C p - heat capacity (J kg-1 )
ρw - seawater density (kg m-3 )
ρm - mantle density (kg m-3 )
k - thermal conductivity (W m-1 C-1 )
Exercise 5.5. Calculate the cooling time for lithospheric subduction on Venus for
crustal thicknesses of 16 km and 24 km. The surface temperature of Venus is 455
◦
C and use a deep mantle temperature of 1400◦ C. Use Earth-like values of thermal
expansion coefficient and thermal diffusivity in Table 5.2.
Chapter 6
Crustal Structure, Isostasy,

Swell Push, and Rheology
6.1 Introduction
This chapter covers four broad topics. First the basic structure of the oceanic and
continental crust is provided. The emphasis is on layer thickness and densities, and
there is little discussion of composition. The second and third topics are the vertical
and horizontal force balances due to variation in crustal thickness. The vertical force
balance, isostasy, provides a remarkably accurate description of variations in crustal
thickness on the Earth based on a knowledge of the topography. The horizontal force
balance provides a lower bound on the force needed to maintain topographic variations
on the Earth. The basic question is: What keeps mountain ranges from spreading
laterally under their own weight?
The fourth topic is the rheology of the lithosphere. How does the lithosphere strain in
response to applied deviatoric stress? We will see that the uppermost part of the litho-
sphere is cold, so frictional sliding along faults governs the strength. At greater depth,
the rocks can yield by non-linear flow mechanisms. The overall strength-versus-depth
profile is called the yield-strength envelope. The integrated yield strength transmits the
global plate tectonic stress field. Moreover, the driving forces of plate tectonics cannot
exceed the integrated lithospheric strength. This provides an important constraint on
the geodynamics of oceans and continents.
The yield-strength-envelope formulation will also be used in the chapter on flexure.

It provides an explanation for the increase in the thickness of the elastic layer as the
lithosphere ages and cools. In addition, it is used to understand the depth of oceanic
68
CHAPTER 6. CRUSTAL STRUCTURE 69
trenches. The first moment of the yield strength versus depth provides an upper bound
on the magnitude of the bending moment that the lithosphere can maintain. This model
strength estimate can be checked by measuring the bending moment of the trench/outer
rise topography. There is remarkably good agreement, assuming mantle dynamics does
not play a large role in the support of subduction zone topography. Throughout these
two chapters, the deviatoric stress levels are typically 100–300 MPa, since this is the
level of stress needed to maintain the topographic features on the Earth. In the follow-
ing chapters on earthquakes, the magnitudes of stresses are typically less than 10 MPa.
Part of this order-of-magnitude stress reduction is due to the fact that many earthquakes
are shallow. However, there is still a major unresolved issue of why the crust appears
weak in regard to earthquakes and strong in regard to topography and flexure. Since
this is a major unresolved issue, it is perhaps a good topic for research.
6.2 Oceanic Crustal Structure
The basic structure of the oceanic crust has been established through seismic refrac-
tion and reflection experiments, seafloor dredging/drilling, gravity/magnetics model-
ing, and studies of ophiolites [59]. Figure 6.1 from [36] illustrates the current model of
crustal generation at a fast spreading ridge. Partial melt that forms by pressure release
in the uppermost mantle (∼40 km depth) percolates to a depth of about 2000 m beneath
the ridge, where it accumulates to form a thin magma lens. Beneath the lens is a mush-
zone which develops into a 3500 m thick gabbro layer by some complicated ductile
flow. Above the lens, sheeted dikes (∼1400 m thick) are injected into the widening
crack at the ridge axis. Part of this volcanism is extruded into the seafloor as pillow
basalts. The pillow basalts and sheeted dikes cool rapidly as cool seawater percolates to
a depth of at least 2000 m. This process forms the basic crustal layers seen by reflection
and refraction seismology methods.
See Table 6.1.
Some key points about the oceanic crust:
• The Mohorovicic Discontinunity (Moho) is defined as the seismic velocity jump

from 6.8 km s−1 to greater than 8 km s−1 .
• The Moho corresponds to a change in composition and density. For thick conti-
nental crust, it also corresponds to a change in strength.
• The oceanic crustal thickness is remarkably uniform throughout the ocean basins.
While the average crustal thickness does not depend on spreading rate, the local
variations in thickness are greater for the slow-spreading crust (<70 mm/yr full
rate). For example, the crust is generally thinner near the end of a spreading
segment and thicker toward the center of the spreading segment.
Figure 6.1: Model of crustal structure derived from reflection and refraction seismol-
ogy [36].
Table 6.1
6.3 Continental Crustal Structure
The basic structure of the continental crust has also been established through seismic
refraction and reflection experiments, sampling and drilling, gravity/magnetics model-
ing and studies of exposed crustal sequences. Of course, the continental crust is highly
variable in thickness, velocity, density and composition. Table 6.2 represents an aver-
age crustal model for elevations close to sea level.
Table 6.2
6.4 Vertical Force Balance - Isostasy
As discussed in Chapter 1, one of the most important and defining features of the Earth
is the bimodal histogram of topography (Figure 1.2). The tallest peak in the histogram
represents continental crust having elevations close to sea level. A second broader
peak represents the oceanic crust having elevations between about -6000 m and -3000
m with a median depth of -4093 m. This bimodal histogram can be largely explained
by simple Airy isostasy where the lithostatic pressure at the base of the lithosphere is
constant over the Earth.
Figure 6.2: Airy compensation model used to determine crustal thickness and crustal
volume. The uniform density crust (ρc = 2800 kg m-3 ) is divided into 5 layers which
float on the higher density mantle (ρm = 3200 kg m-3 ). The seawater has a density ρw , of
1025 kg/m3. Layer 1 lies above sea level, layer 2 lies between sea level and the normal
seafloor depth (base depth) zb , layer 3 corresponds to the thickness of normal oceanic
crust, layer 4 is the compensating root for layer 2 and layer 5 is the compensating root
for layer 1.
A diagram of the Airy compensation model is shown in Figure 6.2 [70]. A uniform
density crust is divided into 5 layers, which float on the higher density mantle. An
major assumption of this model is that the mantle beneath the crustal plateau has no
lateral density variations. To understand how the thickness of each layer is calculated,
first consider a plateau on the seafloor where all of the topography lies below sea level.
By definition, the thicknesses of layers 1 and 5 are zero. The topography of the plateau
above the normal seafloor depth (h2 ) is isostatically compensated by a crustal root with
a thickness of h4 . Isostatic balance means the pressure at the base of the crustal root is
the same as the pressure at the same depth beneath normal oceanic crust. The balance
is
g (h2 ρc + h3 ρc + h4 ρc ) = g (h2 ρw + h3 ρc + h4 ρm ) . (6.1)
This can be simplified to an equation where the root thickness is related to the elevation.
(ρc − ρw )
h4 = h2 (6.2)
(ρm − ρc )
The densities determine the ratio of root thickness to elevation and for typical values
provided in Figure 6.2, this ratio is 4.4. The overall crustal thickness is the sum of the
three layers
(ρc − ρw )
" #
ht = h3 + h2 1 + . (6.3)
(ρm − ρc )
Exercise 6.1. (a) Develop a formula for the total crustal thickness when the top of
the plateau is at sea level. Use the densities provided in Figure 6.2, a normal oceanic
crustal thickness of 6.5 km, and a normal seafloor depth of 4.1 km.
(b) Calculate the thickness of this plateau at sea level.
(c) How does this compare with a typical value for continental crustal thickness at sea
level?
This simple Airy isostasy model provides a first order explanation for spatial variations
in crustal thickness observed using seismic refraction measurements in both the oceans
(Figure 6.3) and continents (Figure 6.4). For a continent elevation of 5 km such as the
Tibetan plateau, this simple model predicts the crust is 70 km thick which is in fair
agreement with the seismic measurements.
The scatter in the seismic thickness with respect to the estimates from Airy isostasy
(Figure 6.4) can be causes by three factors. First there may be significant lateral varia-
tions in crustal density (i.e., Pratt compensation). Second, there may be density varia-
tions in the mantle lithosphere caused by thermal or compositional variations. The most
prominent example is the elevation of the seafloor spreading ridges (-2500 m) above
the average basin depth (-4100 m) caused by thermal isostasy of the cooling oceanic
lithosphere (Chapter 5). Finally the assumption of constant pressure at the base of the
lithosphere could be incorrect because of pressure variations due to mantle flow (i.e.,
dynamic topography). While these processes are important, Airy isostasy provides a
first order explanation for the bimodal elevation of the Earth.
Figure 6.3: Seismic refraction measurements (30) of crustal thickness versus eleva-
tion above base depth for two continental submarine plateaus (Lord Howe Rise and
Mascarene Plateau) and two oceanic plateaus (Ontong-Java and Shatsky Rise) [70].
Predictions of the Airy compensation model are shown for crustal densities ranging
from 2700 to 2900 kg/m3 and a mantle density of 3200 kg/m3.
Figure 6.4: Seismic refraction measurements (1350) of crustal thick-ness versus eleva-
tion [70]. Measurements were selected from continental areas. Antarctica and Green-
land data were not used. Best-fitting Airy compensation model (solid curve) has a base
depth crustal thickness of 20 km. The RMS scatter about the model is 9.09 km.
g -∞
Now take the inverse fourier transform of (14).

o
2πG
N(x) = ∫ Δρ(x,z )zdz (15)
g -∞
This is a remarkable results because this formula provides a way to compute the geoid height
simply by integrating the density anomaly only over depth. This integration can be done for a
variety of models including Airy compensation, Pratt compensation, and thermal compensation
(i.e., spreading ridge or thermal swell). Several of these integrals are done in Geodynamics
[2002]. 6. CRUSTAL STRUCTURE
CHAPTER 75
Isostatic Geoid and the Swell-Push Force
6.5 Horizontal Force Balance - Swell Push Force
The three driving forces of plate tectonics are slab pull, asthenospheric drag, and ridge push.
Ridge-push force is the outward force due to isostatically-compensated topography. This mainly
occurs as a gravitational sliding force on the flanks of the seafloor spreading ridges. However, it
is also associated with the outward force due to thick continental crust or a thermal swell. One
One important question
of your homework that arises
problems is whata general
was to develop is the magnitude
expression forofthis
the stress inforce
swell-push thefor
litho-
sphereisostatically-compensated topography.large-scale
that is needed to maintain Here I'll derive this expression
plateaus and rootsagain[24].
and point out that
This can be
the force integral is identical to the integral for computing isostatic geoid anomaly. Therefore
considered as a minumum deviatoric stress and it places constraints on the long-term
the two are related and one can use geoid height to measure ridge-push force [Parsons and
€ strength
Richter, the
of 1980;crust, especially
Dahlen, theand
1981; Fleitout lower crust. 1982;
Froidevaux, Consider
1983]. isostatically-compensated
topography as shown in Figure 6.5.
Consider isostatically-compensated topography as shown in the diagram below.
(1) (2) P(z) ΔP(z)
ρc
(2) (1)
ρm
-L
Figure 6.5
While this diagram is related to a specific Airy-type compensation mechanism, the

integral relation, presented next, is quite general. To calculate the total outward force
F s due to this isostatically compensated plateau, we integrate the difference in pressure
between column (1) and column (2) over depth, to the depth of compensation −L where
the pressure difference is zero.
Z0
Fs = ∆P(z) dz (6.4)
−L
Integrate by parts
0
∂∆P(z)
0 Z
F s = ∆P(z)z−L − z dz. (6.5)
∂z
−L
Note the first term on the right is zero because of isostasy. The second term can be writ-
ten in terms of the density by noting that the vertical gradient in the pressure difference
is
∂∆P(z)
∂z
= −g∆ρ(z). (6.6)
The result is
Z0
Fs = g ∆ρ(z)z dz. (6.7)
−L
The swell push force depends on two factors: (1) the magnitude of the depth-integrated
surface density contrast which is equal and opposite to the magnitude of the depth-
integrated compensation and (2) the distance between the surface and compensating
density. A larger distance between the topography and it compensation increases the
swell push force.
Swell push force can be computed for a variety of isostatic configurations; we consider
three here. The first case is the calculation of the average crustal stress needed to
maintain the elevation of Tibet with respect to the elevation of India. This is left as an
exercise 6.2; however I provide the answer of 98 MPa to entice you to do the problem.
The important conclusion is that the lower crust must have a strength greater than
98 MPa or the high elevation of Tibet would rapidly collapse by lateral spreading.
Of course, the ongoing collision of India with Asia helps to provide some dynamic
support. Nevertheless, there are mountainous areas not having dynamic support that
can maintain their high elevation so the lower crust must be quite strong.
The second case is the calculation of the minimum ice strength needed to maintain the
configuration of a floating ice shelf. This is also left as an exercise 6.3. A prominent
example of ice sheet failure is the collapse of the Larsen B ice shelf in 2002. Prior to
collapse, ice shelf was size of Rhode Island. The entire collapse occurred in a couple
of months leaving behind thousands of large ice fragments. A team of collaborating
investigators have developed a theory of why the ice disintegrates. The theory is based
on the presence of ponded melt water on the ice shelf surface in late summer as the
climate has warmed in the area. Meltwater acts to enhance fracturing of the shelf thus
reducing its cohesive strength.
The third case is the so-called ridge push force. The name is a bit confusing because
the force is zero at the ridge axis and increases with the age of the lithosphere. It
is really a gravitational sliding force and is commonly termed gravitational potential
energy or GPE. In plate cooling case, the surface topography increases as the square
root of age and the depth of compensation also increases as the square root of age so
the ridge push force increases linearly with age. To calculate the ridge push force we
first construct a density structure following the development in section 5.4 on thermal
subsidence. Figure 6.6 shows this density structure where the temperature from the
cooling half space model (equation 5.23) is

z
T (z, t) = (T m − T o )er f √ + To. (6.8)
2 κt
The depth versus age is provided in equation 5.37 and is
κt 1/2
2ρm α (T m − T o )
d(t) = . (6.9)
(ρm − ρw ) π
The swell push force is

t
∆ρ = (ρw − ρm) d(t)
∆ρ = αρm[Τm−Τ(z, t)]
-z
Figure 6.6
Z0 −d(t)
Z
F= g (ρw − ρm )zdz + gαρm [T m − T (z − d, t)] (z − d) dz. (6.10)
−d(t) −∞
The first integral becomes

0
z2 g
(1) = g (ρw − ρm ) = (ρm − ρw ) d2 (t) . (6.11)
2 −d(t) 2
For the second integral let z0 = d (t) − z so the integral becomes

Z∞
z0
!
(2) = gα (T m − T o ) er f c √ z0 dz0 . (6.12)
2 κt
0
z0
Now let η = √
2 κt
so the integral becomes
Z∞
(2) = gα (T m − T o ) 4κt er f c (η)ηdη. (6.13)
0
Now integrate by parts

Z∞ ∞ Z∞
η2 1 2
er f c (η)ηdη = er f c (η) + √ η2 e−η dη. (6.14)
2 0 π
0 0
The first term on the right side is zero while the second term is 1/4 so the second integral
is simply
(2) = gαρm (T m − T o ) κt. (6.15)
The final result is

" #
2αρm (T m − T o )
F = gαρm (T m − T o ) κ 1 + t. (6.16)
π (ρm − ρw )
Now lets put in some numbers. The ridge push force at an age of 100 Ma is 3.2x1012
Nt m-1 . The lithosphere is approximately 100 km thick at that age so the average ridge
push stress is 32 MPa.
6.6 Rheology of the Lithosphere
Please read the article by Brace and Kohlstedt [4].
6.7 Exercises
Exercise 6.2. What is the average crustal stress needed to maintain the elevation of
Tibet (5 km) with respect to the elevation of India (0 km - sea level). Use the crust and
mantle densities from Figure 6.2. Assume the crustal thickness under India is 30 km.
(a) Use Airy isostasy to solve for the crustal thickness of Tibet.
(b) Calculate the outward driving force using equation 6.7.
(c) Calculate the average stress as the ratio of driving force to total crustal thickness.
Exercise 6.3. An ice sheet of thickness D and density ρi is floating on an ocean of

density ρw under the force of gravity g . Assume the ice sheet is thin relative to its
horizontal dimensions.
P ∆P
-h
g ρw ρi ρw
-D
ρw
Figure 6.7
(a) Sketch the pressure versus depth for the ice and water as well as the pressure differ-
ence versus depth.
(b) Derive an expression for the freeboard (i.e., height above sea level) of the ice sheet.
(c) Derive an expression for the outward driving force per unit length into the page
caused by this density configuration
Z0
Fs = ∆P (z)dz.
−D
Check your results in a limiting case.

Chapter 7
Brief Review of Elasticity
This is a very brief review of the elasticity theory needed to understand the principles
of stress, strain, and flexure in Geodynamics [81]. This review assumes that you have
already taken a course in continuum mechanics. One difference from T&S is that we
follow the sign convention used by seismologists and engineers, where extensional
strain and stress is positive.
7.1 Stress
Stress is a force acting on an area and is measured in Newtons per meter squared (Nm-2 ),
which corresponds to a Pascal unit (Pa). Figure 7.1 shows a cube of solid material.
Each face of the cube has three components of stress, so there are 9 possible compo-
nents of the stress tensor.
We will consider only the symmetric part of the stress tensor, so only 6 of these com-
ponents are independent. The antisymmetric part of the tensor represents a torque. In
Cartesian coordinates the stress tensor is given by
 
 σ σ xy σ xz 
 xx 
 
σi j =  σ xy σyy σyz 
 (7.1)
 

σ xz σyz σzz 
where index notation is the shorthand for dealing with tensors and vectors; a variable
with a single subscript is a vector ~a = ai , a variable with two subscripts is a tensor
σ = σi j , and a repeated index indicates summation over the spatial coordinates. For
80
CHAPTER 7. BRIEF REVIEW OF ELASTICITY 81
z
σ zz
σ zy
σ zx
σ yz
σ xz
σ yy
σ xy σ yx
σ xx
Figure 7.1
example the pressure is given by P = −σii /3. In addition, a comma preceding a sub-
script means differentiation with respect to that variable ∇~a = ai, j or, for example,
a x,y = ∂a x
∂y .
7.2 Strain
Strain is change in length over the original length, so it is a dimensionless variable; we

will assume strains are small ( 10−3 ). Let the displacement vector field inside of a
solid body be given by
~u = ui = u x uy uz (7.2)
The gradient of this vector is a tensor: ∇~u = ui, j . This tensor is commonly decomposed
into a symmetric tensor (strain) and an antisymmetric tensor (rotation).
1 ∂u ∂u j 1 ∂u ∂u j

ui, j = 2 ∂x i + ∂x + 2 ∂x i − ∂x (7.3)
j i j i
We will not consider the rotation tensor further but the strain tensor is given by

εi j = 12 ui, j + u j,i (7.4)
7.3 Stress vs. Strain
If one assumes the material has an isotropic and linear response, then the relationship
between stress and strain is given by
σi j = λδi j εkk + 2µεi j (7.5)
where δi j is equal to 0 except when i = j, and then it is equal to 1. The Lamé constants λ
and µ define the elastic properties. The shear modulus µ (or G in the engineering
literature) relates the shear stress to shear strain on a component by component basis:
∂u ∂uy

σ xy = 2µε xy = µ ∂yx + ∂x (7.6)
7.4 Invariants and Principal Stress
This general relation between stress and strain tensors is rather involved, and it is diffi-
cult to invert this relationship to develop a relationship between strain and stress. One
means of simplifying this relationship is to find a coordinate system rotation that will
cause the stress and strain tensors to be diagonal. Let R be a rotation matrix such
that Rt R = I is the identity matrix. There are three properties (invariants) of the stress
tensor that do not change under coordinate rotation. The invariants are found by first
developing the characteristic equation from the determinant of the following equation

σ − γ σ σ
xx xy xz
σ xy σyy − γ σyz = 0

(7.7)

σ σ σ − γ
xz yz zz
which becomes
γ3 − Iγ2 + IIγ − III = 0 (7.8)
where the stress invariants are
I = σii

II = 1
2 σii σ j j − σi j σi j = σ xx σyy + σyy σzz + σ xx σzz − σ2xy − σ2yz − σ2xz (7.9)

III = σi j
the trace I, the sum of minors II, and the determinant of the stress tensor III. The
first invariant is related to the mean normal stress or pressure P = −I/3. The second
invariant is related to shear stress and thus is commonly used as the Von Mises failure
criteria. We will not consider the third invariant further.
Exercise 7.1. Use symbolic algebra in Matlab to take the determinant of the character-
istic equation 7.7. Identify the first and second invariants in the third order polynomial
and check that they match the invariants in equation 7.9.
Real symmetric matrices have real eigenvalues, orthogonal eigenvectors, and can be
diagonalized. This implies that there always exists some principal coordinate system
where the the shear stresses are zero on planes orthogonal to the coordinate axes, and
where the normal stresses act along the principal axes directions (the eigenvectors)
form the rotation matrix R. The eigenvalues form the principal stress tensor
 
 σ 0 0 
 1 
 
~ p =  0
σ σ2 0  = R σR
t
(7.10)
 

σ3 

0 0
where
σ1 ≥ σ2 ≥ σ3 (7.11)
The principal stress system is important in geophysics and geology. Due to the presence
of the free surface, the stress field close to the Earth’s surface is expected to have one
principal stress vertical and hence two horizontal principal stresses. Also, in the Earth
we sometimes subtract the pressure from the stress tensor. In this case it is called
deviatoric stress. In the principal stress system the pressure and maximum shear stress
are given by
1
P = − 3 (σ1 + σ2 + σ3 )
(7.12)
1
τ= 2
(σ1 − σ3 )
7.5 Principal Stress and Strain
The stress versus strain relation is far simpler in the principal co-ordinate system
    
 σ   λ + 2µ λ λ
  ε 
 1     1 

 σ  =  λ λ + 2µ λ
 
  ε2 

(7.13)
 2     
     
 σ   λ λ λ + 2µ ε3 
3
where ε1 , ε2 , and ε3 are the principal strains. Next we can use this relationship to
develop three important parameters: Poisson’s ratio ν, Young’s modulus E, and bulk
modulus K.
First consider the case of uniaxial stress where σ2 = σ3 = 0. This represents ap-
plication of an end load to an elastic beam fastened to a wall. The second equation
for σ2 is
0 = λε1 + (λ + 2µ) ε2 + λε3 (7.14)
Because of symmetry we know ε2 = ε3 , so we arrive at a relationship between ε2 andε1 :
−λ
ε2 = 2 (λ + µ) ε1 = −νε1 (7.15)
where ν is Poisson’s ratio. Next we can use this relationship between strains in the first
equation to provide a relationship between σ1 and ε1 .
−λ2
σ1 = (λ + 2µ) ε1 + λ + µ ε1
(λ + 2µ) (λ + µ) − λ2
σ1 = λ+µ
ε1
(7.16)
µ (3λ + 2µ)
σ1 = λ+µ
ε1
σ1 = Eε1
where E is Young’s modulus.
Next we consider the case of uniform pressure. In this case, the change in pressure
∆P = − (σ1 + σ2 + σ3 ) /3 is related to a change in volume ∆V = (ε1 + ε2 + ε3 ). Using
the stress-strain relation we find.

2
∆P = − λ + 3 µ ∆V
(7.17)
∆P = −K∆V
where K is the bulk modulus. One can invert this stress vs. strain relationship 7.13 to
obtain a strain vs. stress relationship. We’ll also assume that the principal co-ordinates
are aligned with the x, y, and z axes.
    
 ε   1 −ν −ν   σ 
 xx     xx 
 1 
 εyy  =  −ν 1 −ν   σyy 
  
(7.18)
  E    
     
εzz −ν −ν 1 σzz
Exercise 7.2. Use symbolic algebra in Matlab to invert the stress versus strain rela-
tionship 7.13 to obtain the relationship between strain and stress 7.18. Show that the
product of these two matrices is the identity matrix.
Now we have arrived at equations (3.31), (3.32), and (3.33) in T&S [81]. Before mov-
ing onto the flexure problem we consider the case of a thin elastic plate. Thin plate
means that there are no variations in the vertical displacement field as a function of
depth in the plate, so we can make the approximation σzz = 0. Under this approxima-
tion we have the following
1
ε xx = E σ xx − vσyy
1
εyy = E σyy − vσ xx (7.19)
−ν
εzz = E σ xx + σyy
These equations are the starting point for the development of the relationship between
bending moment and curvature provided in section 3.9 of T&S [81].
7.6 Exercises
Exercise 7.3. Use the thin plate equations 7.19 to develop a linear relationship between
moment and curvature. What are the important parameters that control the flexural
rigidity. It will be helpful to study section 3.9 in T&S [81].
Chapter 8
Flexure of the Lithosphere
This chapter is basically a supplement to Turcotte and Schubert [81, Chapter 3]. The
results of the first derivation are the same as T&S [81, equation (3.130)], but rather
than guessing the general solution, the solution is developed using Fourier transforms.
The approach is similar to the solutions of the marine magnetic anomaly problem,
the lithospheric heat conduction problem, the strike-slip fault flexure problem, and the
flat-Earth gravity problem. In all these cases, we use the Cauchy integral theorem to
perform the inverse Fourier transform. Later we’ll combine this flexure solution with
the gravity solution to develop the gravity-to-topography transfer function. Moreover,
one can take this approach further to develop a Green’s function relating temperature,
heat flow, topography and gravity to a point heat source (e.g., [60]). In addition to
the constant flexural rigidity solution found in the literature, we develop an iterative
solution to flexure with spatially variable rigidity.
Before going over these notes, please reread Turcotte and Schubert [81, Section 3-9]
on the development of moment versus curvature for a thin elastic plate.
The loading problem is illustrated in Figure 8.1. We start with a simple line source,
but the solution method also applies to a point source. Of course, the point source
Green’s function can be convolved with an arbitrary load distribution to make the so-
lution completely general; we’ll do this later. The vertical force balance for flexure of
a thin elastic plate floating on the mantle is described by the following differential:
d2 d2 w d2 w

2 D(x) 2 + F 2 + ∆ρgw = q(x) (8.1)
dx dx dx
end restoring vertical
flexural load
load force
resistance
See also Table 8.1.
86
CHAPTER 8. FLEXURE OF THE LITHOSPHERE 87
Figure 8.1: Thin elastic plate subjected to vertical load q(x) and end load F. The plate
responds with a vertical displacement w(x).
Parameter Definition Value/Unit
w(x) deflection of plate m

(positive down)
D(x) flexural rigidity Nm
h or T e elastic plate thickness m
F end load N m−1
q vertical load N m−2
∆ρ density contrast 2200 kg m−3
(ρm − ρw )
g acceleration of gravity 9.82 m s−2
E Young’s modulus 6.5 × 1010 Pa
v Poisson’s ratio 0.25
Table 8.1
8.1 Constant Flexural Rigidity, Line Load, No End Load
Under these assumptions, the differential equation and boundary conditions become
d4 w
D + (ρm − ρw ) gw(x) = Vo δ(x)
dx4
(8.2)
dw
lim w(x) = 0 and lim =0
|x|→∞ |x|→∞ dx
Take the Fourier transform of the differential equation, where the forward and inverse
transforms are defined as
Z∞ Z∞
1
F(k) = f (x) e−ikx dx f (x) = 2π F(k) eikx dk (8.3)
−∞ −∞
where the wavenumber is now 2π/λ instead of the usual 1/λ. I have switched notation
because it saves writing 2π many times, and also these are old notes. The derivative
property is now = [dw/ dx] = ik= [w]. The Fourier transform of the differential equa-
tion is
Dk4 W(k) + (ρm − ρw ) gW(k) = Vo (8.4)
and the solution for plate deflection is simply
4 −1 Vo

W(k) = k4 + 4 (8.5)
D α
where the flexural parameter α is (see [81, equation (3.127)])

4D
α4 = (8.6)
g (ρm − ρw )
Now take the inverse Fourier transform of equation (8.6).
Z∞
Vo eikx
w(x) = dk (8.7)
2πD k4 + α44
−∞
As in the other solutions, we find the poles in the denominator of equation (8.7) and
integrate around the poles.
! ! !
4 2i 2i
k + 4 = k + 2 k − 2
4 2 2
α α α
(8.8)
1+i −1 + i
! ! ! ! !
4 1−i −1 − i
k + 4 = k−
4
k− k− k−
α α α α α
See also Figure 8.2.

Figure 8.2: Location of poles in the complex plane. The integration path from −∞ to
∞ can be closed in the upper or lower hemisphere.
First consider the case for x > 0. To match the boundary conditions at infinity we want
Im(k) > 0. Thus we close the integration in the upper half of the plane and apply the
Cauchy Residue Theorem I
f (z)
dz = i2π f (zo ) (8.9)
z − zo
The relevant poles are
1+i −1 + i
k= α and k= α
(8.10)
The solution is
1+i
α3 ei( α ) x

Vo
w(x) =

2πi 
2πD (1 + i + 1 − i) (1 + i − 1 + i) (1 + i + 1 + i)
(8.11)
−1+i
α3 ei( α ) x

+

(−1 + i − 1 − i) (−1 + i − 1 + i) (−1 + i + 1 + i)

After some simplification this becomes
Vo α3 −x/α eix/α e−ix/α

" #
w(x) = e + (8.12)
8D (1 + i) (1 − i)
This can be rewritten in terms of cos(x/α) and sin(x/α). Also we know that the solution
should be symmetric about x = 0. The final result for positive x matches [81, equation
(3.130)].
Vo α3 −x/α
w(x) = e [cos(x/α) + sin(x/α)] (8.13)
8D
The important parameters and length scales in this solution are:

α flexural parameter
√
2πα flexural wavelength
xo = 3πα/4 distance to the first crossing
The Figure 8.3 from Turcotte and Schubert [81] show the solution for the continuous
plate where the maximum flexure is normalized to 1. In addition the solution for a
broken plate is shown. This is also the same form used to model plate bending at a
subduction zone. Note for the same downward force, the amplitude of the broken plate
is twice the amplitude of the continuous plate.
Figure 8.4 shows isostasy along the Hawaiian-Emperor Seamount Chain.

Figure 8.3: From Turcotte and Schubert [81].

Figure 8.4: (From Watts and Cochran [86].) The volcanic loading from the Hawaiian-
Emperor seamount chain causes downward flexure of the plate. (lower) Both gravity
and bathymetry profiles are used to estimate the elastic thickness of the plate.
8.2 Variable Flexural Rigidity,

Arbitrary Line Load, No End Load
For this case we need to solve a linear differential equation, but with a variable coeffi-
cient. This will involve an iteration scheme in the Fourier transform domain where the
first iteration is basically equation (8.5) above. See the original derivation in [62]. The
differential equation and boundary conditions are
d2 d2 w(x)

2 D(x) 2 + (ρm − ρw ) gw(x) = q(x)
dx dx
(8.14)
dw
lim w(x) = 0 and lim dx = 0
|x|→∞ |x|→∞
where D(x) is now the spatially variable flexural rigidity, w(x) is the deflection of the
plate, and q(x) is the applied load. It is assumed that D, w, and P are band-limited func-
tions, so that their Fourier transforms exist. The functions D and w can be written as
Z∞
1
D(x) = D(s)eisx ds
2π
−∞
(8.15)
Z∞
1
w(x) = W(r)eirx dr
2π
−∞
Upon substitution of these expressions for D and w into the first term of equation (8.14)
and differentiating under the integral, the following is obtained:
Z∞ Z∞
1
(r + s)2 r2 D(s) W(r)ei(s+r)x dr ds + g (ρm − ρw ) w(x) = P(x) (8.16)
(2π)2
−∞ −∞
The Fourier transform of equation (8.16) is

Z∞ Z∞ Z∞
1
(r + s) r D(s)W(r)
2 2
ei(s+r−k)x dx dr ds + g (ρm − ρw ) W(k) = Q(k)
(2π)2
−∞ −∞ −∞
(8.17)
By making use of the definition of the delta function,
Z∞
1 h i
ei(s+r−k)x dx = δ r − (k − s) (8.18)
2π
−∞
performing the integral with respect to r, and using the band-limited property of D(s)
(i.e., D(s) = 0|s| > β) , equation (8.17) reduces to a Fredholm integral equation:
Zβ
k2
D(s)W(k − s)(k − s)2 ds + g (ρm − ρw ) W(k) = Q(k) (8.19)
2π
−β
Notice that if the flexural rigidity is constant D(x) = Do , then D(s) = 2π Do δ(s). For
this case, the solution for the plate deflection for an arbitrary load is
h i−1
W(k) = Do k4 + g (ρm − ρw ) Q(k) (8.20)
Now consider the more general case of spatially variable flexural rigidity
D(s) = D0 (s) + 2πDo δ(s) (8.21)
Inserting equation (8.21) into equation (8.19) and rearranging terms yields
β
 
i−1  2 Z 
h k
W(k) = Do k4 + g (ρm − ρw ) Q(k) − D0 (s)W(k − s)(k − s)2 ds

(8.22)
 2π 
−β
The plate deflection appears on both sides of equation (8.22), so there is no closed
form solution for W(k). However, if the variations in flexural rigidity D0 are small
compared with the mean value of flexural rigidity Do , then this equation can be solved
by successive approximation. The original derivation in [62] provides the necessary
requirement for convergence, but a numerical illustration is also useful. This variable-
rigidity flexure approach has also been applied in 3-D [26].
Figure 8.5 is a numerical example of flexure of a plate with a sharp reduction in plate
thickness at the origin. The upper curve compares the flexure of a continuous plate,
(continuous curve, analytic solution, equation (8.13) to the Fourier transform solution
to equation (8.2) (dashed curve). The lower plot is a comparison of the analytic solution
to flexure of a broken plate (continuous curve, [81, equation (3.140)]) to the numerical
iterative solution of equation (8.22) (dashed curve). For this case the thickness of the
plate at the origin was reduced by 95%. This approximates the broken plate solution.
continuous plate
0.2
w/wo
0.4
0.6
0.8
1
0 1 2 3 4 5 6 7 8 9 10
x/α
broken plate
0.5
w/wo
1.5
2.5
0 1 2 3 4 5 6 7 8 9 10
x/α
Figure 8.5: (upper) Flexure of a plate with constant flexural rigidity. The Fourier trans-
form (dashed) and analytic solutions show good agreement. (lower) Flexure of a bro-
ken plate. The iterative Fourier transform solution (dashed) and analytic solutions show
good agreement.
8.3 Stability of Thin Elastic Plate Under End Load
Consider the original differential equation 8.1 and now assume constant flexural rigid-
ity with a delta function line load and uniform compressive end load F . The line load
by itself will cause flexure of the plate as given in equation (8.13). For sufficiently
large end load, this initial deflection will become amplified and the plate will buckle.
Here we develop formulas for the critical end load when the amplification becomes
unbounded. In addition, the buckling will occur at a particular wavelength λc . Under
these assumptions, the differential equation becomes
d4 w d2 w
D + F + ∆ρgw(x) = Vo δ(x). (8.23)
dx4 dx2
The Fourier transform of the differential equation is
Dk4 W(k) − Fk2 W(k) + ∆ρgW(k) = Vo . (8.24)

and the solution for plate deflection is
h i−1
W (k) = Vo Dk4 − Fk2 + ∆ρg . (8.25)
The deflection of the plate is singular when the denominator goes to zero. We note that
the denominator is a polynomial so we can use the quadratic formula to search for the
zeros of this equation. We can write this as a quadratic equation in k2
p
F± F 2 − 4D∆ρg
k =
2
. (8.26)
2D
We know the wavenumber must be a real number and this only occurs when
p
F> 4D∆ρg. (8.27)
When F is equal to this critical value, we call this the critical end load Fc . The wave-
length when this occurs is the critical wavelength given by
" #1/4
D
λc = 2π (8.28)
∆ρg
3
where D is the flexural rigidity given by D = 12(Eh
1−ν2 )
and h is the elastic plate thickness.
Note this wavelength λc is also equal to the flexural wavelength discussed in Section
8.1.
An example of the flexure due to a line load on a thin elastic plate is shown in Fig-
ure 8.6. When no end load is applied, the flexure follows the analytic solution for a
continuous plate given in equation (8.13). When an end load is applied the plate begins
to buckle. This example has an end load F = 0.9Fc .
Next consider an example of buckling of oceanic lithosphere that is 50 km thick. The

density contrast ∆ρ = (ρm − ρw ) and other parameters are given in Table 8.1. The
parameters of interest are the average stress at the ends of the plate σ = Fc /h and
the buckling wavelength λc . For this case the values are 4.9 GPa and 475 km. From
the analysis of the yield strength envelope, it is clear that this level of stress cannot
be sustained by even the strongest oceanic lithosphere. Therefore the elastic buckling
model is not relevant for the Earth. One must consider the inelastic properties of the
lithosphere to understand the response to large end loads.
0.5
0.9 Fc
0
w/wo
-0.5
-1
-20 -15 -10 -5 0 5 10 15 20
x/α
Figure 8.6: (solid) Normalized flexure a continuous plate under a line load. (dashed)
Normalized flexure with line load and end load. The plate begins to buckle at the
flexural wavelength.
8.4 Exercises
Exercise 8.1. Continental yield strength envelope model. The continental yield strength
has been described as a jelly sandwich consisting of a weak layer (jelly) between two
strong layers (bread). The flexural rigidity of a single strong later is
Eh3
D= . (8.29)
12 1 − ν2
(a) What is the flexural rigidity of two strong layers, each of thickness h/2, that are not
bonded along their common interface?
(b) What is the effective elastic thickness for this layered case (bottom diagram in
figure 8.7)?
Figure 8.7: (solid) (top) Single plate of thickness h. (bottom) Two plates each of thick-
ness h/2 are not bonded at their interface so they act independently under a bending
moment. .
Exercise 8.2. Write a matlab program to generate the two flexure curves in Figure 8.6.
Start with the code used in Exercise 2.8.
Chapter 9
Flexure Examples
This chapter provides practical examples of flexural models applied to structures in the
oceanic and continental lithosphere. The models are all basically solutions to the thin-
plate flexure equation with a variety of surface loads, sub surface loads, and boundary
conditions. Both gravity and topography data are used to constrain the models. We’ll
see in a following chapter that gravity data provide important constraints on the to-
pography of the moho. Figures and captions from various sources are provided on the
following pages. In addition, a reference is provided for each figure.
The features include:
Seamount undersea volcano loading the oceanic lithosphere.

Trench plate bending at a subduction zones.
Fracture Zone flexure that accumulates due to the differential subsidence across an
oceanic fracture zone.
Normal Fault asymmetric topography due to normal faults mainly in continental crust.
Thrust Fault asymmetric topography due to a thrust fault in the foreland basin adja-
cent to continental mountains.
Rift symmetrical but aborted continental rift which usually fills with sediment to form
a steer head basin.
Passive Margin symmetrical rift that evolves into a young ocean basin and accumu-
lates continental sediments to become a passive continental margin.
Venus Trench flexural topography on Venus that is caused by thrust faulting and pos-
sibly earth-like subduction.
98
CHAPTER 9. FLEXURE EXAMPLES 99
9.1 Large Seamount
Figure 9.1: (From Watts [85, Figure 4.21].) “Comparison of the crustal structure at
the Canary (Watts et al. [83]), Hawaiian (Watts and ten Brink [89]), and Marquesas
(Caress et al. [8]) Islands with the predictions of simple eleastic plate models. ”
9.2 Seamount Model
Figure 9.2: (From Watts [85, Figure 4.16].) “Comparison of observed and calculated
gravity anomalies along a profile that intersects the Hawaiian Islands between Oahu
and Molokai. The observed profile is based on data acquired during cruise KK8213
of R/V Kana Keoki. The calculated profiles are based on an elastic plate model with
T e = 10, 20, 30, 50, and 75 km. The numbers on the right-hand side are the root mean
square differences between observed and calculated anomalies.”
9.3 Seamount Flexure vs. Age
Figure 9.3: (From Watts [85, Figure 6.11].) “Plot ot T e against the age of the litho-
sphere at the time of loading. ”
9.4 Trench Model
Figure 9.4: (From Levitt and Sandwell [40, Figure 2].) “Schematic representation of
topography (w0 , x0 , α, and d0 ) and gravity (w0 , x0 , α, and g0 ) model parameters. See
equations (3) and (5) for details of the model. ”
9.5 Trench Data
Figure 9.5: (From Levitt and Sandwell [40, Figure 5].) “Topography and gravity
data for a subset of the profiles modeled in the study. Profiles are shifted vertically
for presentation. For topography (shown at left), solid lines corresond to the best fit
model for topograv (λ − 0.4) and dashed lines correspond to the best fit model for topo
(λ = 0), and dashed lines correspond to the best fit model for grav (λ = ∞). Profiles
are numbered according to the numbering scheme of Table 1. “TG”, ‘G’, and ‘T’ mark
profiles that were “successfully” modeled by topograv, grav, and topo, respectively.”
9.6 Outer Rise Normal Faults
Figure 9.6: (From Massell [41, Figure 5].) High resolution multibeam bathymetry
data reveals the development of horst and graben structures on the outer trench wall of
the Tonga Trench.
9.7 Brittle and Ductile Deformation
Figure 9.7: (From Watts [85, Figure 6.34].) (a) Stress versus depth in bent thin elastic
plate has extension near the surface and compression at depth. (b) Stresses cannot
exceed the yield strength of the plate resulting in brittle fracture near the surface and
ductile flow at depth. The elastic core is thinner than the original elastic thickness.
Figure 9.8: (From Watts [85, Figure 6.27].) “Yield Strength Envelope and strength
profiles for the oceanic lithosphere. (a) YSE for the cooling plate (solid lines) and
half-space (dashed lines) models and for ages of lithosphere in the range 4–144 Ma.
(b) Integrated strength of the lithosphere obtained by calculating the area under YSE
at each time interval. Parameters defining the flow laws of olivine are as given in
Table 6.4.”
9.8 Trench Flexure and Earthquakes
Figure 9.9: (From Watts [85, Figure 6.37].) “Comparison of the depth and focal
mechanisms of earthquakes in deep-sea trench-outer rise systems with the predictions
of the Yield Strength Envelope (YSE) model. The earthquake data are based on a com-
pilation by Seno and Yamanaka [72]. The YSE model is based on a thermal structure
determined by the cooling plate model, an olivine rheology and the range of strain rates
and applied stresses shown.”
9.9 Continental Yield Strength and Earthquakes
Figure 9.10: (From Watts [85, Figure 6.28].) Continental yield strength envelope for
a variety of crustal rock types.
Figure 9.11: (From Watts [85, Figure 6.40].) “Comparison of T e with the depth of
earthquakes in (a) oceans and (b) continents. The T e data are based on data in Tables 6.1
and 6.2. The earthquake data are based on the compilations of Chen and Molnar [10]
and Wiens and Stein [91].”
9.10 Fracture Zone Model
4658 SANDWELL AND SCHUBERT: LITHOSPHERE FLEXURE AT FRACTURE ZON
c B A
the Juan de Fuca ridge and
identified magneticanomalies
These anomalies were dated
scale of Ness et al. [1980]. Th
offset which varies between 30
140øW.To the west of 140øW,t
C' B
.......
I!,Fault
I age offset gradually shifts from
southern branch. However, on
both FZ's is known, sincethe l
FZ's does not contain identif
PioneerFZ runssubparallelto
age offsetwhich variesbetwee
--dref 4 Myr at 104øW.
Five N-S trendingbathymet
were chosenfor this studyand
other continuousN-S profiles
additional profiles lie within
chosen, and they therefore co
The depthsalongeachof the p
sounding charts used to cons
Northeast Pacific' contour m
dref 1981]. The spacingsbetween
min of latitude. Because of th
shortestwavelengthtopograp
20 Myr wavelength topographicfeatur
The bathymetryalongeacho
'' 40Myr is shown in Figure 3 togethe
spheric segment. The dashe
Fig. 1. Evolution of a fracture zone. (Top) Spreadingridges
Figure 9.12: (From Sandwell and Schubert [63, Figure 1].)
offset by a transformfault. The age offset acrossthe FZ is te, - “Evolution of a profiles
fracture predictedby the dept
(Center) The h's are the differences in ocean floor depth between the following section)and the
zone. (Top) Spreading ridged offset by a transform fault. The age of offset across the
−tB . (Center)
FZ is tB0 locations The
far to h’s are
either theof
side differences
the FZ. The in ocean
initialfloor depth
height between
of the scarplocations
at The predicteddepthsare in ap
far to either
the FZsideisofhA.
theIfFZ.
theThe
FZ initial
doesnotheight ofthe
slip, the scarp
scarp height
at the FZ
mustis hremain
A . If the FZdata for locations more than 1
does notconstant. The constancy
slip, the scarp height mustofremain
scarpheight andThe
constant. theconstancy
decreasein of hscarp
withheight
the FZ's (marked by arrows i
and the age cause
decrease in the lithosphere
h with age cause inthethe vicinityinofthe
lithosphere thevicinity
FZ toofbend.the FZ Theto bend.
flexural amplitude/ie vicinity of each FZ there are
The flexural amplitude δB is theisdifference
the difference
betweenbetween
hA andhA hBand he. Similarly,
. Similarly, δc = hA − hC .
•ic = hA I hE. (Bottom) Sketch of bathymetry
0 0along profiles
(Bottom) Sketch of bathymetry along profiles A-A , B-B , and C-C illustrating the 0 A-A', between the predictedand ob
B-B', and C-C' illustratingthe lithosphericflexuredescribedabove. the ocean floor is anomalous
lithospheric flexure described above.”
of theF•, whileontheolders
expected.We will showthat th
and the vertical offset of distant ocean floor. Therefore the
are a consequenceof lithosph
only adjustable model parameter is the stress relaxation
temperature Te which definesthe base of the elastic layer. MOn•
We will show that good fits between the observed and As the oceanic lithospher
predicted bathymetric profiles are obtainedby varying this accordingto the depth-agerel
singleparameter. These goodfits alsoconfirmthe absenceof
vertical slip on the fossil fault planesof the Mendocinoand
Pioneer FZ's.
9.11 Fracture Zone Data
Figure 9.13: (From Sandwell and Schubert [63, Figure 6].) “Comparisons between
theoretical bathymetric profiles computed from flexure models assuming no slip on
the FZs (dashed lines) and the observed bathymetric profiles A-E (solid lines). The
asymmetric flexure predicted by the model across each FZ is a consequence of the
increase in flexural wavelength with age. The apparent tilt in the bathymetry between
the Pioneer and Menocino FZs occurs because the flexural wavelength is greater than
their separation distance.”
9.12 Normal Fault Model
Figure 9.14: (From Watts [85, Figure 7.4].) “Vening Meinesz’s model for the de-
velopment of a graben. Reproducted from Figure 10D-1 of Heiskanen and Venning
Meinesz [31] with permission of McGraw-Hill Inc.”
9.13 Normal Fault Data
Figure 9.15: (From Watts [85, Figure 7.8b].) “Comparison of the predictions of
the Weissel and Karner [92] model with free-air gravity anomaly, depth to basement
and topograhy profile data over the Western Rifts of East Africa. Reproducted from
Figure 11.a–f of Ebinger et al. [20], copyright the American Geophysical Union.”
9.14 Thrust Fault
Figure 9.16: (From Watts [85, Figure 7.50].) “Comparison of observed and calcu-
lated Bouguer gravity anomalies along a profile of the subAndean thrust and fold belt
in Ecuador. The observed anomalies are based on a compilation by GETECH. The
calculated anomalies are based on a broken plate model which is subject to a surface
(topographic) load and an additinal shear force of 0.5 × 1012 M n−1 that has been ap-
plied at the plate break. (a) Topography, (b) variation in T e , and (c) Bouguer gravity
anomalies. Reproduced from Figure 7 of Stewart and Watts [79].”
9.15 Continental Rift and Basin Evolution
Figure 9.17: (From Watts [85, Figure 7.28].) “The predicted stratigraphy of a basin,
based on the tectonic subsidence model in Figure 7.27. The stratigraphy has been com-
puted by assuming that the sediments represent a load on the underlying lithosphere,
which subsides under their weight. The calculations assume a T e that increases with the
age of loading, as defined by the depth to the (a) 300◦ , (b) 450◦ , and (c) 600◦ isotherms.
Each model is associated with “onlap” of strata at their edges that is due to the increase
in T e with age. Modified from Watts et al. [87, Figure 4].”
9.16 Passive Margin Model
Figure 9.18: (From Watts [85, Figure 4.39].) “Simple model showing the contribu-
tions to the gravity anomaly of a prograding sediment wedge. The net effect is to move
the edge-effect anomaly associated with the initial margin seawards. The amplitude
and wavelength of the calculated anomaly depends strongly on T e with the largest am-
plitude anomalies being associated with the strong margin and smallest weak margin.”
9.17 Passive Margin Data
Figure 9.19: (From Watts [85, Figure 7.25].) “The application of the process oriented
approach to seismic and gravity anomaly data at the East Coast, U.S. rifted margin.
In this approach, sediments in a rift basin are progressively backstripped through time
and the gravity anomalies due to the combined effects of rift and sedimentation are
computed for different assumed values of T e . By comparing observed and calculated
gravity anomalies it is possile to constrain T e and, in some cases, its spatial and tempo-
ral variations. (a) Observed (black dots) and calculated gravity anomalies. (b) Crustal
structure deduced from backstripping and gravity modelling. Reproduced from Fig-
ures 6 and 7 of Watts and Marr [88].”
9.18 Venus Trench Data
Figure 9.20: (From Sandwell and Schubert [64, Figure 1.].) (upper) Shaded
bathymetry map of the Sandwich trench, South Atlantic. (lower) Shaded topography
of Latona Corona, Venus.
9.19 Venus Trench Model
Figure 9.21: (From Sandwell and Schubert [64, ].) Profiles across the Sandwich
Trench (A - A’) and Latona Corona (B - B’) show similar flexural wavelength and
amplitude.
Chapter 10
Elastic Solutions
for Strike-Slip Faulting
(Copyright 2001, David T. Sandwell)
(Reference: [90] [67][11])
This chapter provides the mathematical development of the deformation and strain pat-
tern due to strike-slip deformation on a partially locked fault. The notes come from [81,
Chapter 8], but I’ll focus on section 8-6 through 8-9. While I’ll follow the overall theme
of Chapter 8, I’ll deviate in two respects. First I’ll use a coordinate system with the z-
axis pointed upward, to be consistent with my previous notes on gravity, magnetics, and
heat flow. Second I’ll develop the solution using Fourier transforms, to be consistent
with my other chapters.
10.1 Interseismic Strain Buildup
The first objective is to derive an expression for the surface displacement v(x) and
surface strain δv/δx for the model shown In Figure 10.1. A constant velocity V is
applied at an elastic half space. There is a fault in the half space with locked and
creeping sections. Because of free slip on the faults, the strain will be concentrated
near the fault.
118
CHAPTER 10. ELASTIC SOLUTIONS FOR STRIKE-SLIP FAULTING 119
Figure 10.1
The approach will be as follows:
1. Develop the force balance from basic principles.

2. Establish the line-source Green’s function for an elastic full space.
3. Establish the screw-dislocation Green’s function for an elastic full space.
4. Use the method of images to construct a half-space solution.
5. Integrate the line sources to develop the solutions found in the literature.
6. Compute the geodetic moment accumulation rate for an arbitrary slip distribution.
7. Inclined fault plane
8. Matlab examples
10.1.1 Force Balance
Consider the forces acting on the infinitely long (y-direction) square rod depicted in
Figure 10.2. The body force per unit volume of rod must be balanced by tractions on
the sides of the rod.
Figure 10.2
The equation for this force balance is
τ xy (x + δx) − τ xy (x) δy δz + τzy (z + δz) − τzy (z) δx δy = ρ (x, z) δz δy δz

(10.1)
where τ xy and τzy are the shear tractions on the side and top of the box, respectively,
and ρ(x, y) is the body force that depends only on x and z. Dividing through by δx δy δz
and taking the limit as all three go to zero, we arrive at:
∂τ xy ∂τzy
∂x
+ ∂z = ρ (x, z) (10.2)
Given the following relationship between stress and displacement, the differential equa-
tion reduces to Poisson’s equation
∂v
τ xy = µ ∂x
(10.3)
∂v
τzy = µ ∂z
∂2 v ∂2 v
+ = µ1 ρ (x, z) (10.4)
∂x2 ∂z2
where µ is the shear modulus and v is the displacement in the y-direction.
10.1.2 Line-Source Green’s Function
We can generate the solution to an arbitrary source distribution by first developing the
line-source Green’s function. Consider a line source at a depth of −a. The differential
equation is:
∂2 v ∂2 v
+ = Aµ δ(x) δ(z + a) (10.5)
∂x2 ∂z2
where A is the source strength having units of force/length, or force/length/time if
this will represent an interseismic velocity. The boundary conditions for this second-
order, partial differential are that v must vanish as both |x| and |z| go to infinity. The
2-dimensional forward and inverse Fourier transforms are defined as
Z∞ Z∞
F (k) = f (x) e−i2π(k•x) d2 x
−∞ −∞
(10.6)
Z∞ Z∞
f (x) = F(k) ei2π(k•x) d2 k
−∞ −∞
where k = (k x , ky ) and x = (x, y). Take the 2-dimensional Fourier transform of the
differential equation (10.5).

− (2π)2 k2x + kz2 V(k) = Aµ ei2πkz a (10.7)
so the solution in the Fourier domain is

−Aei2πkz a
V(k) = (10.8)
µ(2π)2 k2x + kz2
Now we need to take the inverse Fourier transform with respect to kz and make sure the
solution goes to zero as |z| goes to infinity. The integral is
Z∞
−A ei2πkz (z+a)
V(k x , z) = dkz (10.9)
µ(2π)2

k2x + kz2
−∞
First consider the case k x > 0, z + a > 0. We can factor the denominator and recognize
that the integrand will vanish for large positive z if we close the contour in the upper
hemisphere.
ei2πkz (z+a)
I
−A
V(k x , z) = dkz (10.10)
µ(2π)2

kz + ik x kz − ik x
From the Cauchy integral formula, we know that for any analytic function the following
holds for a counterclockwise path surrounding the pole.
I
f (z)
dz = i2π f (zo ) (10.11)
z − zo
Figure 10.3
In this case with the pole at ik x , the result is simply
−i2πA e−2πkx (z+a) −A e−2πkx (z+a)

V(k x , z) = = (10.12)
µ4π 2 i2k x 2µ 2πk x
Next consider k x < 0, z + a > 0. In this case we must close the integration path in the
lower hemisphere to satisfy the boundary conditions; during the integration the only
contribution will be from the −ik x pole. The overall result is to replace k x by |k x |.
−A e−2π|kx |(z+a)
V(k x , z) = (10.13)
2µ 2π k x
Note this is exactly the same functional form as the heat flor solution in Chapter 2.
The Green’s function is the inverse cosine transform of equation (10.13), or ln(r2 ). The
final result is
−A h i
v(x, z) = 4πµ ln x2 + (z + a)2 (10.14)
10.1.3 Screw Dislocation for Line Source Green’s Function
In order to produce a fault plane with strike-slip displacement, we need to construct a

line-source screw dislocation. This can be accomplished by abutting equal but opposite
line source dislocations, as shown in Figure 10.4.
A simple way of constructing the screw source is to take the derivative of the line
source Green’s function in a direction normal to the fault plane. So we need to develop
the Green’s function for the following differential equation:
h i. ∂
∇2 vscrew = δ(z + a) δ(x + dx) − δ(x) dx = δ(z + a) ∂x δ(x) (10.15)
line screw
Figure 10.4
To do this we take the derivative of the line source Green’s function in equation (10.14):
−A ∂ h i −Ax h i−1
vscrew (x, z) = 4πµ ∂x ln x2 + (z + a)2 = 2πµ x2 + (z + a)2 (10.16)
So the Green’s function for a line-source screw dislocation at depth is

−A x
vscrew (x, z) = 2πµ h 2 i (10.17)
x + (z + a)2
10.1.4 Surface Boundary Condition: Method of Images
The surface boundary condition is that the shear stress τzy must be equal to zero, but the
full-space result provides a non-zero result. This boundary condition will be satisfied if
we place an image source at z = a . When the combined source and image are evaluated
at the surface z = 0, the result is to double the strength of the Green’s function.
h i−1 h i−1
−A
v(x, z) = 2πµ x x2 + (z + a)2 + x2 + (z − a)2
(10.18)
−A x
v(x, 0) = πµ x2 + a2

10.1.5 Vertical Integration of Line Source to Create a Fault Plane
The final step in the development is to integrate the line-source screw dislocation over
depth. We consider three cases: (1) deep slip to represent interseismic deformation
above a locked fault, (2) shallow slip to represent shallow creep, and (3) shallow slip
on a stress-free crack to represent an earthquake.
Case 1 First consider a fault that is free-slip between a depth −D and infinity. This is
the solution considered by Savage [1990] [66]. See Figure 10.5.
Figure 10.5
The integral of the line source Green’s function is

Z−D
−A x
v(x) = dz (10.19)
πµ x2 + z2
−∞
To integrate 10.19 make the following substitution

η = −xz−1 so dη = xz−2 dz (10.20)
The integral becomes
Zx/D
−A 1 −A x
v(x) = dη = πµ tan−1 D (10.21)
πµ 1+η 2
0
We know that v(±∞) = +V/2, so A = −Vµ. Note that A has units of force per unit
area times a velocity. This corresponds to a moment rate per area of fault. The familiar
results for displacement and shear stress are
x
v(x) = Vπ tan−1 D
(10.22)
µV 1
τ xy = πD 2
1 + Dx
Consider the extreme cases of a completely unlocked fault such that D = 0. The
displacement field will be a step function and the stress will be everywhere zero except
at the origin, where it will be infinite.
Case 2 Next consider a fault that is free-slip between the surface, and a depth −d. In
this case the integral is
Z0
V 1 V
∞
v(x) = dη = tan η
−1
(10.23)
π 1 + η2 π −x/d
−x/d
There are two cases depending on whether x is positive or negative:

V π

x
v(x) = π − tan−1 d x>0
2
(10.24)
−π
V x
v(x) = π
− tan−1 d x<0
2
By combining these, the displacement and shear stress are

h i V x
v(x) = V H(x) − 12 − π tan−1 d
  (10.25)
1 1
τ xy = µV δ(x) −
 
πd 1 +
 x 2 
d
If the fault is completely unlocked so that as d goes to infinity, the displacement be-
comes a step and the shear stress is infinite at the origin, in agreement with our concepts
of a free-slipping fault.
Case 3 The third case considered also has shallow slip between depth −d and the
surface. However, in this case we consider a so-called crack model, where the slip
versus depth function results in zero stress on the fault. This derivation will lead to the
crack solution given in [81, equation (8.110)]. The Case 2 solution has uniform slip
with depth. This leads to a stress singularity at the base of the fault. In contrast, the
model in [81] has a stress-free crack imbedded in a pre-stressed elastic half space. Us-
ing the Green’s function developed above it can be shown that the two solutions are in
fundamental agreement. The only difference is related to the slip versus depth function.
From the dislocation theory developed in equation (10.19), the y-displacement as a

function of distance from the fault is given by
Z0
1 s(z)x
ν (x) = dz (10.26)
π x 2 + z2
−d
where z is depth, x is distance from the fault, s (z) is the slip versus depth, and v (x)
is the displacement. Now consider the two slip versus depth functions between the
surface and −d.
s1 = S
1/2 (10.27)
s2 = S 1 − z2 /d2
The first slip function is constant with depth while the second corresponds to the stress-
free crack and has the form provided in [81, equation (8-93)]. Using the approach
described above, the integral of the constant slip with depth s1 is
!
S x π x
v (x) = π − tan (10.28)
|x| 2 d
The integral of the slip function for the crack model s2 is given by
Z0 1/2 Zd 1/2
S 1 − z2 /d2 S 1 − z2 /d2
v (x) = π
x dz = π
x dz (10.29)
x 2 + z2 x2 + z2
−d 0
Now we let x = x/d and z = z/d, so the integral becomes

0 0
Z1 1/2
S 0 1 − z0 2
v x = 0
π
x dz (10.30)
x0 2 + z0 2
0
This integral can be performed in Matlab using the following code with the sym-
bolic toolbox.
%
clear
syms x positive
syms z
arg=sqrt(1-z*z)/(x*x+z*z);
int(arg,z,0,1)
%
% ans=-1/2*pi*(x-(xˆ2+1)ˆ(1/2))/x
%
Note that the integrand contains x02 , so the results for positive and negative x0 are
identical. Therefore in the integrated result, the x0 should be replaced by |x0 |. The
result is
S π 0 2 1/2
0
v x0 = π x0 1 + x − x (10.31)
2 |x0 |
Finally, substitute for x0 and we arrive at
x S
" 1/2 #
x2 |x|
v x0 = 1+ 2

− d (10.32)
|x| 2 d
This matches [81, equation (8-110)].
One can now make a direct comparison between the displacement versus distance for
the two slip functions, to note their similarities and differences. However, note that the
arctangent slip function will have a larger seismic moment (i.e., slip integrated over
depth) than the crack model slip function. The magnitude of the difference is found
by integrating the slip versus depth for the two cases. For the arctangent function the
integrated slip is simply S d. For the crack model the integrated slip is
Z1
1/2 π
Sd (1 − z2 ) dz = Sd 4 (10.33)
0
Figure 10.6 compares the two displacement functions when the depth of faulting for
Figure 10.6
the arctangent model is reduced by π/4, so the moments are matched; at this scale the
plots are nearly identical. This illustrates the fact that measurements of displacement
versus distance across a fault are not very sensitive to the shape of the slip versus
depth function although they do provide an important constraint on the overall seismic
moment. In the next section we highlight this issue that geodetic measurements of
surface displacement are relatively insensitive to the shape of the slip (versus depth
function), but provide a good estimate of the overall seismic moment.
10.2 Geodetic Moment Accumulation Rate
The geodetic moment accumulation rate M per unit length of fault L is given by the
well known formula
M
L
= µS D (10.34)
where D is the thickness of the locked zone and S is the slip deficit rate or backslip
rate used in block models. This is the standard formula provided in all the seismology
textbooks although they usually consider the co-seismic moment release due to co-
seismic slip. Here we are considering the gradual accumulation of geodetic moment
during the interseismic period. These moments must balance over many earthquake
cycles. In the general case where the slip rate s varies with depth z, the moment rate M
is given by
Z0
M
=µ s (z) dz (10.35)
L
−Dm
where Dm is the maximum slip depth. The objective of the following analysis is to
show that the total moment rate per unit length of fault can be measured directly from
geodetic data; no slip vs. depth model is needed. The only assumptions are that the
strike-slip fault is 2-D and the earth behaves as an elastic half space. From the disloca-
tion theory developed in equation (10.26), the y-velocity as a function of distance from
the fault is given by
Z0
1 s(z)x
ν (x) = dz (10.36)
π x 2 + z2
−Dm
Next we guess that the integral of the displacement rate times distance from the x-origin
is a proxy for the moment accumulation rate. We call this proxy Q and later show how
it is related to the moment rate M. We integrate to an upper limit W and then take the
limit as W → ∞:
 
 ZW ZW Z0
 
 2 
1  1 x

Q = lim  xν (x) dx = lim  s(z) 2 dz dx  (10.37)

W→∞  W  W→∞  πW x + z2 
0 0 −Dm
After rearranging the order of integration one finds

Z0  ZW
 
2
1 1 x

Q= s(z) lim  dx  dz (10.38)
 
π W→∞  W x2 + z2 
−Dm 0
The integral over x can be done analytically:

ZW
1 x2 x z
W
−1 x z W
dx = − tan = 1 − W tan−1 z (10.39)
W x 2 + z2 W W z o
0
Figure 10.7: Schematic of surface velocity due to uniform backslip rate over a depth
D.
In the limit as W → ∞ the second term on the right side is zero because z has an upper
bound of Dm , so the total integral is simply 1. Overall we find this proxy is
Z0
1
Q= s(z) dz (10.40)
π
−Dm
Comparing equation (10.37) with equation (10.40) it is clear that the geodetic moment
can be directly related to the integral of the displacement times the distance from the
origin. Note we have extended the integral to both sides of the fault to enable the use
of geodetic measurements on both sides.
 ZW
 
M µπ

= lim  xν (x) dx (10.41)
 
L W→∞  W 
−W
As a check we can insert equation (10.36) into equation (10.41) and make sure we
arrive at equation (10.35)
   
 ZW  Z0
 µπ
 
M  1 s (z) x  
= lim   x  dz dx
L W→∞  W π x2 + z2  
−W −Dm
(10.42)
Z0 ZW
 
2
1 x
 
=µ s (z)  lim dx  dz
 
W→∞ W x2 + z2 
−Dm −W
We perform the integral over x first and multiply by 2 after changing the limits because
the integrand is symmetric about x = 0.
ZW
x2 W
−1 x
2 dx = 2 x − z tan (10.43)
x 2 + z2 z 0
0
In the limit as W → ∞ the final result is

2 W
lim W W − z tan−1 z = 2, for zmax = Dm W (10.44)
W→∞
The moment accumulation rate is

Z0
M
= 2µ s (z) dz (10.45)
L
−Dm
This agrees with our original estimate of moment except for a factor of 2. The corrected
formula for the moment accumulation rate is
ZW
M µπ
= lim xν (x) dx (10.46)
L W→∞ 2W
−W
The main utility of this formula is to demonstrate that geodetic measurements of y-

displacement rate rate across an infinitely long strike slip fault provide a direct estimate
of the geodetic moment rate. It is unnecessary to attempt the unstable inverse problem
to calculate slip versus depth and then integrate this function.
10.3 Inclined Fault Plane
Now consider a model where the fault plane is not perpendicular to the free surface
of the earth. The angle α between the vertical and the fault plane will introduce an
Figure 10.8
asymmetry in the model. Later we’ll consider the effect of frictional heating on an
inclined fault on the surface heat flux.
To develop this solution we’ll start with the surface displacement due to a screw dislo-
cation. We’ll integrate over depth and rotate from the inclined frame into the horizontal
frame. Finally we’ll introduce the image source to reconcile the free surface boundary
condition.
From equation (10.17) we have
−A x0
v(x0 , z0 ) = h i. (10.47)
2πµ x0 2 + (z0 + a)2
The rotation from the x, z frame to the x0 , z0 frame is
x0 = x cos α + z sin α
(10.48)
z0 = −x sin α + z cos α
Also note that D = D0 cos α. As before, consider free slip between a depth of −D0 and
minus infinity.
Z−D0
V x0
v(x0 , z0 ) = 2π da0 (10.49)
x2 + (z0 + a0 )2
−∞
Let η = z + a , so dη = da .
0 0 0
Z0 −D0
−z
V x0
v(x0 , z0 ) = dη (10.50)
2π x02 + η2
−∞
We have performed this integration before (equations (10.19)–(10.20)) so it is not re-

peated here. The result is
0
V x
v(x0 , z0 ) = 2π tan−1 D0 + z0 (10.51)
To match the surface boundary condition, we introduce an image source extending

from +D0 to infinity, but along an image fault inclined at an angle of −α with respect
to the vertical. The displacement from the image is
00
!
00 00 V x
vimage (x , z ) = 2π tan −1
00 00 (10.52)
D −z
Finally, combining the source and the image and substituting x and z we find
( )
x cos α + z sin α x cos α − z sin α

V
v(x, z) = 2π tan −1
D0 − x sin α + z cos α
+ tan −1
D0 − x sin α − z cos α
(10.53)
Now calculate the displacement at z = 0 and substitute D0 = D/cos α.

x cos2 α

v(x) = Vπ tan−1 D − x sin α cos α (10.54)
If one plots this solution there are two differences from the vertical strike-slip fault
case. First, the displacement pattern is shifted along the x-axis by an amount D tan α.
Therefore one can identify a dipping fault by recognizing that the position of the fault
based on geodetic measurements is shifted from the position of the fault trace based on
field geology.
The second difference is that the solution given in equation (10.54) solution does not
match the far-field boundary conditions of ±V/2. The hanging wall has more displace-
ment than the foot wall. In the extreme case of a near horizontal fault plane, the hanging
wall has the full displacement +V while the foot wall has none. This is to be expected
because in our model is driven by a force couple. One can “correct” this asymmetry by
subtracting a constant α from the arctangent in equation (10.54). It is left as an exercise
for the reader to show the final solution is
x cos2 α

V
v(x) = π tan−1 D − x sin α cos α − α (10.55)
We see that for α = 0, this matches the previous solution, equation (10.22). Also, we
can superimpose several of these solutions to simulate any combination of shallow and
deep slip. The stress is the shear modulus times the x-derivative of the displacement.
After a little algebra one finds:
µV
" 2 #−1
x cos2 α x cos3 α sin α

τ xy = 1+ cos2
α + (10.56)
πDα Dα Dα
where Dα = D − x sin α cos α.

10.4 MATLAB Examples
The first example is a MATLAB program to calculate the strain and displacement fields
due to a vertical strike-slip fault with free-slip on both shallow and deep fault planes.
%
% program to generate displacement and strain for a screw
% dislocation. fault slip occurs both shallow and deep.
%
clear
clf
hold off
%%
V=-.01;
D=12000.;
d=800.;
d0=200;
x = -40000:8:40000;
xp = x/1000.;
%
% this first model has shallow creep between depths of d0 and d
%
v1 = (V/pi)*(atan(x/d0)-atan(x/d));
dv1 = (V/(pi*d0))*1./(1.+(x/d0).ˆ2) - (V/(pi*d))*1./(1.+(x/d).ˆ2);
%
% this second model has free-slip for depths greater than D.
%
v2 = (V/pi)*atan(x/D);
dv2 = (V/(pi*D))*1./(1.+(x/D).ˆ2);
%
subplot(2,1,2);plot(xp,(v1+v2)*1000,xp,v2*1000,’:’);
xlabel(’distance (km)’);
ylabel(’displacement (mm/a)’)
subplot(2,1,1);plot(xp,1.e6*(dv1+dv2),xp,1.e6*dv2,’:’);
ylabel(’strain (microradian/a)’); axis([-40,40,-3,1])
See Figure 10.9

Figure 10.9
The second example is a MATLAB program to illustrate the effect of fault dip that
simply shifts the arctangent function by an amount D tan α. In this example the shift
is 6.9 km.
%
% Compute the displacement due to a dipping fault using equation (34).
% Note the function atan2() must be used.
%
V=-10;
alph=30*pi/180.;
D=12;
x=-40:40;
%
%
cosa=cos(alph);
sina=sin(alph);
num=x.*cosa*cosa;
dem=D-x.*sina*cosa;
vel0=V*atan2(x,D)/pi;
vel1=V*(atan2(num,dem)/pi-alph/pi);
subplot(2,1,1);plot(x,vel0,x,vel1,’--’);
xlabel(’distance (km)’);ylabel(’displacement (mm/a)’);
title(’dipping 30 degrees in positive x-direction’)
grid
%
See Figure 10.10.
Figure 10.10
Chapter 11
Heat Flow Paradox
(Reference: [6] [37])
11.1 Paradox
The seismogenic zone extends from the surface to a depth of about 10 km. According
to Byerlee’s law, the shear stress on the fault should be some fraction of the hydrostatic
stress:
τ(z) = f ( ρc − ρw ) gz (11.1)
f static coefficient of friction ∼0.60
ρc crustal density 2600 kg m−3
ρw water density 1000 kg m−3
D depth of seismographic zone 12 km
This assumes that water percolates to 12 km depths to lower friction on the fault. We
can compute the average shear stress on the fault.
ZD
1
τ= D
f ( ρc − ρw ) gz dz = 1
2 f ( ρc − ρw ) gD = 56 MPa (11.2)
o
136
CHAPTER 11. HEAT FLOW PARADOX 137
The observed stress drop during an earthquake ranges from 0.1 to 10 MPa with a typical
value of 5 MPa, which is about 10 times smaller than the average stress from Byerlee’s
Law [6]. This implies that only a fraction of the total stress is released during an earth-
quake. The average stress during the earthquake times the earthquake displacement
produces energy both as seismic radiation (small fraction) and as heat (large fraction).
If this heat energy is averaged over many earthquake cycles, then this average heat/area
generated on the fault plane will appear as a heat flow anomaly on the surface having a
similar heat/area as along the fault.
To calculate this heat anomaly for a variety of frictional heating models, first consider
a line source of heat. (See Figure 11.1 .) The differential equation and boundary
z
2
y
-V/2 V/2
x
Q(x,z) = Vτ(z)δ(x)
The differential equation and boundary conditions for a unit-amplitude, line source at depth -a is
Figure 11.1
1 1
conditions∇2for Q( x, z ) = δ ( x )δ ( zline
T =a unit-amplitude, + a)source at depth −a is (3)
k k
1 1
∇2 T = k Q(x, z) = k δ(x) δ(z + a) (11.3)
T ( x, 0) = 0
lim T ( x, z ) = 0
z →∞
lim T ( x, z ) = 0 T (x, o) = 0
x →∞
lim T (x, z) = 0
where T is the temperature anomaly|z|→∞ in ˚K, k is the thermal conductivity (3.3 Wm -1˚K-1), and Q is
the heat generation in Wm-3. Note this is the same differential equation as equation (5) of the last
lim
section. The only difference is|x|→∞the surface =0
T (x, z)boundary condition. The surface stress problem has
vanishing shear stress at the surface (i.e., vertical derivative of displacement is v is zero) so we
−1◦ −1
where T is the temperature
introduced a positiveanomaly in ◦ K,
image source k is the
to force thethermal conductivity
displacement field to be(3.3Wm
symmetric K
about) z = 0.
−3
and Q is Inthethisheat
heatgeneration
flow case, we
in have
Wmvanishing
. Notetemperature
this is theanomaly at the surface
same differential so we introduce
equation as a
equation negative
(10.5). lineTheheat
onlysource at z = a is
difference to form an anti-symmetric
the surface boundarytemperature
condition.function. The solution
The surface
to the full-space problem is identical to equation (14) of the previous section.
stress problem has vanishing shear stress at the surface (i.e., vertical derivative of dis-
placement of v is−zero), so we introduced a positive image source to force the dis-
placementT (field
1
2πk
[ ]
2 1/ 2
x 2 + ( z + a)about
x, z ) =to belnsymmetric z = 0. (4)vanishing
In this heat flow case, we have
After including the image source, the result is
T ( x, z ) =
−1
2πk { [
ln x 2 + ( z + a)
2
]
1/ 2
[
− ln x 2 + ( z − a) ]
2 1/ 2
} (5)
temperature anomaly at the surface, so we introduce a negative line heat source at z = a

to form an anti-symmetric temperature function. The solution to the full-space problem
is identical to equation (10.14).
−1 h i1/2
T (x, z) = 2πk ln x2 + (z + a)2 (11.4)
After including the image source, the result is

h i1/2 i1/2
−1 h
T (x, z) = 2πk ln x2 + (z + a)2 − ln x2 + (z − a)2 (11.5)
The quantity of interest is the surface heat flow versus distance from the fault.
δT 1 δ
h i1/2 h i1/2
q(x, z) = −k δz = 2π ln x2 + (z + a)2 − ln x2 + (z − a)2 (11.6)
δz
After a little algebra one arrives at the heat flow:
(z + a)
( )
1 (z − a)
q(x, z) = 2π − (11.7)
x2 + (z + a)2 x2 + (z − a)2
Thus the surface heat flow for a line source of unit strength at depth a is
1 a
q(x) = π 2 (11.8)
x + a2
For an arbitrary shear stress distribution with depth τ(z) the surface heat flow is
Z0
V z τ(z)
q(x) = π dz (11.9)
x2 + z2
−∞
Now let’s assume that the stress follows equation 11.1, Byerlees’s Law (i.e., high stress
and high heat flow). Also allow hydrothermal circulation to extend from the surface to
some depth d, which effectively removes all the heat produced between the surface and
that depth. The integration is
Z−d
f ( ρc − ρw ) gV z2
q(x) = π
dz (11.10)
x2 + z2
−D
This integral is done with help from the table of integrals;

Z Z
x2 x a 1
dx = − 2 dx (11.11)
a + bx
2 b b a + bx
After some algebra one arrives at the following analytic formula for the heat flow
f (ρc − ρw )gV

d D
q(x) = π
(D − d) + x tan−1 − x tan−1
x x
(11.12)
It is interesting to compare this heat flux to the heat flux at a mid-ocean ridge for the
same total opening rate V (see Figure 11.2). The formula is
q(x) = k(T m − T o ) (2πκx/V)−1/2 (11.13)
Tm mantle temperature 1600 ◦ K
To surface temperature 273 ◦ K
k thermal conductivity 3.3 Wm−1 ◦ K−1
κ thermal diffusivity 8. × 10−7 m2 s−1
11.1.1 MATLAB Example
The following is a MATLAB program simulates a high-stress fault (i.e., Byerlee’s Law)
extending to a depth of 12 km and sliding at a rate of 30 mm/yr. Two cases are con-
sidered; the first case (the solid curve in Figure 11.2) has hydrothermal heat removal
extending to a depth of 1 km, while the second (the dotted curve in the same figure)
has heat removal to a depth of 5 km. These models are compared with the heat flow
measurements across the San Andreas Fault [37]. It is clear that the shallow heat re-
moval model is inconsistent with the data. However, the deep heat removal model is
not precluded by the observations, especially if the background level of the model heat
flow is allowed to vary from the spatial average. One argument against hydrothermal
removal of heat is the absence of hot springs along the fault with sufficient vigor to
remove this heat. Hydrothermal circulation is the dominant heat removal mechanism
at the mid-ocean ridges and hydrothermal vents are common. However, as shown in
Figure 11.2, heat generation along a strike-slip fault is 2–3 orders of magnitude less
than a mid-ocean ridge, so it is not clear that the same mechanism should operate at
a fault. Even if heat loss is concentrated in small areas it may be difficult to detect at
the surface.
%
% program to calculate the surface heat flux due to frictional
% heating on a strike-slip fault
D=12;
d1=1; d5=5;
rc=2600; rw=1000; g=9.8;
V=.03/3.15e7; f=.60;
q0=1.e6*f*(rc-rw)*g*V/pi;
%
% calculate the heat flow for the two models of shallow
% and deep heat removal
x=-60:.1:60;
q1=q0*((D-d1)+x.*atan(d1./x)-x.*atan(D./x));
140
120
heat flow (mWm-2 ) 100
80
60
40
20
0
100 50 0 50 100
distance (km)
104
heat flow (mWm-2)
mid-ocean ridge
102
strike-slip fault
100
60 40 20 0 20 40 60
distance (km)
Figure 11.2
q5=q0*((D-d5)+x.*atan(d5./x)-x.*atan(D./x));
% plot the results
plot(x,q1+73,x,q5+73,’:’);
xlabel(’distance (km)’);
ylabel(’heat flow (mWm-2)’)
axis([-120,120,0,167]);
11.2 Moment Paradox: Seismic Moment

versus Tectonic Saturation Moment
The Moment Paradox described next is really part of the heat-flow paradox, except it
is expressed in a different way. As discussed in a previous lecture, and in Brace and
Kohlstedt [4], measurements of stress difference in the uppermost crust to depths of
several kilometers are consistent with a yield strength model following Byerlee’s Law.
The static frictional resistance to sliding is related to a coefficient of friction f of about
0.60 times the overburden pressure of ∆ρgz. This leads to differential stress difference
of 140 MPa at a depth of only 10 km. We also found that these high stresses are required
to support the 5000 m elevation of Tibet relative to India. This isostatic model is the
minimum stress needed to support topography, so it is clear that high stresses exist at
shallow depths in the crust. Similarly, one can calculate the bending moment needed
to support the trench and outer rise topography at a subduction zone. The moment
calculation is model-independent [45]:
Z∞
M(xo )/L = g∆ρw(x) (x − xo ) dx (11.14)
xo
where M is the moment per unit length along the trench, ∆ρ is the mantle-to-seawater
density contrast, w(x) is the height of the outer rise above the normal depth and x− xo is
the distance between the first zero crossing of the trench flexure profile and some point
out on the outer rise. The integral converges because w(x) goes to zero exponentially
with distance. Observed bending moments at outer rises vary from 5 × 1016 N for
young lithosphere (10 Ma) to 3 × 1017 at old oceanic lithosphere (140 Ma) [40]. Next
we’ll compare these numbers to typical seismic moments of large earthquakes using
the Alaska 1964 and Landers 1992 events as examples. The Landers rupture was about
70 km long, so we’ll divide its moment by length for the comparison with models. The
results are provided in Table 11.1.
This comparison highlights two issues: first, the moment of the Alaska 1964 earthquake
was sufficient to cause a collapse of the outer rise; second, the seismic/geodetic moment
of the Landers 1992 earthquake is 10–20 times smaller than the moment estimated next
using a the simple elastic dislocation model, where stress is limited only by Byerlee’s
Law.
11.2.1 Seismic Moment Released During an Earthquake
The moment released during an earthquake can be estimated in two ways, either by
analysis of the seismic radiation pattern or by the geodetic analysis of the geodetic
ground motion. They usually provide similar values, although in the case of the Lan-
ders 1992 rupture, the seismic moment estimate is about 2 times the geodetic moment
Tectonic Example Moment per Length (N)
Outer rise flexure (10 Ma) 5 × 1016 N
Outer rise flexure (140 Ma) 3 × 1017 N
Alaska flexure 1.2 × 1017 N
Alaska 1964 earthquake, M9.2 1.1 × 1017 N = 8.2 × 1022 /750km
Landers 1992 earthquake, M7.4 1.4 × 1015 N geodetic/length
2.8 × 1015 N seismic/length
Byerlee’s criterion (0–12 km only) 1.3 × 1016 N m
Table 11.1
estimate. The moment is defined as
M s = µLD∆y (11.15)
f static coefficient of friction ∼0.60

ρc crustal density 2600 kg m−3
ρw water density 1000 kg m−3
µ shear modulus 2.6 × 1010 Pa
L length of rupture 70 km
D depth of rupture 12 km
∆y rupture offset 4.5 m
V plate velocity 0.0015 m/yr
t earthquake recurrence interval
Now we can do some simple calculations. First, a check on the geodetic moment
1.4×1015 N provides a match to the published value. The recurrence interval of ∆y/V =
3000 years seems OK for a fault out in the Mojave Desert away from the San Andreas
Fault. So everything seems consistent. Next let’s assume that the stress on the fault,
as a function of depth, matches Byerlee’s law for the case of hydrostatic pore pressure.
We’ll compare this saturation moment and recurrence interval with the observations
from earthquakes.
Landers 1992 earthquake, M7.4 1.4 x 1015 N geodetic/length
2.8 x 1015 N seismic/length
Byerlee's criterion (0 - 12 km only) 1.3 x 1016 N m
This comparison highlights two issues: first, the moment of the Alaska 1964 earthquake was
sufficient to cause a collapse of the outer rise(??); second, the seismic/geodetic moment of the
Landers 1992 earthquake is 10-20 times smaller than the moment estimated next using a the
simple elastic dislocation model where stress is limited only by Byerlee's law.
Seismic Moment Released During an Earthquake
L y
x
D
Figure 11.3
The
11.3of the moment released
Tectonic during anMoment
Saturation earthquake can be estimated in two ways, either by analysis
seismic radiation pattern or by the geodetic analysis of the geodetic ground motion. They
usually provide similar values; although in the case of the Landers 1992 rupture, the seismic
moment
Assume estimate
that the simpleishalf-space
about 2 times the geodetic
solution moment
(developed above)estimate.
providesThe
the moment
stress andis defined as
strain field for a fault locked from the surface to a depth D. Further, assume that the
maximum µLD∆that
Ms =stress y can be maintained on a fault is given by Byerlee’s Law (2)
h i
f τ(z)
static coefficient of =
friction ρw ) gz + τn
f − (ρc ~− 0.60 (11.16)
whereρτcn is the
-3
crustal density
additional 2600applied
tectonic normal stress kg m to the fault plane. The tectonic
ρ water density
moment per unit length is given by
w 1000 kg m-3
g acceleration of gravity 9.8 m s-2
µ shear modulus Z0 x 1010 Pa
2.6
L length of rupture M T /L = ∆y 70 kmµ(z) dz (11.17)
D depth of rupture 12
−Dkm
What is µ(z)? This is the effective shear modulus needed to keep the stress below the
upper bound provided by Byerlee’s law, so
τ(z) ∂v(x, z)
µ(z) = where ε(z) = (11.18)
ε(z) ∂x
Now assume that v(x, z) is provided by the interseismic strain solution developed in the
previous chapter (equation 10.18). It is left as an exercise to finish the problem. You
will find that ε(z) is proportional to ∆y, so this factor cancels in equation 11.17. The
final result is 
 ρc − ρw gD 2 

MT
L
= f πD 
2
4
+ 3 τn  (11.19)
Using the values in Table 11.1 and for zero normal stress, we find the saturation mo-
ment per unit length is 1.3 × 1016 N. Again, this is 10 times larger than the observed
moment. Given the fault parameters above, this moment implies a potential seismic
offset of 45 m and a recurrence time of 3000 years—a giant earthquake indeed!
There are only two ways to understand this dilemma:

1. Faults are somehow lubricated ( f ∼.05), so the average stress on the fault is 10–20
times smaller than predicted by Byerlee’s law. In this case one has the difficulty
of maintaining the elevation of the topography in California. For example, San
Jacinto Mountain, which is less than 25 km from the San Andreas Fault, has a
relief of about 3000 m, which implies stresses of 80 MPa (16 times the stress
drop in an earthquake).
2. Faults are strong as predicted by Byerlee’s Law. In this case, faults are always
very close to failure and each earthquake relieves only a small fraction (∼10%)
of the tectonic stress. As we saw in the last section, this model implies a large
amount of energy dissipation along the fault; friction from both aseismic creep
and seismic rupture will generate heat. It has been proposed that perhaps during
the earthquake, the coefficient of friction drops from 0.60 to, say, 0.05, to tem-
porarily disable the heat generation. However, it seems that such a slippery fault
would release all of the elastic energy during an earthquake (∼60 m of offset).
Another possibility is that heat is generated but a large fraction of the heat is ad-
vected to the surface by circulation of water in the upper couple km of crust. The
unfortunate implication of this high-stress model is that since faults are always
close to failure, it will be almost impossible to predict earthquakes.
11.4 Exercises
Exercise 11.1. (a) Provide an approximate formula for the magnitude of the shear
stress that is needed to induce slip on a dry fault at 10 km depth in continental crust
(density 2800 kg m−3 )? Which parameter is least well known and what is a possible
range for this parameter.
(b) Suppose the crust is saturated with water to 10 km depth. How does this change the
stress magnitude?
Chapter 12
The Gravity Field of the Earth,

Part 1
12.1 Introduction
This chapter covers physical geodesy, the shape of the Earth and its gravity field. This
is just electrostatic theory applied to the Earth. Unlike electrostatics, geodesy is a
nightmare of unusual equations, unusual notation, and confusing conventions. There is
no clear and concise book on the topic, although Chapter 5 of T&S [81, ] is OK.
The things that make physical geodesy messy include:
• Earth rotation,
• latitude is measured from the equator instead of the pole,

• latitude is not the angle from the equator but is referred to the ellipsoid,
• elevation is measured from a theoretical surface called the geoid,
• spherical harmonics are defined differently from standard usage,
• anomalies are defined with respect to an ellipsoid having parameters that are
constantly being updated,
• there are many types of anomalies related to various derivatives of the potential,
• and MKS units are not commonly used in the literature.
145
CHAPTER 12. THE GRAVITY FIELD OF THE EARTH, PART 1 146
In the next couple of chapters, I’ll try to present this material with as much simplifica-
tion as possible. Part of the reason for the mess is that prior to the launch of artificial
satellites, measurements of elevation and gravitational acceleration were all done on the
surface of the Earth (land or sea). Since the shape of the Earth is linked to variations
in gravitational potential, measurements of acceleration were linked to position mea-
surements both physically and in the mathematics. Satellite measurements are made in
space well above the complications of the surface of the Earth, so most of these prob-
lems disappear. Here are the two most important issues related to old-style geodesy.
12.1.1 Elevation
Prior to satellites and the global positioning system (GPS), elevation was measured
with respect to sea level, orthometric height. See Figure 12.1. Indeed, elevation is still
Figure 12.1
defined in this way however, most measurements are made with GPS. The pre-satellite
approach to measuring elevation is called leveling.
1. Start at sea level and call this zero elevation. (If there were no winds, currents,
and tides then the ocean surface would be an equipotential surface and all shore-
lines would be at exactly the same potential.)
2. Sight a line inland perpendicular to a plumb line. Note that this plumb line will
be perpendicular to the equipotential surface and thus is not pointed toward the
geocenter.
3. Measure the height difference and then move the setup inland and to repeat the
measurements until you reach the next shoreline. If all measurements are correct
you will be back to zero elevation assuming the ocean surface is an equipotential
surface.
With artificial satellites measuring geometric height is easier, especially if one is far
from a coastline.
1. Calculate the x, y, z position of each GPS satellite in the constellation using a

global tracking network.
2. Measure the travel time to 4 or more satellites, 3 for position and one for clock
error.
3. Establish your x, y, z position and convert this to height above the spheroid,
which we’ll define below.
4. Go to a table of geoid height and subtract the local geoid height to get the ortho-
metric height used by all surveyors and mappers.
Orthometric heights are useful because water flows downhill in this system, while it
does not always flow downhill in the geometric height system. Of course the problem
with orthometric heights is that they are very difficult to measure, or one must have a
precise measurement of geoid height. Let geodesists worry about these issues.
12.1.2 Gravity
The second complication in the pre-satellite geodesy is the measurement of gravity.

Interpretation of surface gravity measurement is either difficult or trivial depending
whether you are on land or at sea, respectively. Consider the land case illustrated in
Figure 12.2.
Figure 12.2
Small variations in the acceleration of gravity (<106 g) can be measured on the land
surface. The major problem is that when the measurement is made in a valley, there
are masses above the observation plane. Thus, bringing the gravity measurement to
a common level requires knowledge of the mass distribution above the observation
point. This requires knowledge of both the geometric topography and the 3-D density.
We could assume a constant density and use leveling to get the orthometric height but
we need to convert to geometric height to do the gravity correction. To calculate the

geometric height we need to know the geoid. However, the geoid height measurement
comes from the gravity measurement, so there is no exact solution. Of course one can
make some approximations to get around this dilemma, but it is still a problem and
this is the fundamental reason why many geodesy books are so complicated. Of course
if one could make measurements of both the gravity and topography on a plane (or
sphere) above all of the topography, our troubles would be over.
Ocean gravity measurements are much less of a problem because the ocean surface is
nearly equal to the geoid so we can simply define the ocean gravity measurement as
free-air gravity. We’ll get back to all of this again later when we discuss flat-Earth
approximations for gravity analysis.
12.2 Global Gravity
(References: [81, Chapter 5, 5.1–5.5 ], [77, Chapters 3-4], [35, Chapter 3], [25, Chapter
5].)
The gravity field of the Earth can be decomposed as follows:
• the main field due to the total mass of the Earth,

• the second harmonic due to the flattening of the Earth by rotation,
• and anomalies which can be expanded in spherical harmonics or Fourier series.
The combined main field and the second harmonic make up the reference Earth model
(i.e., spheroid, the reference potential, and the reference gravity). Deviations from
this reference model are called elevation, geoid height, deflections of the vertical, and
gravity anomalies.
12.2.1 Spherical Earth Model
The spherical Earth model is a good point to define some geodetic terms. There are
both fundamental constants and derived quantities. See Table 12.1.
We should say a little more about units. Deviations in acceleration from the reference
model, described next, are measured in units of milligal (1 mGal = 10−3 cm s−2 =
10−5 m s−2 = 10 gravity units (gu)). As noted above the vertical gravity gradient is also
called the free-air correction, since it is the first term in the Taylor series expansion for
Parameter Description Formula Value/Unit
Re mean radius of Earth - 6371000 m
Me mass of Earth - 5.98 × 1024 kg
G gravitational constant - 6.67 × 10−11 m3 kg−1 s−2
ρ mean density of Earth Me (4/3πR3e )−1 5520 kg m−3
U mean potential energy −GMe R−1

e −6.26 × 10−7 m2 s−2
needed to take a unit
mass from the surface
of the Earth and place
it at infinite distance
g mean acceleration on −δU/δr = −GMe R−2

e −9.82m s−2
the surface of the
Earth
δg/δr gravity gradient or δg/δr = −2GMe R−3

e 3.086 × 10−6 s−2
free-air correction = −2g/Re
Table 12.1
gravity about the radius of the Earth.

∂g
g(r) = g(Re ) + ∂r (r − Re ) + · · · (12.1)
Example 12.1. How does one measure the mass of the Earth? The best method is
to time the orbital period of an artificial satellite. Indeed measurements of all long-
wavelength gravitational deviations from the reference model are best done be satel-
lites. See Figure 12.3.
GMe m mv2
=
r2 r
(12.2)
GMe
= rω s 2 ⇒ GMe = r3 ω s 2
r2
The mass is in orbit about the center of the Earth, so the outward centrifugal force is
balanced by the inward gravitational force; this is Kepler’s third law. If we measure the
radius of the satellite orbit r and the orbital frequency ω s , we can estimate GMe . For
example, the satellite Geosat has a orbital radius of 7168 km and a period of 6037.55
Figure 12.3
sec, so GMe is 3.988708 × 1014 m3 s−2 . Note that the product GMe is tightly constrained
by the observations, but that the accuracy of the mass of the Earth Me is related to the
accuracy of the measurement of G.
12.2.2 Ellipsoidal Earth Model
The centrifugal effect of the Earth’s rotation causes an equatorial bulge that is the prin-
cipal departure of the Earth from a spherical shape. If the Earth behaved like a fluid and
there were no convective fluid motions, then it would be in hydrostatic equilibrium and
the Earth would assume the shape of an ellipsoid of revolution also called the spheroid.
See Figure 12.4 and Table 12.2.
a
c
{ θ θg
{
Figure 12.4
Parameter Description Formula Value/Unit (WGS84)
GMe 3.986004418 × 1014 m3 s−2

a equatorial radius - 6378137 m
c polar radius - 6356752.3 m
ω rotation rate - 7.292115 × 10−5 rad s−1
f flattening f = (a − c)/a 1/298.257223563
θg geographic latitude - -
θ geocentric latitude - -
Table 12.2
z
x
θ
y
φ
Figure 12.5
The formula for an ellipse in Cartesian coordinates is
x2 y2 z2
+ + =1 (12.3)
a2 a2 c2
where the x-axis is in the equatorial plane at zero longitude (Greenwich), the y-axis is
in the equatorial plane and at 90◦ E longitude, and the z-axis points along the spin-axis.
The formula relating x, y, and z to geocentric latitude and longitude is (see Figure 12.5)
x = r cos θ cos φ
y = r cos θ sin φ (12.4)
z = r sin θ
Now we can rewrite the formula for the ellipse in polar coordinates and solve for the
radius of the ellipse as a function of geocentric latitude:
−1/2
cos θ sin2 θ
2
r= 2 + 2 a 1 − f sin2 θ (12.5)
a c
Before satellites were available for geodetic work, one would establish geographic lati-
tude by measuring the angle between a local plumb line and an external reference point,
such as the star Polaris. Since the local plumb line is perpendicular to the spheroid (i.e.,
local flattened surface of the Earth), it points to one of the foci of the ellipse. The con-
version between geocentric and geographic latitude is straightforward and its derivation
is left as an exercise. The formulas are
c2
tan θ = tan θg or tan θ = (1 − f )2 tan θg (12.6)
a2
Example 12.2. What is the geocentric latitude at a geographic latitude of 45◦ ? The
answer is θ = 44.8◦, which amounts to a 22 km difference in location!
12.2.3 Flattening of the Earth by Rotation
Suppose the Earth is a rotating, self gravitating ball of fluid in hydrostatic equilibrium.
Then density will increase with increasing depth and surfaces of constant pressure and
density will coincide. The surface of the Earth will be one of these equipotential sur-
faces; it has a potential Uo :
Uo = V(r, θ) − 12 ω2 r2 cos2 θ (12.7)
where the second term on the right side of equation (12.7) is the change potential due
to the rotation of the Earth at a frequency ω. The potential due to an ellipsoidal Earth
in a non-rotating frame can be expressed as
a 2
GM a
V = − r e 1 − J1 r P1 (θ) − J2 r P2 (θ) − · · · (12.8)
The center of the coordinate system is selected to coincide with the center of mass so,
by definition, J1 is zero. For this model, we keep only J2 (dynamic form factor or
“jay two” = 1.08 × 10−3 ), so the final reference model is
GM GMe J2 a2
V=− re + 3 3 sin2 θ − 1 (12.9)
2r
This parameter J2 is related to the principal moments of inertia of the Earth by Mac-
Cullagh’s formula. For a complete derivation see [77, Chapter 3]. Let C and A be the
moments of intertia about the spin axis and equatorial axis, respectively. For example,
Z
C= x2 + y2 dm (12.10)
V
After a lot of algebra one can derive a relationship between J2 and the moments
of inertia:
C−A
J2 = (12.11)
Ma2
In addition, if we know J2 , we can determine the flattening. This is done by inserting
equation (12.9) into equation (12.7) and noting that the value of Uo is the same at the
equator and the pole. Solving for the polar and equatorial radii that meet this constraint,
one finds a relationship between J2 and the flattening:
1a ω
3 2
a−c 3
f = = 2 J2 + 2 (12.12)
a GM
Thus if we could somehow measure J2 , we would know quite a bit about our planet.
12.2.4 Measurement of J2
Just as in the case of measuring the total mass of the Earth, the best way to measure is
to monitor the orbit of an artificial satellite. In this case we measure the precessional
period of the inclined orbit plane. To the second degree, external potential is
GM GMe J2 a2
V=− re + 3 sin2 θ − 1 (12.13)
2r3
The force acting on the satellite is −∇V.

∂V 1 ∂V 1 ∂V
g = − ∂r r̂ − θ̂ − φ̂ (12.14)
r ∂θ r cos θ ∂φ
If we were out in space, the best way to measure J2 would be to measure the θ̂-
component of the gravity force:
1 ∂V 3GMe a2 J2
gθ = = − sin θ cos θ (12.15)
r ∂θ r4
This component of force will apply a torque to the orbital angular momentum and it
should be averaged over the orbit. Consider the diagram, Figure 12.6. For a prograde
orbit, the precession ω p is retrograde; that it is opposite to the Earth’s spin direction.
The complete derivation is found in [77, page 76]. The result is
ω p −3a2
= J2 cos i (12.16)
ωs 2r2
∆L
m
L
g
θ
Figure 12.6: Satellite of mass m orbiting the Earth at an inclination i. The orbit has
an angular momentum vector L perpendicular to the orbital plane. The θ compnent of
the gravity vector applies a torque to the orbit causing a retrograde precession of the
angular momentum vector and thus a precession of the orbital plane.
where i is the inclination of the satellite orbit with respect to the equatorial plane, ω s
is the orbit frequency of the satellite, and ω p is the precession frequency of the orbit
plane in inertial space.
Example 12.3. LAGEOS: As an example, the LAGEOS satellite orbits the Earth every
13673.4 seconds, at an average radius of 12,265 km and an inclination of 109.8◦ . Given
the parameters in Table 12.3 and that J2 = 1.08 × 10−3 , the predicted precession rate is
0.337◦ /day. This can be compared with the observed rate of 0.343◦ /day. Figure 12.7 is
an illustration of the LAGEOS satellite.
12.2.5 Hydrostatic Flattening
Given the radial density structure, the Earth rotation rate and the assumption of hy-
drostatic equilibrium, one can calculate the theoretical flattening of the Earth (see [27,
Appendix 2]). This is called the hydrostatic flattening fh = 1/299.5. From Table 12.3
we have the observed flattening f = 1/298.257, so the actual Earth is flatter than the
theoretical Earth. There are two reasons for this. First, the Earth is still recovering
from the last ice age when the poles were loaded by heavy ice sheets. When the ice
melted, polar dimples remained and the glacial rebound of the viscous mantle is still
incomplete. Second, the mantle is not in hydrostatic equilibrium, because of mantle
convection. Finally it should be noted that J2 is changing with time due to the contin-
Description Value
Semimajor axis 12265 km
Eccentricity 0.004
Inclination 109.8◦
Perigee height 5858 km
Apogee height 5958 km
Perigee rate −0.215◦ /d
Node rate +0.343◦ /d
Semimajor axis decay rate −1.1 mm / d
Oribital acceleration 3 × 10−12 m/s
Table 12.3: LAGEOS Orbital Parameters
ual post-glacial rebound. This is called jay two dot and it can be observed in satellite
orbits as a time variation in the precession rate.
9218 COHEN AND SMITH' LAGEOS SCmNTIFIC RESULTS•INTRODUCTION TO THE SPECIAL ISSUE
LAGEOSASSEMBLY
47.G2 MM
HALF SPHERE CORE RETROFLEC TORS
MATERIAL - ALUMINUM MATERIAL BRASS PATTERN
131.76
DIA
27.50 ½M •
1.905 CM DIA STUD

MATERIAL BRASS
30.00 cm
SPHERICAL RADIUS
Fig. 1. Structuraldetail of LAGEOS satellite.
Figure 12.7: Structural detail of thebyLAGEOS

discussed satellite
Smith and Dunn [1980],
ogy, modeling techniques,and computerprocedureshave im- Afonso et al.[12].
[1980],
Rubincam[1982], Alselmoet al. [1983], and Smith [1983].
proved the accuracyof site determinationsto a few centime-
ters in all coordinates.Contemporary measurementsthen are Features of a special gravity field developedto model the
in the range where important geodynamicand geodeticobser- LAGEOS orbit, GEM L2, is discussed by Lerch et al. [1982].
vations can be made. The use of LAGEOS data to study postglacialrebound has
The Crustal Dynamics Project (CDP) of the National Aero- beenreportedby Yoderet al. [ 1983], Rubincam[1984], Peltier
12.3 Exercises
nautics and Space Administration has the primary respon-
sibility for conducting laser ranging measurements on
LAGEOS. Numerous other domesticand foreign institutions
[1983], and Alexander[1983]. Lerch et al. [1978] have deter-
mined the geocentricgravitationalconstant(GM). A number
of conferenceproceedings contain LAGEOS results.Many of
are cooperatingin data collection,analysis,and interpretation. theseare referencedin the geodesyportion of the 1979-1982
The CDP was formed in 1979 to apply spacetechnology,both U.S. National Report to the InternationalUnion of Geodesy
in the form of satellite laser ranging and very long baseline and Geophysics.
interferometry (see below) to the measurement of tectonic
SUMMARY OF TECHNICAL PAPERS
Exercise 12.1. Derive equation 12.6 using the ellipsoidal Earth model.
plate motions, regional crustal deformations,polar motion
and earth rotation, and other phenomena associatedwith
For descriptivepurposethe paperspresentedin this special
crustal movements. The CDP is contained with NASA's Geo- issueof the Journal of GeophysicalResearchcan be divided
dynamicsProgram which has evolvedfrom earlier earth phys- into nine interrelatedcategoriesthat cut acrossthe traditional
Exercise 12.2. Derive equation 12.12 by following the instructions in
ics programssuchas the National GeodeticSatelliteProgram, earth sciencesdisciplinesof geodesyand tectonophysics
the Earth and Ocean PhysicsApplicationsProgram, and the involve (1) referenceframes for earth dynamics, (2) station
and
the paragraph
LAGEOS Project. coordinatesand intersitebaselines,(3) gravity fields,(4) polar
preceding this This
equation.
specialissueof JGR presentscurrent scientificfindings motions, (5) earth rotation, (6) earth tides, (7) orbital pertur-
based on analysesof LAGEOS data. It brings together work bations, (8) mantle structure, and (9) geodetictechniques.In
on satellite geodesy,gravity, plate tectonics,polar motion, the following paragraphswe will discusseach of thesetopics
and the paperspresentedthat relate to them.
Exercise 12.3. Derive equation 12.16. You may need to refer to the book by Stacey [77,
postglacialrebound, length-of-day,angular momentum,earth
tides, and mantle structure. Before summarizingthesereports The establishmentof a geodynamicreferenceframe is a
it is appropriate to mention some of the papers that have basicstep for virtually all LAGEOS analyses.Both the geo-
Chapter 4, page 76]. already been publishedusingLAGEOS data. The evolution of
the LAGEOS orbit and various orbit perturbationshave been
detic and geophysicalinferencesto be drawn from SLR data
depend on having a well defined coordinate systemwith a
Exercise 12.4. A sun synchronous orbit has a prograde precession rate of once per
year. This type of orbit is used by optical remote sensing satellites to have the same sun
illumumination for every repeat cycle. Use equation 12.16 to calculate the inclination
of the orbit plane needed to match this rate for a satellite in a circular orbit at 800 km.
Exercise 12.5. Assume the density of the Earth is uniform and the earth is a prefect
sphere.
(a) Develop a formula for the gravitational acceleration as a function of radius inside
the Earth and check the dimensions.
(b) Develop a formula for the pressure at the center of this earth and check dimensions.
Chapter 13
Reference Earth Model: WGS84
13.1 Some Definitions
{
ω
c
a θ θg
{
spheroid
Radius of spheroid
−1/2
cos2 θ sin2 θ

r(θ) = + a 1 − f sin2 θ (13.1)
a2 c2
157
CHAPTER 13. REFERENCE EARTH MODEL: WGS84 158
Parameter Description Formula Value/Unit
GMe (WGS84) 3.986004418 × 1014 m3 s−2
Me mass of Earth – 5.98 × 1024 kg
G gravitational constant – 6.67 × 10−11 m3 kg−1 s−2
a equatorial radius – 6378137 m

(WGS84)
c polar radius (derived) – 6356752.3 m
ω rotation rate – 7.292115 × 10−5 rad s−1

(WGS84)
f flattening f = (a − c)/a 1/298.257223563

(WGS84)
J2 dynamic form factor – 1.081874 × 10−3

(derived)
θg geographic latitude – –
θ geocentric latitude – –
Conversion between geocentric θ and geographic θg latitude

c2
tan θ = tan θg or tan θ = (1 − f )2 tan θg (13.2)
a2
Gravitational potential in frame rotating with the Earth

GM GMe J2 a2
Uo = − r e + 3 sin2 θ − 1 − 21 ω2 r2 cos2 θ (13.3)
2r3
Calculation of the second degree harmonic J2 from WGS84 parameters

2 a3 ω2
J2 = 3 f − 3GM (13.4)
e
Calculation of J2 from the polar C and equatorial A moments of inertia

C−A
J2 = (13.5)
Me a2
Kepler’s third law relating orbit frequency, ω s , and radius r to Me
ω2s r3 = GMe (13.6)
Measurement of J2 from orbit frequency ω s radius r, inclination i, and precession

rate ω p
ω p −3a2
= 2 J2 cos i (13.7)
ωs 2r
Hydrostatic flattening is less than observed flattening

1 1
fH = 299.5 < f = 298.257 (13.8)
13.2 Disturbing Potential and Geoid Height
To a first approximation, the reference potential Uo is constant over the surface of the
Earth. Now we are concerned with deviations from this reference potential. This is
called the disturbing potential Φ; over the oceans the anomalous potential results in a
deviation in the surface away from the spheroid:
U = Uo + Φ (13.9)
where the reference potential Uo is given in equation 13.3. The geoid is the equipoten-
tial surface of the Earth that coincides with the sea surface when it is undisturbed by
winds, tides, or currents. The geoid height is the height of the geoid above the spheroid
and it is expressed in meters. Consider the following mass anomaly in the Earth and its
effect on the ocean surface.
Because of the excess mass, the potential on the spheroid is higher than the reference
level U = Uo + Φ. Thus, the ocean surface must move farther from the center of the
Earth to remain at the reference level Uo . To determine how far it moves, expand the
potential in a Taylor series about the radius of the spheroid at ro :
∂U
Uo (r) = U (ro ) + ∂r (r − ro ) + · · · (13.10)
Notice that g = −δU/δr, so we arrive at

U(r) − Uo g (r − ro )
(13.11)
Φ = gN
This is Brun’s formula, which relates the disturbing potential to the geoid height N.
13.3 Reference Gravity and Gravity Anomaly
The reference gravity is the value of total (scalar) acceleration one would measure on
the spheroid assuming no mass anomalies inside of the Earth.
∂U 1 ∂U 1 ∂U
g = −∇Uo = − ∂ro r̂ − r ∂θo θ̂ − r cos θ ∂φo φ̂ (13.12)
The total acceleration on the spheroid is

 !2 1/2
 ∂Uo 2 1 ∂U
g = − ∂r + o

∂θ
 (13.13)
r
The second term on the right side of equation 13.13 is negligible because the normal to
the ellipsoid departs from the radial direction by a small amount, and the square of this
value is usually unimportant. The result is
3J a2

GM
g(r, θ) = − 2 e 1 − 2 2 3 sin2 θ − 1 + ω2 rcos2 θ (13.14)
r 2r
To calculate the value of gravity anomaly on the spheroid, we substitute

r(θ) = a 1 − f sin2 θ (13.15)
After substitution, expand the gravity in a binomial series and keep terms of order f ,
but not f 2 :
h i
g(θ) = ge 1 + 52 m − f sin2 θ
(13.16)
ω2 a2
m = GM
e
The parameter ge is the value of gravity on the equator and m is approximately equal
to the ratio of centrifugal force at the equator to the gravitational acceleration at the
equator. In practice, geodesists get together at meetings of the International Union

of Geodesy and Geophysics (IUGG) and agree on such things as the parameters of
WGS84. In addition, they define something called the international gravity formula:

go (θ) = 978.03185 1 + 0.00527889 sin2 θ + 0.000023462 sin4 θ (13.17)
This version was adopted in 1967, so for a real application, you should use a more up-
to-date value or use a value that is consistent with all of the other data in your database.
13.4 Free-Air Gravity Anomaly
The free-air gravity anomaly is the negative radial derivative of the disturbing potential,
but it is also evaluated on the geoid. The formula is
∂Φ 2g (θ)
∆g = − ∂r − r(θ)
o
N (13.18)
13.5 Summary of Anomalies
Disturbing potential Φ
U = Uo + Φ (13.19)
total reference disturbing
potential potential potential
Geoid height N
Φ
N= (13.20)
go (θ)
Free-air gravity anomaly

∂Φ 2g (θ)
∆g = − ∂r − r(θ)
o
N (13.21)
Deflection of the vertical The final type of anomaly, not yet discussed, is the deflec-
tion of the vertical. This is the angle between the normal to the geoid and the normal
to the spheroid. There are two components: north ξ and east η.
1 ∂N
ξ=−
a ∂θ
(13.22)
1 ∂N
η=−
a cos θ ∂φ
Chapter 14
Laplace’s Equation in Spherical

Coordinates
14.1 Introduction
As discussed in Chapter 12, the gravity field of the Earth can be decomposed into a ref-
erence gravity model (e.g., WGS84) and anomalies which can be expanded in spherical
harmonics and/or Fourier series. The spherical harmonic decomposition should be used
for longer wavelength anomalies (i.e., λ > 1000 km). However, for shorter wavelength
anomalies (e.g., λ < 1000 km), the Fourier series representation is more practical and
computationally efficient.
We begin by introducing spherical harmonics and their properties. We explain how the
spherical harmonic decomposition of a function on a sphere is analogous to the Fourier
series decomposition of a 2-D function in Cartesian coordinates. We then use this
spherical harmonic formulation to solve Laplace’s equation. Finally, we describe how
the Earth’s gravity field is represented as spherical harmonic coefficients. These notes
follow from Chapter 3 in Jackson [1975] [34]. However, when we apply this mathe-
matical development to the Earth, we replace the colatitude measured from the z-axis,
commonly used by mathematicians and physicists, with geocentric latitude measured
from the equator, commonly used by Earth scientists.
163
CHAPTER 14. LAPLACE’S EQUATION IN SPHERICAL COORDINATES 164
14.2 Spherical Harmonics
The coordinate system used for this presentation of spherical harmonics follows from
the definition used by mathematicians and physicists and is shown in Figure 14.1. Lon-
gitude is measured from the x-axis and colatitude θ is measured from the z-axis. Any
function f (θ, φ) can be expanded in terms of spherical harmonic coefficients as follows
∞ X
X m=l
f (θ, φ) = Flm Ylm (θ, φ) (14.1)
l=0 m=−l
where Ylm (θ, φ) are the spherical harmonic functions, where l is the spherical harmonic
degree and m of the spherical harmonic order. The spherical harmonic coefficients Flm
are computed by integrating the function over the sphere as follows
z
x
r
θ
y
φ
Figure 14.1
Z 2π Z π
Flm = f (θ, φ)Y m
l
(θ, φ) sin θdθdφ (14.2)
0 0
where the overbar signifies complex conjugate. The fully normalized spherical har-
monic functions are
#1/2
(2l + 1) (l − m)!
"
Ylm (θ, φ) = l (cos θ)e
Pm imφ
4π (l + m)! (14.3)
Yl−m (θ, φ) = (−1)m Ylm (θ, φ)
where Pm l (cos θ) is the associated Legendre function. Note that e

imφ
represents the
Fourier series which form a complete basis set of orthonormal functions of order m on
the interval 0 6 φ 6 2π. Also the functions Pm l
(cos θ) form a complete basis set in the
degree l for each m on the interval 0 6 θ 6 π . Therefore the Ylm (θ, φ) functions form a
complete orthonormal basis on the unit sphere. The orthonormal condition is
Z 2π Z π
0
Ylm0 (θ, φ)Y m
l
(θ, φ) sin θdθdφ = δl0 l δm0 m . (14.4)
0 0
The completeness relation is
∞ X
X m=l
Ylm θ0 , φ0 Ylm (θ, φ) = δ φ − φ0 δ cos θ − cos θ0 .

(14.5)
l=0 m=−l
Some examples of spherical harmonic functions are
1
Y00 = √
4π
r
3
Y1−1 = sin θe−iφ
8π
r
3
Y1 =
0
cos θ
4π
r
3
Y11 = − sin θeiφ
8π
r (14.6)
15
Y2−2 = sin2 θe−i2φ
32π
r
15
Y2−1 = sin θ cos θe−iφ
8π
r
5
Y2 =
0
3cos2 θ − 1
16π
r
15
Y21 = − sin θ cos θeiφ
8π
r
15
Y2 =
2
sin2 θei2φ .
32π
The spherical harmonic order l is similar to the wavenumber in Fourier series. For a
sphere of radius a the characteristic wavelength of the spherical harmonic function is
2πa
λ= (14.7)
(l + 1)
An examination of these spherical harmonic functions shows that there are always l
nodes around the sphere. There are m nodes in longitude so there are l − m nodes
in latitude. As discussed above, spherical harmonic decomposition of a function on a
sphere is analogous to Fourier decomposition as shown in the table below.
Cartesian coordinates Spherical coordinates

R∞ R∞ R 2π R π
F(k) = f (x)e−i2π(k•x) d2 x Flm = 0 0 f (θ, φ)Y ml
(θ, φ) sin θdθdφ
−∞ −∞
R∞ R∞ ∞ m=l
f (x) = F(k)ei2π(k•x) d2 k f (θ, φ) = Flm Ylm (θ, φ)
P P
−∞ −∞ l=0 m=−l

Re f (x) ⇒ F −k x , ky = F k x , ky Re f (θ, φ) ⇒ Fl−m = (−1)m Flm

(14.8)
The last row of the table shows the Hermitian property of the Fourier transform in the
case where the function is real valued in the space domain. A similar property holds
for spherical harmonic coefficients. In both cases this property enables one to skip the
computation of half of the coefficients which saves both computer time and computer
memory.
14.3 Laplace’s Equation
Now we use this spherical harmonic decomposition to solve Laplace’s equation in

spherical coordinates. Similarly, in the following chapter, we will use the Fourier
transform to solve Laplace’s equation in Cartesian Coordinates. Laplace’s equation
is
1 ∂2 ∂ ∂Φ 1 ∂2 Φ
!
1
(rΦ) + sin θ + =0 (14.9)
r ∂r2 r2 sin θ ∂θ ∂θ r2 sin2 θ ∂φ2
where Φ (r, θ, φ) is the potential, θ is colatitude and φ is longitude. The boundary

conditions are
lim Φ (r, θ, φ) = 0
r→∞
(14.10)
Φ (1, θ, φ) = Φo (θ, φ) .
Now we expand the potential function in spherical harmonics
∞ X
X m=l
Φ (r, θ, φ) = l (r) Yl (θ, φ).
Φm m
(14.11)
l=0 m=−l
With some work it is possible to show that Laplace’s equation reduces to
∂2 m
r rΦl − l (l + 1) Φm
l =0 (14.12)
∂r2
The general solution to this ordinary differential equation is
Φm
l (r) = Al r
m −(l+1)
+ Bm
l r.
l
(14.13)
Note that only the first term in this solution satisfies the boundary condition as r goes
to infinity so the solution is
∞ X
X m=l
Φ (r, θ, φ) = Am
l r Yl (θ, φ).
−(l+1) m
(14.14)
l=0 m=−l
The solution that satisfies the surface boundary condition is
∞ X
X m=l
Φ (r, θ, φ) = Φm
ol r Yl (θ, φ)
−(l+1) m
(14.15)
l=0 m=−l
where
Z 2π Z π
Φm
ol = Φo (θ, φ)Ylm (θ, φ) sin θdθdφ (14.16)
0 0
This solution is used to calculate the potential anywhere exterior to the sphere using the
following approach. First expand the surface potential in spherical harmonics (equa-
tion (14.16)). Then multiply each coefficient by the factor r−(l+1) . Finally sum this
product over all l and m (equation (14.14)). The factor r−(l+1) is called the upward
continuation and it reduces the amplitude of the potential especially at large l which
corresponds to short wavelength. For the Cartesian coordinate system the analogous
upward continuation factor is e−2π|k|x . In Figure 14.2 we compare the two upward
continuation formulas to establish the wavelength where the Cartesian formulation can
replace the spherical harmonic formulation.
1
z = 40 km sphere40
flat40
sphere400
0.8 flat400
sphere4000
flat4000
0.6
gain
0.4 z = 40
0 km
0.2
z = 400
0 km
0
0 5 10 15 20 25 30 35 40
degree
Figure 14.2: Amplitude gain versus spherical harmonic degree for spherical upward
continuation (solid) compared with flat Earth upward continuation (dashed) for three
altitudes 40 km, 400 km, 4000 km. For these altitudes, the two formulas show some
disagreement for degrees less than 20 (1900 km wavelength) but good agreement for
larger degrees. At degree 40 they are indistinguishable and the flat-Earth approximation
is accurate enough for most applications.
14.4 Earth’s Gravity Field
The spherical harmonic formulation commonly used for the Earth’s gravity field uses
geocentric latitude rather than colatitude. With this change, the fully normalized spher-
ical harmonic function becomes
#1/2
(2l + 1) (l − m)!
"
Ylm (θ, φ) = l (sin θ)e
Pm imφ
(14.17)
4π (l + m)!
where θ now refers to geocentric latitude and there is a sin θ in the associated Legendre
function instead of a cos θ. As discussed in the Chapter 12, the l = 1 spherical harmonic
coefficient is zero so the disturbing potential becomes
∞ m=l
−GMe X X m a l m
Φ (r, θ, φ) = Al Y (θ, φ) (14.18)
r l=2 m=0
r l
where GMe and a are two components of the WGS84 reference model provided in the
previous chapter. Note that the C20 term is sometimes equivalent to the J2 term used in
the reference model 13.3. This formulation has complex valued coefficients Am l but it is
customary to use real coefficients Clm and S lm applied to the sine and cosine components
of the complex exponential. Moreover, the standard gravity approach has a different
normalization and uses the fact that the gravity field is real valued to eliminate the need
for negative order coefficients. With these changes, the disturbing potential is
∞ m=l
−GMe X X a l m h i
Φ (r, θ, φ) = P̃l (sin θ) Clm cos (mφ) + S lm sin (mφ) (14.19)
r l=2 m=0
r
where this associated Legendre function has a slightly different normalization
2 − δ0,m (2l + 1) (l − m)! 1/2 m

" #
P̃m (sin θ) = Pl (sin θ). (14.20)
l
(l + m)!
One can relate the real valued spherical harmonic coefficients used for Earth gravity
analysis to the complex coefficients used by mathematical physicists as
 √
2π Clm − iS lm (−1)m m>0





√


Am =

4πClm m=0 (14.21)

l 


√ m



2π Cl + iS lm m < 0.



In practice geodesists have a standard way of distributing the spherical harmonic coef-
ficients. Below are the tide-free coefficients, up to degree 6, for the EGM2008 gravity
model [Pavlis et al., 2012] [55]
degree order Clm Slm sigClm sigSlm

2 0 -0.484165143790815D-03 0.000000000000000D+00 0.7481239490D-11 0.0000000000D+00
2 1 -0.206615509074176D-09 0.138441389137979D-08 0.7063781502D-11 0.7348347201D-11
2 2 0.243938357328313D-05 -0.140027370385934D-05 0.7230231722D-11 0.7425816951D-11
3 0 0.957161207093473D-06 0.000000000000000D+00 0.5731430751D-11 0.0000000000D+00
3 1 0.203046201047864D-05 0.248200415856872D-06 0.5726633183D-11 0.5976692146D-11
3 2 0.904787894809528D-06 -0.619005475177618D-06 0.6374776928D-11 0.6401837794D-11
3 3 0.721321757121568D-06 0.141434926192941D-05 0.6029131793D-11 0.6028311182D-11
4 0 0.539965866638991D-06 0.000000000000000D+00 0.4431111968D-11 0.0000000000D+00
4 1 -0.536157389388867D-06 -0.473567346518086D-06 0.4568074333D-11 0.4684043490D-11
4 2 0.350501623962649D-06 0.662480026275829D-06 0.5307840320D-11 0.5186098530D-11
4 3 0.990856766672321D-06 -0.200956723567452D-06 0.5631952953D-11 0.5620296098D-11
4 4 -0.188519633023033D-06 0.308803882149194D-06 0.5372877167D-11 0.5383247677D-11
5 0 0.686702913736681D-07 0.000000000000000D+00 0.2910198425D-11 0.0000000000D+00
5 1 -0.629211923042529D-07 -0.943698073395769D-07 0.2989077566D-11 0.3143313186D-11
5 2 0.652078043176164D-06 -0.323353192540522D-06 0.3822796143D-11 0.3642768431D-11
5 3 -0.451847152328843D-06 -0.214955408306046D-06 0.4725934077D-11 0.4688985442D-11
5 4 -0.295328761175629D-06 0.498070550102351D-07 0.5332198489D-11 0.5302621028D-11
5 5 0.174811795496002D-06 -0.669379935180165D-06 0.4980396595D-11 0.4981027282D-11
6 0 -0.149953927978527D-06 0.000000000000000D+00 0.2035490195D-11 0.0000000000D+00
6 1 -0.759210081892527D-07 0.265122593213647D-07 0.2085980159D-11 0.2193954647D-11
6 2 0.486488924604690D-07 -0.373789324523752D-06 0.2603949443D-11 0.2466506184D-11
6 3 0.572451611175653D-07 0.895201130010730D-08 0.3380286162D-11 0.3347204566D-11

6 4 -0.860237937191611D-07 -0.471425573429095D-06 0.4535102219D-11 0.4489428324D-11
6 5 -0.267166423703038D-06 -0.536493151500206D-06 0.5097794605D-11 0.5101153019D-11
6 6 0.947068749756882D-08 -0.237382353351005D-06 0.4731651005D-11 0.4728357086D-11
In summary, one can calculate the gravity field at a radius greater than the equatorial
radius of the Earth. If the calculation is performed in and Earth-fixed coordinate system
rotating with the Earth then the gravity computed using the WGS84 reference model in
equation (13.3) is added to the disturbing potential computed using spherical harmonic
coefficients (14.19).
 ∞ Xm=l l

−GMe  X a m h  1 2 2 2
i
U (r, θ, φ) = 1+ P̃l (sin θ) Cl cos (mφ) + S l sin (mφ) 
m m
− ω r cos θ

 
r


 r  2

l=2 m=0
(14.22)
There is an excellent web service (http://icgem.gfz-potsdam.de) where one can select

their favorite gravity model as spherical harmonic coefficients and then compute a wide
variety of gravity products (e.g., geoid height, free-air anomaly, gravity gradient . . .)
on a grid defined by the user.
14.5 Exercises
Chapter 15
Laplace’s Equation in
Cartesian Coordinates and
Satellite Altimetry
15.1 Solution to Laplace’s Equation
Variations in the gravitational potential and the gravitational force are caused by lo-
cal variations in the mass distribution in the Earth. As described in Chapter 12, we
decompose the gravity field of the Earth into three fields:
• the main field due to the total mass of the Earth,
• the second harmonic due to the flattening of the Earth by rotation,

• and anomalies which can be expanded in spherical harmonics or Fourier series.
Here we are interested in anomalies due to local structure. Consider a patch on the
Earth having a width and length less than about 1000 km, or 1/40 of the circumference
of the Earth. Within that patch we are interested in features as small as perhaps 1 km
wavelength. Using a spherical harmonic representation would require 40,000 squared
coefficients! To avoid this enormous computation and still achieve accurate results, we
will treat the Earth as being locally flat. Here is a remove/restore approach that has
worked well in our analysis of gravity and topography:
171
CHAPTER 15. LAPLACE’S EQUATION 172
1. Acquire a spherical harmonic model of the gravitational potential of the Earth

and generate models of the relevant quantities (e.g., geoid height, gravity anomaly,
deflection of the vertical, . . . ) out to, say, harmonic 80. You may want to taper
the harmonics between, say, 60 and 120 to avoid Gibb’s phenomenon; this de-
pends on the application.
2. Remove that model from the local geoid, gravity, . . . An alternate method is to
remove a trend from the data and then apply some type of window prior to per-
forming the Fourier analysis. I do not recommend this practice because the trend
being removed will contain a broad spectrum; it is dependent on the size of the
area, and it cannot be restored accurately.
3. Project the residual data onto a Mercator grid so that the cells are approximately
square, and use the central latitude of the grid to establish the dimensions of the
grid for Fourier analysis.
4. Perform the desired calculation (e.g., upward continuation, gravity/topography
transfer function, . . . ).
5. Restore the appropriate spherical harmonic quantity using the exact model that
was removed originally.
Consider the disturbing potential
U = Uo + Φ (15.1)
where, in this case, the reference potential comprises the reference Earth model plus
the reference spherical harmonic model described in Step (1) above. The disturbing
potential satisfies Laplace’s equation for an altitude, z, above the highest mountain in
the area, while it satisfies Poisson’s equation below this level, as shown in Figure 15.1.
Figure 15.1
Φ(x, y, z) disturbing potential (total - reference)
G gravitational constant
ρ density anomaly (total - reference)
Laplace’s equation is a second order partial differential equation in three dimensions.
∂2 Φ ∂2 Φ ∂2 Φ
+ 2 + 2 = 0, z>0 (15.2)
∂x2 ∂y ∂z
Six conditions are needed to develop a unique solution. Far from the region, the dis-
turbing potential must go to zero; this accounts for five of the boundary conditions
lim Φ = 0, lim Φ = 0, lim Φ = 0 (15.3)

|x|→∞ |y|→∞ z→∞
At the surface of the Earth (or at some elevation), one must either prescribe the potential
or the vertical derivative of the potential.
Φ(x, y, 0) = Φo (x, y) (Dirichlet)

∂Φ
(15.4)
∂z
= −∆g(x, y) (Neumann)
To solve this differential equation, we’ll use the 2-D Fourier transform again where the
forward and inverse transform are
Z∞ Z∞
F(k) = f (x)e−i2π(k·x) d2 x
−∞ −∞
(15.5)
Z∞ Z∞
f (x) = F(k)ei2π(k·x) d2 k
−∞ −∞
where x = (x, y) is the position vector, k = (1/λ x , 1/λy ) is the wavenumber vector, and
k · x = k x x + ky y . Fourier transformation reduces Laplace’s equation and the surface
boundary to
∂2 Φ
−4π2 k2x + ky2 Φ(k, z) + 2 = 0
∂z
(15.6)
lim Φ(k, z) = 0, Φ(k, 0) = Φo
z→∞
The general solution is
Φ(k, z) = A(k)e2π|k|z + B(k)e−2π|k|z (15.7)

To satisfy the boundary condition as z → ∞ , the A(k) term must be zero. To satisfy
the boundary condition on the z = 0 plane, B(k) must be Φ(k, 0). The final result is
Φ(k, z) = Φo (k, 0) e−2π|k|z (15.8)

potential at potential at upward
altitude z=0 continuation
Figure 15.2: Gain of upward continuation kernel as a function of the altitude of the
observation z, divided by the wavelength of the anomaly λ.
The upward continuation physics, as shown in Figure (15.2), is the same for the po-
tential and all of its derivatives. For example, if one measured gravity anomaly at the
surface of the Earth ∆g(x, 0), then to compute the gravity at an altitude of z, one takes
the Fourier transform of the surface gravity, multiplies it by the upward continuation
kernel, and inverse transforms the result. This exponential decay of the signal with
altitude is a fundamental barrier to recovery of small-scale gravity anomalies from a
measurement made at altitude. Here are two important examples.
1. Marine gravity Consider making marine gravity measurement on a ship which

is 4 km above the topography of the ocean floor. (Most of the short-wavelength
gravity anomalies are generated by the mass variations associated with the topog-
raphy of the seafloor.) At a wavelength of, say, 8 km, the ocean surface anomaly
will be attenuated by 0.043 from the amplitude of the seafloor anomaly.
2. Satellite gravity The typical altitude of an artificial satellite used to sense vari-
ations in the gravity field is 400 km, so an anomaly having a 100-km wavelength
will be attenuated by a factor of 10−11 ! This is why radar altimetry (below),
which measures the geoid height directly on the ocean surface topography, is
so valuable.
15.2 Derivatives of the gravitational potential
This solution to Laplace’s equation can be used to construct all of the common deriva-
tives of the potential. Suppose one has a complete survey over a patch on the surface
of the Earth so that a Fourier method can be used to convert among the different repre-
sentations of the gravity field. This is particularly true for computing gravity anomaly
from geoid height or deflection of the vertical. The general relation between the po-
tential in the space domain (at any altitude) and the Fourier transform of the surface
potential is
Z∞ Z∞
Φ(x, z) = Φ(k, 0)e−2π|k|z ei2π(k·x) d2 k (15.9)
−∞ −∞
Table 15.1 uses equation (15.9) and the definitions of the derivatives of the potential to
construct the variety of anomalies. Before examining these relationships however, let’s
review some of the definitions in relation to what can be measured.
Gravitational potential
N – Geoid Height: since the ocean surface is an equipotential surface, variations in

gravitational potential will produce variations in the sea surface height. This can be
measured by a radar altimeter.
First derivative of potential
∆g – Gravity Anomaly: this is the derivative of the potential with respect to z. It can be
measured by an accelerometer, such as a gravity meter.
η, ξ – Deflection of the Vertical: these are the derivatives of the potential with respect
to x and y, respectively. These are also forces, however, that are usually measured by
recording the tiny angle between a plumb bob and the vector pointing to the center
of the Earth. Over the ocean this is most easily measured by taking the along-track
derivative of radar altimeter profiles.
Second derivative of potential
 
 ∂2 Φ ∂2 Φ ∂2 Φ 
 ∂x2 ∂x∂y ∂x∂z 
 
 ∂2 Φ ∂2 Φ 
 ∂y2 ∂y∂z 
 
 ∂2 Φ 
∂z2
Gravity Gradient: this is a symmetric tensor of second partial derivatives of the grav-
itational potential. A direct way of making this measurement is to construct a set of
accelerometers each spaced at a distance of ∆ in the x, y, and z directions. What is
the minimum number of accelerometers needed to measure the full gravity gradient
tensor? What can you say about the trace of this tensor when the measurements are
made in free space? Below we’ll develop an alternate method of measuring ∂∂zΦ2 over
2
the ocean using a radar altimeter.
As an exercise, use Laplace’s equation and the various definitions to develop gravity
anomaly from vertical deflection (equation (15.13)) and vertical gravity gradient from
ocean surface curvature (equation (15.14)).
Here is a practical example. Suppose one has measurements of geoid height N(x) over
a large area on the surface of the ocean and wishes to calculate the gravity anomaly,
∆g(x, z) at altitude. The prescription is:
1. Remove an appropriate spherical harmonic model from the geoid;

2. Take the 2-D Fourier transform of N(x);
3. Multiply by g2π |k| e−2π|k| ;
4. Take inverse 2-D Fourier transform;
5. Restore the matching gravity anomaly calculated from the spherical harmonic
model at altitude.
Table 15.1: Relationships among the various representations of the gravity field in free space.
Space Domain Wavenumber Domain
Geoid height from N(x) 1g Φ(x, 0) N(k) g1 Φ(k, 0) (15.10)

the potential
∂Φ
Gravity anomaly ∆g(x, z) − ∂z (x, z) ∆g(k, z) 2π |k| e−2π|k|z Φ(k, 0) (15.11)
from the potential
Deflection of the vertical ∂N 1 ∂Φ i2πk

η(x) = − ∂x − ∂x η(k) − g x Φ(k, 0)
from the potential (east g
CHAPTER 15. LAPLACE’S EQUATION
slope and north slope) (15.12)

i2πky
∂N 1 ∂Φ ξ(k) − g Φ(k, 0)
ξ(x) = − ∂y − ∂y
g
ig h i
Gravity anomaly ∆g(k) = k x η(k) + ky ξ(k) (15.13)
from deflection of the |k|
vertical [29]
!
Vertical gravity gradient ∂∆g ∂2 N ∂2 N
=g + 2 (15.14)
from the curvature of ∂z ∂x2 ∂y
the ocean surface
177
15.3 Conversion of Geoid Height to Vertical Deflection,

Gravity Anomaly, and Vertical Gravity Gradient
from Satellite Altimeter Profiles
As described above, geoid height N(x) and other measurable quantities such as grav-
ity anomaly g(x) are related to the anomalous gravitational potential Φ(x, z) through
Laplace’s equation. It is instructive to go through an example of how measurements of
ocean surface topography from satellite altimetry can be used to construct geoid height,
deflection of the vertical, gravity anomaly, and vertical gravity gradient.
The surface of the ocean is displaced both above and below the reference ellipsoidal
shape of the Earth (Figure 15.3). These differences in height arise from variations
Figure 15.3: Schematic diagram of a radar altimeter orbiting the Earth at an altitude of
800 km.
in gravitational potential (i.e., geoid height) and oceanographic effects (tides, large-
scale currents, el Nino, eddies, . . . ). Fortunately, the oceanographic effects are small
compared with the permanent gravitational effects, so a radar altimeter can be used to
measure these bumps and dips. At wavelengths less than about 200 km, bumps and
dips in the ocean surface topography reflect the topography of the ocean floor and can
be used to estimate seafloor topography in areas of sparse ship coverage.
Radar altimeters are used to measure the height of the ocean surface above the refer-
ence ellipsoid (i.e., the satellite above the ellipsoid H ∗ minus the altitude above the
ocean surface H). A GPS, or ground-based tracking system, is used to establish the
position of the radar H ∗ (as a function of time) to an accuracy of better than 0.1 m.
The radar emits 1000 pulses per second using a carrier wavelength of about 2 cm (Ku-
band). These spherical wave fronts reflect from the closest ocean surface (nadir) and
return to the satellite, where the two-way travel times is recorded to an accuracy of 3
nanoseconds (1-m range variations mostly due to ocean waves). Averaging thousands
of pulses reduces the noise to about 30 mm. If one is interested in making an accurate
geoid height map such, as shown in Figure 15.5, then many sources of error must
Figure 15.4: Ground tracks for two radar altimeters that have provided adequate cov-
erage of the ocean surface. This is a 1500-km by 1000-km area around Hawaii.
Geosat/ERM (U.S. Navy) is the exact repeat orbit phase of the Geosat altimeter mis-
sion (U.S. Navy, 1985–89). These ground tracks repeat every 17 days, to monitor
small changes in ocean surface height caused by time-varying oceanographic effects.
Geosat/GM ground tracks were acquired during a 1.5-year period when the orbit was
allowed to drift. Similarly ERS/GM was a European Space Agency phase of the ERS
mission that acquired non-repeat profiles for 1 year. ERS/ERM is the 35-day repeat
phase of the ERS mission.
be considered and somehow removed. However, if the final product of interest is one
of the derivatives of the potential then it is best to take the along-track derivative of
each profile to develop along-track sea surface slope. In this case, the point-to-point
precision of the measurements is the limiting factor.
To avoid a crossover adjustment of the data, ascending and descending satellite altime-
ter profiles are first differentiated in the along-track direction, resulting in geoid slopes
Figure 15.5: Geoid height above the WGS84 ellipsoid in meters (1 m contour interval)
from EGM2008 [55]. The geoid height is dominated by long wavelengths, so it is
difficult to observe the small-scale features caused by ocean-floor topography. These
can be enhanced by computing either the horizontal derivative (ocean surface slope) or
the vertical derivative (gravity anomaly).
or along-track vertical deflections. These along-track slopes are then combined to pro-
duce east η and north ξ components of vertical deflection [61]. Finally the east and
north vertical deflections are used to compute both gravity anomaly and vertical gravity
gradient. The details for converting along-track slope into east and north components
of deflection of the vertical are provided in Section 15.4 and also in a reference [65].
You probably don’t need to know these details unless you plan to do research in marine
gravity.
To compute gravity anomaly (Figure 15.6) from a dense network of satellite altimeter
profiles of geoid height (Figure 15.3), one constructs grids of east η and north ξ ver-
tical deflection (Figures 15.7 and 15.8). The grids are then Fourier transformed and
equation (15.3) is used to compute gravity anomaly. At this point one can add the
long wavelength gravity field from the spherical harmonic model to the gridded grav-
ity values in order to recover the total field; the resulting sum may be compared with
gravity measurements made on board ships. A more complete description of gravity
field recovery from satellite altimetry can be found in [33, 65, 57].
There is an important issue for constructing gravity anomaly from sea surface slope
that is revealed by a simplified version of equation (15.13). Consider the sea surface
Figure 15.6: Gravity anomaly ∆g(x) derived from east and north components of sea
surface slope using equation (15.13). (20 mGal contour interval.)
slope and gravity anomaly across a two-dimensional structure which depends on x but
not y. The y-component of slope is zero so conversion from sea surface slope to gravity
anomaly is simply a Hilbert transform:
∆g(k x ) = ig sgn(k x )η (k x ) (15.15)
Now it is clear that one µrad of sea surface slope maps into 0.98 mGal of gravity
anomaly and similarly one µrad of slope error will map into 1 mGal of gravity anomaly
error. Thus the accuracy of the gravity field recovery is controlled by the accuracy of
the sea surface slope measurement.
Figure 15.7: East component of sea surface slope η(x) derived from Geosat and ERS
altimeter profiles. Note this component is rather noisy because the altimeter tracks
(Figure 15.4) run mainly in a N-S direction.
Figure 15.8: North component of sea surface slope ξ(x) derived from Geosat and ERS
altimeter profiles. Note this component has lower noise because the altimeter tracks
(Figure 15.4) run mainly in a N-S direction.
Figure 15.9: Vertical gravity gradient δg(x)/δz derived from east and north components
of sea surface slope using equation (15.14). Note this second derivative of the geoid
amplifies the shortest wavelengths; compare with the original geoid (Figure 15.4).
Noise in the altimeter measurements has been amplified resulting in an artificial tex-
ture.
15.4 Vertical Deflections from Along-Track Slopes
Consider for the moment the intersection point of an ascending and a descending satel-
lite altimeter profile. The derivative of the geoid height N with respect to time t along
the ascending profile is
∂N ∂N ∂N
Ṅa = ∂ta = ∂θ θ˙a + ∂φ φ˙a (15.16)
and along the descending profile is

∂N ∂N
Ṅd = ∂θ θ˙d + ∂φ φ˙d (15.17)
where θ is geodetic latitude and φ is longitude. The functions θ and φ are the latitudinal
and longitudinal components of the satellite ground track velocity. It is assumed that
the satellite altimeter has a nearly circular orbit so that its velocity depends mainly
on latitude; at the crossover point the following relationships are accurate to better
than 0.1%.
θ˙a = −θ˙ d φ˙a = φ˙d (15.18)
The geoid gradient (deflection of the vertical) is obtained by solving (15.16) and (15.17)
using (15.18).
∂N 1
∂φ
= (Ṅa + Ṅd ) (15.19)
2φ̇
∂N 1
∂θ
= (Ṅa − Ṅd ) (15.20)
2|θ̇|
It is evident from this formulation that there are latitudes where either the east or north
component of geoid slope may be poorly determined. For example, at ±72◦ latitude,
the Seasat and Geosat altimeters reach their turning points where the latitudinal veloc-
ity θ goes to zero and thus equation (15.19) becomes singular. In the absence of noise
this is not a problem because the ascending and descending profiles are nearly parallel
so that their difference goes to zero at the same rate that the latitudinal velocity goes to
zero. Of course in practice altimeter profiles contain noise such that the north compo-
nent of geoid slope will have a signal to noise ratio that decreases near ±72◦ latitude.
Similarly for an altimeter in a near polar orbit, the ascending and descending profiles
are nearly anti-parallel at the low latitudes; the east component of geoid slope is poorly
determined and the north component is well determined. The optimal situation occurs
when the tracks are nearly perpendicular so that the east and north components of geoid
slope have the same signal-to-noise ratio.
When two or more satellites with different orbital inclinations are available, the situ-
ation is slightly more complex but also more stable. Consider the intersection of four
passes as shown in the following diagram. The along-track derivative of each pass can
be computed from the geoid gradient at the crossover point
   
Ṅ  θ˙ φ˙1 
 1   1   
  
Ṅ2  θ˙2   ∂N 
  =  φ˙2   ∂θ 
      (15.21)
Ṅ3  θ˙3 φ˙3   ∂N 
 
    ∂φ
θ˙ φ˙ 
   
4 Ṅ 4 4
or in matrix notation
Ṅ = Θ ∆N (15.22)
Since this is an overdetermined system, the four along-track slope measurements can-
not be matched exactly unless the measurements are error-free. In addition, an a priori
estimate of the error in the along-track slope σi measurements can be used to weight
each equation in equations (15.21) (i.e., divide each of the four equations by σi ). The
least squares solution to equation (15.22) is
∆N = (Θt Θ)−1 Θt Ṅ (15.23)
where t and −1 are the transpose and inverse operations, respectively. In this case a
2-by-4 system must be solved at each crossover point, although the method is easily
extended to three or more satellites. Later we will assume that every grid cell corre-
sponds to a crossover point of all the satellites considered, so this small system must
be solved many times.
In addition to the estimates of geoid gradient, the covariances of these estimates are
also obtained  
 σθθ σ2θφ 
 2 
  = (Θt Θ)−1 (15.24)
σφθ σφφ 
 2 2 
Since Geosat and ERS-1 are high inclination satellites, the estimated uncertainty of
the east component is about 3 times greater than the estimated uncertainty of the north
component at the equator. At higher latitudes of 60◦ –70◦ , where the tracks are nearly
perpendicular, the north and east components are equally well determined. At 72◦
north, where the Geosat tracks run in a westerly direction, the uncertainty of the east
component is low and the higher inclination ERS-1 tracks prevent the estimate of the
north component from becoming singular at 72◦ .
Finally, the east η and north ξ components of vertical deflection are related to the two
geoid slopes by
1 ∂N
η=− (15.25)
a cos θ ∂φ
1 ∂N
ξ=− (15.26)
a ∂θ
where a is the mean radius of the Earth.
15.5 Exercises
Chapter 16
Poisson’s Equation in
Cartesian Coordinates
16.1 Solution to Poisson’s Equation
As in the Chapter 15 on Laplace’s equation, we are interested in anomalies due to local

structure and will use a flat-Earth approximation. However unlike Chapter 15, the em-
phasis is on generating models of the disturbing potential and it derivatives from a 3-D
model of the variations in density and topography of the Earth. In a Chapter 17 we’ll
combine this approach to calculating gravity models with the models for isostasy and
flexure to develop a topography-to-gravity transfer function. Consider the disturbing
potential
U = Uo + Φ (16.1)
where, in this case, the reference potential comprises the ellipsoidal reference Earth
model plus the reference spherical harmonic model. The disturbing potential satisfies
Laplace’s equation for an altitude z above the highest mountain in the area, while it
satisfies Poisson’s equation below this level, as shown in Figure 16.1.
First consider a density model consisting of an infinitesimally-thin sheet at a depth zo

having a surface-density of σ(x, y) (units of mass per unit area). Later we’ll construct
a more complicated 3-D structure from a stack of many layers. Poisson’s equation is
an inhomogeneous second-order partial differential equation in three dimensions:
∂2 Φ ∂2 Φ ∂2 Φ
+ 2 + 2 = −4πGσ(x) δ(z − zo ), (16.2)
∂x2 ∂y ∂z
187
CHAPTER 16. POISSON’S EQUATION 188
Figure 16.1
Φ(x, y, z) disturbing potential (total-reference)
G gravitational constant
ρ density anomaly (total-reference)
Six conditions are needed to develop a unique solution. Far from the region, the dis-
turbing potential must go to zero; this accounts for the five boundary conditions.
lim Φ = 0, lim Φ = 0, lim Φ = 0 (16.3)

|x|→∞ |y|→∞ |z|→∞
The sixth condition is prescribed by the density model. To solve this differential equa-
tion, we’ll use the 2-D Fourier transform again where the forward and inverse trans-
forms are
Z∞ Z∞
F(k) = f (x)e−i2π(k·x) d2 x
−∞ −∞
(16.4)
Z∞ Z∞
f (x) = F(k)ei2π(k·x) d2 k
−∞ −∞
where x = (x, y) is the position vector, k = (1/λ x , 1/λy ) is the wavenumber vector, and
(k • x) = k x x + ky y. Fourier transformation reduces Poisson’s equation and the surface
boundary to
∂2 Φ
−4π2 k2x + ky2 Φ(k, z) + 2 = −4πGσ(k) δ(z − zo ) (16.5)
∂z
lim Φ(k, z) = 0 (16.6)

z→∞
Next take the Fourier transform with respect to z:

π k2x + ky2 + kz2 Φ(k, kz ) = Gσ(k)e−i2πkz zo (16.7)
We have used the definition of the delta function
Z∞
δ(z − zo )e−i2πkz dz = e−i2πkzo
−∞
Next we solve the differential equation for Φ and take the inverse Fourier transform
with respect to kz .
Z∞
Gσ(k) ei2πkz (z−zo )
Φ(k, z) = dkz (16.8)
π

kz2 + k2x + ky2
−∞
Use Calculus of Residues to do the integration. The denominator can be factored as

follows:
kz2 + k2x + ky2 = (kz + i |k|) (kz − i |k|) (16.9)
1/2
where |k| = k2x + ky2 . If z > zo , then to satisfy the boundary condition as z → ∞,
one must integrate around the i |k|-pole. See Figure 16.2.
Figure 16.2
The result is
Z∞
ei2πkz (z−zo ) e−2π|k|(z−zo )
dkz = 2πi (16.10)
(kz + i |k|) (kz − i |k|) 2i |k|
−∞
The solution for the potential for z > zo is
e−2π|k|(z−zo )
Φ(k, z) = Gσ(k) (16.11)
|k|
The gravity anomaly is
∂Φ
∆g(k, z) = − ∂z = 2πGσ(k)e−2π|k|(z−zo ) (16.12)
16.2 Gravity Due to Seafloor Topography: Approximate

Formula
Consider topography on the ocean floor t(x) where the maximum amplitude of the
topography is much less than the mean ocean depth s, as shown in Figure 16.3.
Figure 16.3
Because the topography has low amplitude we can replace the surface density in equa-
tion (16.12) with the topography times the density contrast across the seafloor:
∆g(k) = 2πG ( ρc − ρw ) T (k)e−2π|k|s (16.13)
The result shows that, to a first approximation, the relationship between gravity and
topography is linear and isotropic.
∆g
T
= 2πG ( ρc − ρw ) e−2π|k|s (16.14)
At long wavelengths |k| → 0 so the exponential upward continuation term is 1 and the
gravity/topography ratio is simply the Bouguer correction term.
∆g
T
= 2πG ( ρc − ρw ) = 75 mGal/km (16.15)
Suppose the wavelength of the topography is equal to the ocean depth. In this case the
exponential, upward-continuation reduces the gravity measured on the ocean surface
by a factor of e−2π = 0.0017. Because of this upward-continuation, topography having
wavelength less than the ocean depth become increasingly difficult to observe in the
gravity field at the ocean surface.
16.3 Gravity Anomaly from a 3-D Density Model
Using this formulation, one can stack, or integrate, these surface density layers over a
range of depths to construct the gravity field due to a full 3-D density model. Also see
Figure 16.4.
Zo
e−2π|k|(z−zo )
Φ(k, z) = G ρ(k, zo ) dzo (16.16)
|k|
−∞
Figure 16.4
Zo
e−2π|k|(z−zo )
Φ(k, z) = G ρ(k, zo ) dzo (16.17)
|k|
−∞
The equivalent expression in the space domain is

Z∞ Z∞ Zo h i−1/2
Φ(x, z) = G ρ (xo , yo , zo ) (x − xo )2 + (y − yo )2 + (z − zo )2 dzo dyo dxo
−∞ −∞ −∞
(16.18)
Indeed this is just a statement of the convolution theorem where
−1/2 e−2π|k|z
= x2 + y2 + z2 = (16.19)
|k|
16.4 Computation of Geoid Height and Gravity Anomaly
Table 16.1 provides the two approaches for calculating geoid height and gravity anomaly
from a 3-D density model. The Fourier approach involves 2-D Fourier transformation
of each layer, summing (or integrating) the upward-continued contribution from each
layer, and inverse Fourier transformation of the total. The space-domain approach
involves a 3-D convolution of the density model with the 1/r (geoid) or z/r3 (gravity)
kernel. For a model with 1024 points in both horizontal directions the Fourier approach
will be about 50,000 times faster to compute than the space-domain convolution. More-
over the Fourier approach will have higher numerical accuracy, because there are fewer
additions and subtractions.
Space Domain Wavenumber Domain
Z∞ Z∞ Zo Zo
G ρ (xo , yo , zo ) G e2π|k|zo
N(x) = i1/2 dzo dyo dxo N(k) = ρ(k, zo ) dzo
g h g |k|
−∞ −∞ −∞ (x − xo )2 + (y − yo )2 + z2o −∞
Z∞ Z∞ Zo Zo
ρ (xo , yo , zo ) zo
∆g(x) = G h i3/2 dzo dyo dxo ∆g(k) = 2πG ρ(k, zo )e2π|k|zo dzo
−∞ −∞ −∞ (x − xo )2 + (y − yo )2 + z2o −∞
Table 16.1
16.5 Gravity Anomaly for a Slab of Thickness H

and a Density of ρo
The equation relating gravity to the 3-D density anomaly in the wavenumber domain
can be used to calculate the gravity anomaly due to a slab of thickness H and a den-
sity of ρo . This is used for the Bouguer correction in land gravity surveys. The 3-D
density is 
 ρo −H < z < 0

ρ(x, z) = 

(16.20)
 0 z < −H, z > 0

The Fourier transform of this density is

 δ (k x ) δ ky ρo
 −H < z < 0
ρ(k, z) = 

(16.21)
0
 z < −H, z > 0
The gravity anomaly integral simplifies to

o
Z
∆g(k) = 2πGρo δ (k x ) δ ky e2π|k|zo dzo
(16.22)
−H
1
= 2πGρo δ (k x ) δ ky
1 − e−2π|k|H
2π |k|
Since only the zero wavenumber component is extracted by the delta function, we
expand 16.24 in a Taylor series about |k| and take the limit as |k| → 0 .
1 (2π |k| H)2

lim 1 − 1 + 2π |k| H − 2!
+ · · · =H (16.23)
|k|→0 2π |k|
The result in the wavenumber domain is

∆g(k) = 2π Gρo δ (k x ) δ ky H. (16.24)
The inverse Fourier transform provides the gravity field due to an infinite slab
∆g(x) = 2π Gρo H. (16.25)
16.6 Bouguer Gravity Anomaly
Over the ocean one measures the total acceleration of gravity and subtracts the Interna-
tional Gravity Formula (IGF) to obtain the free-air gravity anomaly. Indeed, the free-air
anomaly is defined on the geoid, which is closely-approximated by the ocean surface.
Therefore no corrections are needed for marine gravity measurements.
In contrast, over the land one measures total gravitational acceleration at some eleva-
tion h above the geoid; assume this elevation is known from leveling. To reduce these
gravity measurements to the geoid, two corrections are commonly applied:
1. The free-air correction accounts for the decrease in gravity because the observa-
tion point is farther from the center of the Earth.
2. The Bouguer correction uses the infinite-slab approximation to account for the
gravitational attraction of the rock between the measurement point and the geoid.
Note unless the topography is very flat over a large area, this infinite-slab approx-
imation may not be very accurate, and a more accurate terrain correction should
be applied.
2GMe
∆gB = gt − 2πGρo h + h − γo (θ) (16.26)
R3e
Bouguer measured slab International
gravity correction free-air Gravity Formula
gravity
correction
(-0.1118 mGal/m)
(0.3086 mGal/m)
(ρo = 2670kgm3 )
16.7 Gravity Anomaly from Topography: Parker’s Ex-

act Formula
In the previous development (section 16.2), we compressed the topography t(x) into
a thin sheet with varying surface density. The approximation is accurate when the
amplitude of the topography is less than the upward continuation distance. For example
in the ocean where the mean seafloor depth is 4 km, the approximation works quite well
for topography that extends 2 km above that depth. However when the topography
is rugged and approaches the observation plane (e.g. sea surface), a more accurate
treatment is needed. Parker [52] derived a more accurate formula that results in a
Taylor series expansion in powers of topography.
Consider the exact formula for the disturbing potential Φ (xo , zo ) due to a uniform den-
sity slab having a flat bottom and upper surface topography t (x) (Figure 16.5) is given
by the following 3-D convolution integral
z
Φ (xo, zo)
zo
y
t(x)
ρ
0
x
Figure 16.5
Z∞ Z∞ Zt(x)h i−1/2
Φ (xo , zo ) = ρG (x − xo )2 + (y − yo )2 + (z − zo )2 dzd2 x. (16.27)
−∞ −∞ 0
Take the 2-D Fourier transform of the disturbing potential on the observation plane.
Z∞ Z∞ Zt(x)Z∞ Z∞ h i−1/2
= [Φ] = Φ (k, zo ) = ρG (x − xo )2 + (y − yo )2 + (z − zo )2 e−i2π(k•xo ) d2 xo dzd2 x
−∞ −∞ 0 −∞ −∞
(16.28)
In section 16.3 we showed that the 2-D Fourier transform of convolution the inverse
distance Greens function is given by equation 16.19.
−1/2 e−2π|k|z
= x2 + y2 + z2 = (16.29)
|k|
Also recall the shift property of the Fourier transform equation (2.21).
= f (x−xo ) = e−i2π(k•x) F (k)

(16.30)
Using these tools we can write the Fourier transform of the disturbing potential as
Z∞ Z∞ Zt(x)
e−2π|k|(zo −z) −i2π(k•x) 2
Φ (k, zo ) = ρG e dzd x. (16.31)
|k|
−∞ −∞ 0
The integral over z can be performed analytically.
Zt(x)
e−2π|k|zo h 2π|k|t(x) i
e −2π|k|zo
e2π|k|z dz = e −1 (16.32)
2π |k|
0
We can expand the term in brackets in a Taylor series about |k| = 0.
|2πk|2 2
" #
1 + 2π |k| t (x) + t (x) + . . . − 1 (16.33)
2!
Now we can rewrite the Fourier transform of the disturbing potential on the plane as
∞
X |2πk|n−2
Φ (k, zo ) = 2πρGe −2π|k|zo
= tn (x) .

(16.34)
n=1
n!
Finally note that the gravity anomaly is the negative vertical derivative of the potential
∆g = − ∂Φ
∂z so the result is
∞
X |2πk|n−1
∆g (k, zo ) = 2πρGe−2π|k|zo = tn (x) .

(16.35)
n=1
n!
This exact formula for computing gravity anomaly from topography involves an infinite
series of Fourier transforms of the topography raised to the power n. In the derivation
of the approximate formula for gravity due to seafloor topography (Section 16.2), we
compressed the topography, times the density contract across the seafloor, into surface
density at an average seafloor depth. We see now that this is equivalent to keeping just
the n = 1 term in equation 16.35 to arrive at
∆g (k, zo ) = 2πρGe−2π|k|zo T (k) (16.36)
Parker [52] proves that this series converges as long as the highest peak in the topog-
raphy does not extend above the observation plane. Moreover, the convergence of the
series is optimal when the z = 0 level is selected such that it is half way between the
maximum and minimum topography.
One can use this more exact formula for calculating gravity due to flexurally-compensated
topography as discussed in the next chapter.
16.8 Exercises
Exercise 16.1. Abyssal hills on the seafloor have a characteristic wavelength of 10 km

and a peak-to-trough amplitude of 500 m.
(a) What is the amplitude of the gravity anomaly on the seafloor assuming the topog-
raphy (density 2800 kg m−3 ) can be compressed into a thin sheet?
(b) What is the amplitude of the gravity anomaly at the sea surface where the mean
ocean depth is 3 km?
(c) Over a time period of 50 Ma, the abyssal hills will be carried by plate tectonics
into the deep ocean where the depth is 5 km. What is the new amplitude of the gravity
anomaly?
(d) In addition to a deeper ocean, the topography of the abyssal hills will be covered
with sediment so the seafloor is now flat. What is the new value of the amplitude of the
gravity? (Use a sediment density of 2300 kg m−3 ).
Exercise 16.2. Derive the Bouguer formula (equation 16.25) for the gravity due to
a slab of uniform thickness H and uniform density ρo from the Parker expansion for
gravity due to topography of uniform density (equation 16.35). Replace t(x) by H and
continue the calculation.
Chapter 17
Gravity/Topography
Transfer Function and
Isostatic Geoid Anomalies
17.1 Introduction
This chapter combines thin-elastic plate flexure theory with the solution to Poisson’s
equation to develop a linear relationship between gravity and topography. This rela-
tionship can be used in a variety of ways:
1. If both the topography and gravity are measured over an area that is several times
greater than the flexural wavelength, then the gravity/topography relationship (in
the wavenumber domain) can be used to estimate the elastic thickness of the
lithosphere and/or the crustal thickness. There are many good references on this
topic, including [18, 42, 1, 84, 44].
2. At wavelengths greater than the flexural wavelength, where features are isostati-
cally compensated, the geoid/topography ratio can be used to estimate the depth
of compensation of crustal plateaus and the depth of compensation of hot-spot
swells [30].
3. If the gravity field is known over a large area, but there is rather sparse ship-track
coverage, the topography/gravity transfer function can be used to interpolate the
seafloor depth among the sparse ship soundings [75].
197
CHAPTER 17. GRAVITY/TOPOGRAPHY TRANSFER FUNCTION 198
17.2 Flexure Theory
In Chapter 8 we developed an analytic solution for the response of a thin-elastic plate

floating on a fluid mantle that is subjected to a line load. Here we follow the same
approach but solve the flexure equation for an arbitrary vertical load representing, for
example, the loading of the lithosphere due to the weight of a volcano as shown in
t(x) w
-s
w(x) c
-s-d
m
Figure 17.1
Figure 17.1, where s is the mean ocean depth (∼4km) and d is the thickness of the
crust (∼6km). The topography of the Moho is equal to deflection of the elastic plate
w(x). The topography of the seafloor, t(x), has two components: the topographic load,
to (x), and the deflection of the elastic plate w(x).
t(x) = to (x) + w(x) (17.1)
For this calculation, we make the following assumptions: the thickness of the elastic
plate is less than the flexural wavelength, the deflection of the elastic plate is much
less than the flexural wavelength, the flexural rigidity, D, is constant, and there is no
end-load on the plate, so F = 0 . The vertical force balance for flexure of a thin elastic
plate floating on the mantle is described by the following differential equation:
∂ ∂2 ∂4
4
D 4 + 2 4 2 + 4 w(x) + ( ρm − ρw ) gw(x) = − ( ρc − ρw ) gto (x) (17.2)
∂x ∂x ∂y ∂y
where the parameters are defined in Table 17.1.

w(x) deflection of plate m

(positive up)
Eh3
d= 12(1−v2 )
flexural rigidity Nm
h elastic plate thickness m
ρw seawater density 1025 kg m3
ρc crust density 2800 kg m3
ρm mantle density 3300 kg m3
v Poisson’s ratio 0.25
Table 17.1
Take the 2-D Fourier transform of equation (17.2) to reduce the differential equation to
an algebraic equation:
h i
D(2π)4 k4x +2k2x ky2 +ky4 W(k)+( ρm − ρw ) gW(k) = − ( ρc − ρw ) g T (k)−W(k) (17.3)
where we have used equation (17.1) to replace T o (k). With a little algebra and noting
that |k|4 = (k2x + ky2 )2 this can be rewritten as
D (2π |k|)4 W(k) + ( ρm − ρc ) gW(k) = − ( ρc − ρw ) gT (k) (17.4)

Now one can solve for the deflection of the elastic plate in terms of the observed
topography
#−1
− ( ρc − ρw ) D (2π |k|)4
"
W(k) = 1+ T (k) (17.5)
( ρm − ρc ) g ( ρm − ρc )
This equation is called the isostatic response function because it describes the topog-
raphy of the Moho in terms of the topography of the seafloor. Define the flexural
wavelength
1/4 √
D
λ f = 2π g ρ − ρ = 2πα (17.6)
m c
(Note α is the flexural parameter from Chapter 8. When the wavelength of the topog-
raphy is much greater than the flexural wavelength, then the topography of the Moho
follows the Airy-compensation model; this is compensated topography.
− ( ρc − ρw )
W(k) = T (k) (17.7)
( ρm − ρc )
In contrast, when the wavelength of the topography is much less than the flexural wave-
length, the topography of the Moho is zero; this is uncompensated topography. The
gravity field of the earth is very sensitive to the degree of isostatic compensation, so it
is useful to develop the gravity field for this model.
17.3 Gravity/Topography Transfer Function
The gravity anomaly for this model is approximated by compressing the topography
into a sheet mass where the surface density is ( ρc − ρw ) t(x). Similarly the Moho to-
pography is compressed into a sheet mass with surface density ( ρm − ρc ) w(x). Finally
the gravity anomaly in each layer is upward-continued to the ocean surface. See Fig-
ure 17.2.
y
x
t(x)(ρc - ρw)
-s
w(x)(ρm - ρc)
-s-d
Figure 17.2
The solution to Poisson’s equation (16.12) provides an approximate method of con-

structing a gravity model for the combined model.
∆g(k) = 2πG ( ρc − ρw ) e−2π|k|s T (k) + 2πG ( ρm − ρc ) e−2π|k|(s+d) W(k) (17.8)
Using equation (17.5), this can be rewritten in terms of the observed topography
( −1 )
D (2π |k|)4

∆g(k) = 2πG ( ρc − ρw ) e −2π|k|s
1 − 1 + g(ρ − ρ ) e −2π|k|d
T (k) (17.9)
m c
Δg(k) = 2π G( ρ c − ρ w )e −2 π k sT (k) + 2π G( ρ m − ρ c )e −2 π k (s+ d) W(k) (8)
Using equation 5, this can be re-written in terms of the observed topography
  4 −1 
  D( 2π k )  −2π k d 
−2 π k s
Δg(k) = 2π G( ρ c − ρ w )e 1 − 1+ e T(k). (9)
  g(ρ m − ρ c )  
This formulation provides a direct approach to constructing gravity anomaly models from
seafloor topography:
This formulation i) take athe
provides 2-Dapproach
direct fourier transform of thegravity
to constructing topography;
anomalyii)models
multiply by the
gravity-to-topography transfer
from seafloor topography: function;
i) take and
the 2-D iii) take
Fourier the inverse
transform of thefourier transform
topography; of the result.
ii) mul-
Thetiply
most
byimportant parameter is thetransfer
the gravity-to-topography elastic plate thickness
function, that
and iii) is the
take usedinverse
to estimate the flexural
Fourier
rigidity. Theoffigure
transform below
the result. shows
The mostthe gravity/topography
important parameter is thetransfer
elasticfunction for a range
plate thickness that of elastic
thicknesses.
is used to estimate the flexural rigidity. Figure 17.3 shows the gravity/topography trans-
Figure 17.3
Since the asthenosphere relieves stresses on geological timescales, there is no truly-
fer function fortopography.
uncompensated a range of elastic
Thusthicknesses.
the gravitySince the asthenosphere
anomaly relievesstructures
for very large-scale stresses such as
on geological timescales, there is no truly uncompensated topography. Thus the
continents and hot-spot swells is small or zero far from the edges of these features. gravity
It is only the
anomaly for very large-scale structures, such as continents and hot-spot swells,
sharp topographic features such as large seamounts that will have prominent gravity is small
expressions.
or zero far from the edges of these features. It is only the sharp topographic features,
such as large seamounts, that will have prominent gravity expressions.
The homework problem is to take a topography profile that crosses the Hawaiian Ridge,
construct a model gravity profile for a variety of elastic thicknesses, and compare the
result with the observed gravity.
17.4 Geoid/Topography Transfer Function
Using the formulas for converting between geoid and gravity (derived in the Chapter
14 on Laplace’s equation) it is a simple matter to develop the geoid/topography transfer
function:
1
N(k) = 2π g ∆g(k)
|k|
(17.10)
N(k) G ( ρc − ρw ) −2π|k|s
( −1 )
D (2π |k|)4

= e 1 − 1 + g ( ρm − ρc )
e−2π|k|d
T (k) |k| g
This geoid topography transfer function has some interesting properties, as illustrated
in Figure 17.4. The amplitude of the geoid/topography transfer function is typically
wavelength (km)
1000 100 10
3.5
geoid/topography (m/km)
3
2.5
km
30
2
h=
isostatic geoid
1.5 m
anomaly
20k
1 m
h= 10 k
h=
0.5 Airy
0
-4 -3 -2 -1
10 10 10 10
wavenumber 1/km
Figure 17.4
0 to 4 meters per kilometer. This means that a seamount that is 1 km tall will pro-
duce a bump on the ocean surface that is about 1 meter tall. Since satellite altimeters
have accuracy of better than 0.1 m, such seamounts will be apparent in global geoid
height maps. The shape of the geoid/topography transfer functions also depends on the
thickness of the elastic plate.
There is one important difference between the geoid and the gravity anomaly. At long
wavelengths, the gravity/topography transfer function goes to zero because the topog-
raphy is isostatically compensated, so the gravity signal from the topography is exactly
cancelled by the gravity signature from the Moho. In contrast, the geoid/topography
transfer function goes to a constant value at long wavelengths. We’ll exploit this long-
wavelength behavior in the next section to develop a very simple approach to construct-
ing geoid height models.
17.5 Isostatic Geoid Anomalies
Section 5-12 of Geodynamics [81] has a nice discussion of isostatic geoid anomalies,
including several of the more important applications of this approximation. However,
their derivation is performed in the space domain. I think the wavenumber domain
derivation is easier to understand and it follows the methods used in this book. In
Chapter 15 on Poisson’s equation, we derived the formula for computing geoid height
from an arbitrary 3-D density model ∆ρ(x, y, z):
Z0
G e2π|k|z
N(k) = g
∆ρ (k, z) dz (17.11)
|k|
−∞
where ∆ρ(k, z) = =2 ∆ρ(x, z) . Also see Figure 17.5.

y
x
∆ρ(x, y, z)
Figure 17.5
If the topography is isostatically-compensated, then

Z0 Z0
0= ∆ρ(x, z) = ∆ρ(k, z) dz (17.12)
−∞ −∞
Now expand the exponential in equation (17.11) in a Taylor series about zero wavenumber:
Z0
G
(2π |k| z)2

N(k) = ∆ρ(k, z) 1 + 2π |k| z + 2!
+ · · · dz (17.13)
g |k|
−∞
Note that the first term in the brackets represents the integral of the density anomaly
over depth and that this is zero because of isostasy (equation (17.12)). Next we assume
that the wavelength of the anomaly is much greater than the depth of compensation
λ z or |k|z 1. In this long wavelength limit, the third and higher-order terms are
small compared with the second term. The integral simplifies to
Z0
2πG
N(k) = ∆ρ(k, z) z dz. (17.14)
g
−∞
Now take the inverse Fourier transform of 17.14.

Z0
2πG
N(x) = ∆ρ(x, z) z dz. (17.15)
g
−∞
This is a remarkable results because this formula provides a way to compute the geoid
height simply by integrating the density anomaly only over depth. This integration
can be done for a variety of models, including Airy compensation, Pratt compensation,
and thermal compensation (i.e., spreading ridge or thermal swell). Several of these
integrals are done in Geodynamics [81].
17.6 Isostatic Geoid and the Swell-Push Force
The three driving forces of plate tectonics are slab pull, asthenospheric drag, and ridge
push. Ridge-push force is the outward force due to isostatically-compensated topog-
raphy. This mainly occurs as a gravitational sliding force on the flanks of the seafloor
spreading ridges. However, it is also associated with the outward force due to thick
continental crust or a thermal swell. One of your homework problems was to develop a
general expression for this swell-push force for isostatically-compensated topography.
Here I’ll derive this expression again and point out that the force integral is identical to
the integral for computing isostatic geoid anomaly. Therefore the two are related and
one can use geoid height to measure ridge-push force [53, 14, 22, 23].
Consider isostatically-compensated topography as shown in Figure 17.6. While this di-

agram is related to a specific Airy-type compensation mechanism, the integral relation
is, in fact, quite general. To calculate the total outward force F s due to this isostati-
cally compensated plateau, we integrate the difference in pressure between column (1)
and column (2) over depth to the depth of compensation −L (i.e. where the pressure
difference is zero).
Z0
Fs = ∆P(z) dz (17.16)
−L
Isostatic Geoid and the Swell-Push Force
The three driving forces of plate tectonics are slab pull, asthenospheric drag, and ridge push.
Ridge-push force is the outward force due to isostatically-compensated topography. This mainly
occurs as a gravitational sliding force on the flanks of the seafloor spreading ridges. However, it
is also associated with the outward force due to thick continental crust or a thermal swell. One
of your homework problems was to develop a general expression for this swell-push force for
isostatically-compensated topography. Here I'll derive this expression again and point out that
the force integral is identical to the integral for computing isostatic geoid anomaly. Therefore
the two are related and one can use geoid height to measure ridge-push force [Parsons and
€ Richter, 1980; Dahlen, 1981; Fleitout and Froidevaux, 1982; 1983].
Consider isostatically-compensated topography as shown in the diagram below.
(1) (2) P(z) ΔP(z)
ρc
(2) (1)
ρm
-L
Figure 17.6
Integrate by parts
0
∂∆P(z)
0 Z
F s = ∆P(z)z−L − z dz. (17.17)
∂z
−L
Note the first term on the right is zero because of isostasy. The second term can be
written in terms of the density by noting that the vertical gradient in the pressure dif-
ference is
∂∆P(z)
∂z
= −g∆ρ(z). (17.18)
The result is
Z0
Fs = g ∆ρ(z)z dz. (17.19)
−L
Comparing equation (17.15) to equation (17.19), we see the integrals are the same, so
there is a direct relationship between swell-push force and geoid height.
g2
F s = 2πG N (17.20)
We’ll use this result later to calculate the ridge-push force versus age from the geoid
height versus age relation.
17.7 Exercises
Exercise 17.1. (a) Use the formula for the geoid height due to long wavelength isostatically-
compensated topography (equation 17.15) to calculate the geoid height due to the fol-
lowing density model ∆ρ = σ [δ (z) − δ (z + a)].
(b) What is the change in geoid height for a topography step of 4 km, a density of 2800
kg m−3 and a compensation depth a of 40 km?
(c) What is the magnitude of the outward swell push force? What is the depth-averaged
stress needed to maintain this topography?
Chapter 18
Postglacial Rebound
18.1 Introduction and Dimensional Analysis
This chapter considers the classic problem of the viscous response of the mantle to
rapid melting of the ice sheets following the last glacial maximum. The approach
is similar to section 6-10 in Turcotte and Schubert [81] but is for an arbitrary shaped
initial topography rather than a single wavelength cosine function. The initial condition
is shown in Figure 18.1.
Figure 18.1: Viscous half space with an initial surface topography that decays expo-
nentially with time under the restoring force of gravity.
206
CHAPTER 18. POSTGLACIAL REBOUND 207
The main parameters are:
T (x) initial topography (m)

η viscosity (Pa s)
ρ density (kg m−3 )
g acceleration of gravity (m s−2 )
u x-velocity (m s−1 )
w z-velocity (m s−1 )
Guess at a Solution: Dimensional Analysis

We can make an initial guess at the time evolution of the topography T (t)
assuming a single wavelength λ for the initial surface topography T (0). The
guess is
T (t) = T (0)e−t/τr . (18.1)
The relaxation time should increase as the viscosity increases and decrease
as the restoring force increases, so we put these in the numerator and de-
moninator, respectively: η/ρg. However, the result has dimensions of m s.
To make this into a time we can divide by the wavelength resulting in our
initial guess at the relaxation time τr = η/ρgλ. We’ll compare this with the
relaxation time based on the full derivation.
18.2 Exact Solution
We’ll assume the mantle is incompressible ∇ • ~u = 0 and the model is 2-D, so this
requires that
∂u ∂w
=− (18.2)
∂x ∂z
As discussed in T&S [81], we can define a stream function ψ(x, z) such that
∂ψ ∂ψ
u=− , w= . (18.3)
∂z ∂x
This ensures that the material is incompressible
∂u ∂w ∂2 ψ ∂2 ψ
+ =− + =0 (18.4)
∂x ∂z ∂x∂z ∂z∂x
The stresses can be related to the stream function as

∂u ∂2 ψ
τ xx = 2η ∂x = −2η ∂x∂z
∂w ∂2 ψ
τzz = 2η ∂z = 2η ∂z∂x (18.5)
∂2 ψ ∂2 ψ

τ xz = η −
∂x2 ∂z2
The force balance equations become (i.e., equations (6.72) and (6.73) in [81])
∂3 ψ ∂3 ψ ∂P
!
η + + =0
∂x2 ∂z ∂z3 ∂x
(18.6)
∂3 ψ ∂3 ψ ∂P
!
η + − =0
∂x3 ∂z2 ∂x ∂z
Following T&S [81] we take the derivative of the first equation with respect to z, the
second equation with repect to x, and add them to obtain the biharmonic equation.
∂4 ψ ∂4 ψ ∂4 ψ
+ 2 + = ∇4 ψ = 0 (18.7)
∂x4 ∂z2 ∂x2 ∂z4
The boundary conditions for this problem are that the solution must vanish for large z,
and that the surface shear stress is zero.
lim ψ (x, z) = 0
x→∞
(18.8)
∂P ∂T

∂x 0
= −ρg ∂x
Note that uniform topography does not drive any flow. The flow is driven by the hori-
zontal pressure gradient set up by the topography gradient. Take the Fourier transform
of the biharmonic equation with respect to x to arrive at
∂2 ψ ∂4 ψ
(i2πk)4 ψ + 2(i2πk)2 + =0 (18.9)
∂z2 ∂z4
We guess a general solution of the form
ψ (k, z) = A (k) e−2π|k|z + B (k) ze−2π|k|z + C (k) e2π|k|z + D (k) ze2π|k|z (18.10)
The solution must go to zero for large z, so C = D = 0 and the remaining terms are
ψ (k, z) = (A + Bz) e−2π|k|z (18.11)

Next we use the boundary condition that the shear traction must vanish at the surface:

∂ ψ ∂2 ψ
2
τ xz | 0 = η − 2 = 0 (18.12)
∂x ∂z
2

0
We need to compute these derivatives

∂2 ψ
= −(2πk)2 ψ,
∂x2
∂ψ
∂z
= [B − 2π |k| (A + Bz)] e−2π|k|z
(18.13)
and
∂2 ψ h i
= −2B (2π |k|) + (2π |k|)2 (A + Bz) e−2π|k|z
∂z2
The boundary condition becomes

τ xz

= −2(2π |k|)2 A + 2 (2π |k|) B = 0 so B = 2π |k| A (18.14)
η 0
The stream function and the two velocity components are
ψ (k, z) = A (1 + 2π |k| z) e−2π|k|z
u = A(2π |k|)2 ze−2π|k|z (18.15)
w = A (i2πk) (1 + 2π |k| z) e−2π|k|z

Finally we need to match the surface pressure gradient boundary condition:
∂P ∂T

∂x 0
= −ρ g ∂x (18.16)
From the force balance equation we have

∂P ∂3 ψ ∂3 ψ
!
= −η + (18.17)
∂x ∂x2 ∂z ∂z3
But we know that
∂ψ

∂z 0
=0 (18.18)
so only one term remains. In the transform domain this boundary condition becomes
∂3 ψ
η = ρ g (i2πk) T (k) (18.19)
∂z3
Taking the third derivative of the stream function and solving for A we find
(i2πk) ρ gT (k)
A= . (18.20)
2η(2πk)3
Putting this result in to the equation for the vertical velocity we find
−ρ g
w (k, 0) = 2η(2π T (k) (18.21)
|k|)
We also know that w = ∂T/∂t, so we end up with a differential equation for the time
evolution of the topography
∂T ρg
=− T. (18.22)
∂t 4π |k| η
The solution to this differential equation is
ρg
T (k, t) = T (k, 0) e− 4π|k|η t . (18.23)
From this we find the characteristic relaxation time is

4πη
τr = ρ gλ (18.24)
Note that this exact solution differs from the initial guess by 4π, which is OK for a first
order calculation.
Let’s consider the example of Fennoscandia which has a characteristic wavelength

of 3000 km a mantle density of 3300 kg m−3 and a characteristic relaxation time of
4400 yr. Using the formula we arrive at a viscosity of 1.1 × 1021 Pa s.
18.3 Elastic plate over a viscous half space
We can construct a more realistic model by placing an elastic lithosphere over a viscous
half space. We begin with an ice load that has deformed the lithosphere for infinite time
so the deflection of the elastic plate T (x) follows the flexural response function given
in Chapter 8. The differential equation for a line load of magnitude Vo is
∂4 T
D + ρgT (x) = Vo δ (x) (18.25)
∂x4
where D is the flexural rigidity defined in Chapters 8 and 17. The solution for the
deflection of the plate in the Fourier transform domain is
 −1
Vo  k4 
T (k) = 1 + 4  (18.26)
ρg  kf
1 ρg 1/4
h i
where k f = 2π D is the flexural wavenumber and is equal to the inverse flexural
wavelength. At time zero we remove this line load from the lithosphere and solve for
the viscous rebound of the lithosphere and mantle. From equation 18.23 we find the
topography at some later time is
 −1
Vo  k4  − 4π|k|η
ρg
T (k, t) = 1 + 4  e . (18.27)
ρg  kf
The response at a later time is easily computed with the following matlab program
where we construct the Fourier transform of the line-load flexure, multiply each wavenum-
ber by the appropriate exponential decay, and inverse transform.
%
% Matlab program to compute rebound of an elastic plate over a viscous half space
%
clear;clf;
L=10000000; % make region wide enough to avoid edge effects
g0=9.8; % acceleration of gravity
E=7.e10; % Young’s modulus
h=100000; % elastic plate thickness
nu=.25; % Poisson’s ratio
rho=3300; % mantle density
eta=5.e20; % dynamic viscosity
V0=4000*rho*g0; % ice sheet load in Pascals
D=E*h.ˆ3/(12*(1-nu*nu)); % flexural rigidity
kf=(rho*g0/D).ˆ.25/(2.*pi); % flexural wavenumber = 1/flexural_wavelength
tmax=10000*86400*365; % maximum time
%
% set the location of the line load at L/2
%
N=1024; dx=L/N; x=(1:N)*dx;
topo=zeros(N,1); topo(N/2)=V0/(rho*g0);
%
% take the Fourier transform of the load and generate wavenumbers
%
ctopo=fftshift(fft(topo));
k=-N/2:1:(N/2-1); k=(k./L)’; ak=abs(k);
%
% compute the Fourier transform of the load as well as the
% response time for each wavenumber
%
CW0=-1./(1+(k./kf).ˆ4);
TR=4*pi*eta.*abs(k)/(rho*g0);
%
% calculate the topography at 4 time steps
%
nstep=4; dt=tmax/nstep;
for i = 1:nstep
time=dt*(i);
years=time/(86400*365)
cmod=ctopo.*CW0.*exp(-time./TR);
mod=real(ifft(fftshift(cmod)));
subplot(3,1,1),plot(x/1000.,mod);axis([3000,7000,-70,40]);ylabel(’topography (m) ’)
if i == 1
hold
%
% also plot stream function for this case
%
nd=20; dmax=2000000; dz=dmax/(nd-1);
stream=zeros(nd,N);
Ak=i*2*pi*rho*g0*k.*cmod./(2*eta*(2*pi*ak).ˆ3);
for n = 1:nd
z=(n-1)*dz; argz=2*pi*ak*z;
cstr=Ak.*(1+argz).*exp(-argz);
str=imag(ifft(fftshift(cstr)));
for jj = 1:N
stream(n,jj)=str(jj);
end
end
end
zz=-(0:nd-1)*dz;
subplot(3,1,2),contour(x/1000,zz/1000,stream,40); axis(’equal’);
axis([3000,7000,-1000,0]); xlabel(’distance (km)’); ylabel(’depth (km)’);
end
We show the topography at 4 times following the removal of the load in Figure 18.2 (top).
The topography rebounds in the area where the load was removed and subsides on ei-
ther side of the load. The characteristic wavelength of this signature is about equal to
the flexural wavelength, which is related to the thickness of the lithosphere. We found
in Chapter 9 that a typical elastic thickness for flexure on million year and greater
timescales is 30 km. However, to explain the width of the peripheral bulge seen in
vertical velocity profiles across Laurentide and Fenoscandia (e.g., Figure 18.3) a much
thicker elastic lithosphere (∼100km) is needed.
This vertical land motion has a can have significant effect on relative sea level at shore-
lines. The mean global sea level rise due to ocean volume changes (mass and thermal)
is ∼2.8 mm/yr. However, if you live in Churchill, Canada where emergence rate is
maximum you will record a sea level fall of 9 mm/yr but if you live on the peripheral
bulge in Hampton Roads, United States, you will record a sea level rise of more than 4
mm/yr.
Of course, this 2-D line load model with uniform mantle viscosity is only a crude
approximation to the actual 3-D loading and viscosity structure. Moreover the redis-
tribution of mass as shown by the stream function in Figure 18.2 (bottom) produces a
change in the geoid height [48] which further complicates the global variations in sea
level. Nevertheless, this simple 2-D model explains the basic features observed in past
and present-day postglacial rebound.
40
subsidence rebound subsidence
20
topography (m)
0
-20 10000 yr
7500 yr
-40
5000 yr
-60 2500 yr
3000 4000 5000 6000 7000
stream function
-200
-400
depth (km)
-600
-800
-1000
3000 4000 5000 6000 7000
distance (km)
Figure 18.2: (top) Topography calculated at 4 times after the removal of a line load
from an elastic plate over a viscous half space. (bottom) Stream function 10,000 years
after the removal of the load shows upward flow of the mantle beneath the rebound
area and downward flow in surrounding areas.
Relative Sea Level Trends and Distance from Hudson Bay

6
Hampton Roads, VA
Pensacola, FL
4 Halifax, NS Bermuda Key West Balboa Honolulu
New York
Relative Sea Level Rise, MM/Yr
2
Portland, ME
San Diego
0
0 1000 2000 3000 4000 5000 6000 7000
-2 Great Circle Distance from Churchill (Hudson Bay), KM
-4
-6
-8
Churchill
-10
Figure 18.3: Rate of relative sea level rise versus radial distance from Hudson Bay,
Canada derived from tide guage records. The average global sea level rise over the
past 100 years was about 2 mm/yr (green dashed line) although it has increased to
about 2.8 mm/yr today. Areas within the perimeter of the ancient Laurentide ice sheet
are still rebounding at rates up to 9 mm/yr so relative sea level is falling. Areas on the
peripheral bulge are subsiding due to unflexing of the 100 km thick elastic plate. Ocean
islands far from tectonic features provide the best sites for measuring global sea level
rise. From [46].
18.4 Exercises
h wine cellar depth 3m
η dynamic viscosity 1020 Pa s
ρ mantle 3300 kg m3
λ wavelength of surface 3000 km

deformation
Table 18.1: Parameters needed for exercise 18.1
Exercise 18.1. Given the parameters in table 18.1, develop (guess) characteristic times
for the following processes.
a) Heat diffusion - Describe this timescale in terms of an experiment or process.
b) Maxwell viscoelastic relaxation - Describe this timescale in terms of an experiment
or process.
c) Glacial rebound, viscosity - Describe this timescale in terms of an experiment or
process.
Bibliography
[1] R.J. Banks, R.L. Parker, and S.P. Heustis. Isostatic compensation on a continental
scale: Local versus regional mechanisms. Geophys. J. R. Astr. Soc., 51:431–452,
1977.
[2] J.J. Becker, D.T. Sandwell, W.H.F. Smith, J. Braud, B. Binder, J. Depner,
D. Fabre, J. Factor, S. Ingalls, S.H. Kim, and R. Ladner. Global bathymetry and
elevation data at 30 arc seconds resolution: SRTM30-PLUS. Marine Geodesy,
32(4):355–371, 2009.
[3] J. Blakley. Potential Theory in Gravity and Magnetics. Cambridge University
Press, New York, 1995.
[4] W.F. Brace and D.L. Kohlstedt. Limits on lithospheric stress imposed by labora-
tory experiments. J. Geophys. Res., 85(B11):6248–6252, 1980.
[5] R.N. Bracewell. The Fourier Transform and Its Applications. McGraw-Hill Book
Co., New York, second edition, 1978.
[6] J. Byerlee. Friction of rocks. In Rock friction and earthquake prediction, pages
615–626. Springer, 1978.
[7] S.C. Cande, J.L. LaBrecque, R.L. Pitman, X. Golovchenko, and W.F. Haxby.
Magnetic Lineations of the World’s Ocean Basins. American Association of
Petroleum Geologists, Tulsa, OK, 1989.
[8] D.W. Caress, M.K. McNutt, R.S. Detrick, and J.C. Mutter. Seismic imaging
of hotspot-related crustal underplating beneath the Marquesas Islands. Nature,
373:600–603, February 1995.
[9] H.S. Carslaw and J.C. Jaeger. Conduction of Heat in Solids. Oxford University
Press, Oxford, UK, second edition, 1959.
[10] C. P. Chen and P. Molnar. Focal depths of intracontinental and intraplate earth-
quakes and their implications for the thermal and mechanical properties of the
lithosphere. J. Geophys. Res., 88:4183–4214, 1983.
215
BIBLIOGRAPHY 216
[11] S.C. Cohen. Numerical models of crustal deformation in seismic zones. Advances
in Geophysics, 41:134–231, 1999.
[12] S.C. Cohen and D.E. Smith. Lageos scientific results: Introduction. Journal of
Geophysical Research: Solid Earth, 90(B11):9217–9220, 1985.
[13] S.T. Crough. Hotspot swells. Annual Review of Earth and Planetary Sciences,
11(1):165–193, 1983.
[14] F.A. Dalen. Isostasy and the ambient state of stress in the oceanic lithosphere. J.
Geophys. Res., 86:7801–7807, 1981.
[15] C. DeMets, R.G. Gordon, D.F. Argus, and S. Stein. Current plate motions. Geo-
phys. J. R. Astr. Soc., 101:425–478, 1990.
[16] C. DeMets, R.G. Gordon, D.F. Argus, and S. Stein. Geologically current plate
motions. Geophys. J. Int., 181:1–80, 2010.
[17] M.P. Doin and L. Fleitout. Thermal evolution of the oceanic lithosphere: An
alternate view. Earth Planet. Sci. Lett., 142:121–136, 1996.
[18] L.M. Dorman and B.T.R. Lewis. Experimantal isostasy 3: Inversion of the iso-
static Green’s function and lateral density changes. J. Geophys. Res., 77:3068–
3077, 1972.
[19] Sandwell D.T., D.L. Anderson, and P. Wessel. Global tectonic maps. SPECIAL
PAPERS-GEOLOGICAL SOCIETY OF AMERICA, 388:1, 2005.
[20] C.J. Ebinger, G.D. Karner, and J.K. Weissel. Mechanical strength of extended
continental lithosphere: Constraints from the Western rift system. Tectonics,
10:1239–1256, 1991.
[21] E.R. Engdahl, R. VanderHilst, and R. Buland. Global teleseismic earthquake
relocation with improved travel times and procedures for depth determination.
IBull. Seismo. Soc. Am., 88:722–743, 1998.
[22] L. Fleitout and C. Froidevaux. Tectonics and topography for a lithosphere con-
taining density heterogeneities. Tectonics, 1:21–56, 1982.
[23] L. Fleitout and C. Froidevaux. Tectonic stresses in the lithosphere. Tectonics,
2:315–324, 1983.
[24] L.M. Flesch, A.J. Haines, and W.E. Holt. Dynamics of the India-Eurasia collision
zone. Journal of Geophysical Research: Solid Earth, 106(B8):16435–16460,
2001.
[25] C.M.R. Fowler. The Solid Earth: An Introduction to Global Geophysics. Cam-
bridge University Press, Cambridge, MA, 1990.
[26] E.S. Garcia, D.T. Sandwell, and K.M. Luttrell. An iterative spectral solution
method for thin elastic plate flexure with variable rigidity. Geophysical Journal
International, 200(2):1012–1028, 2014.
BIBLIOGRAPHY 217
[27] G.D. Garland. The Earth’s Shape and Gravity. Permagon Press, 1977.
[28] J. Gee and D. Kent. Variation in layer 2A thickness and the origin of the central
anomaly magnetic high. Geophys. Res., Lett., 21(4):297–300, 1994.
[29] W.F. Haxby, G.D. Garner, J.L. LaBrecque, and J.K. Weissel. Digital images of
combined oceanic and continental data sets and their use in tectonic studies. EOS
Trans. Amer. Geophys. Un., 64:995–1004, 1983.
[30] W.F. Haxby and D.L. Turcotte. On isostatic geoid anomalies. J. Geophys. Res.,
83:5,473–5,478, 1978.
[31] W.A. Heiskanen and F.A. Vening Meinesz. The Earth and Its Gravity Field.
McGraw-Hill, 1958.
[32] C. Hogg and G. Laske. The cooling oceanic lithosphere as constrained by surface
wave dispersion data. In SOS Trans. AGU. Fall AGU Meeting. AGU, 2003.
[33] C. Hwang and B. Parsons. An optimal procedure for deriving marine gravity from
multi-satellite altimetry. J. Geophys. Int., 125:705–719, 1996.
[34] J Jackson. Classical Electrodynamics. New York: Wiley, 1975.
[35] J.D. Jackson. Classical Electrodynamics. John Wiley & Sons, New York, third
edition, August 1998.
[36] G. Kent, A. Harding, and J. Orcutt. Distribution of magma beneath the East Pa-
cific rise between the Clipperton Transform and the 9◦ 17’ deval from forward
modeling of common depth point data. J. Geophys. Res., 98:13,945–13,969,
1993.
[37] A.H. Lachenbruch and J.H. Sass. Heat flow and energetics of the San Andreas
fault zone. J. Geophys. Res., 85:6185–6222, 01 1980.
[38] A.R. Leeds and E.G. Kausel. Variations of upper mantle structure under the Pa-
cific Ocean. Science, 186:141–143, 1974.
[39] F.G. Lemoine, S.C. Kenyon, J.K. Factor, R.G. Trimmer, N.K. Palvis, D.S.
Chinn, C.M. Cox, S.M. Klosko, S.B. Luthcke, M.H. Torrence, Y.M. Wang, R.G.
Williamson, E.C. Palvis, R.H. Rapp, and T.R. Olson. The development of the joint
NASA GSFC and the National Imagery and Mapping Agency (NIMA) geopoten-
tial model EGM96. Technical report, NASA, 1998.
[40] D.A. Levitt and D.T. Sandwell. Lithospheric bending at subduction zones based
on depth soundings and satellite gravity. J. Geophys. Res.: Solid Earth, 100:379–
400, 1995.
[41] C.G. Massell. Large Scale Structural Variation of Trench Outer Slopes and Rises.
Ph.D. dissertation, Scripps Inst. of Oceanography, La Jolla, CA, 2002.
BIBLIOGRAPHY 218
[42] D. McKenzie and C. Bowin. The relationship between bathymetry and gravity in
the Atlantic Ocean. Journal of Geophysical Research, 81:1903–1915, 1976.
[43] D.P. McKenzie. Some remarks on heat flow and gravity anomalies. J. Geophys.
Res., 72:6,261–6,273, 1976.
[44] M. McNutt. Compensation of oceanic topography: An application of the response

function technique to the surveyor area. J. Geophys. Res., 84:7,589–7,598, 1979.
[45] M.K. McNutt and H.W. Menard. Constraints on yield strength in the oceanic
lithosphere derived from observations of flexure. Geophs. J. R. astr. Soc., 71:363–
394, 1982.
[46] Laury Miller and Bruce C Douglas. On the rate and causes of twentieth century
sea-level rise. Philosophical Transactions of the Royal Society of London A:
Mathematical, Physical and Engineering Sciences, 364(1841):805–820, 2006.
[47] J.B. Minster and T.H. Jordan. Present-day plate motions. J. Geophys. Res.,
85:5331–5354, 1978.
[48] Jerry X Mitrovica, Natalya Gomez, and Peter U Clark. The sea-level fingerprint
of West Antarctic collapse. Science, 323(5915):753–753, 2009.
[49] R.D. Muller, W.R. Roest, J.Y. Royer, L.M. Gahagan, et al. Digital isochrons of
the world’s ocean floor. J. Geophys. Res., 102:3211–3214, 1997.
[50] ER Oxburgh and EM Parmentier. Compositional and density stratification in

oceanic lithosphere-causes and consequences. Journal of the Geological Society,
133(4):343–355, 1977.
[51] J.E. Pariso and H.P. Johnson. Alteration processes at deep sea drilling
project/ocean drilling program hole 504B at the Costa Rica Rift: Implications
for magnetization of oceanic crust. J. Geophys. Res., 96:11,703–11,722, 1991.
[52] R.L. Parker. The rapid calculation of potential anomalies. Geophys. J. Roy. Astr.
Soc., 31:441–455, 1973. Required reading for anyone interested in marine grav-
ity/magnetics.
[53] B. Parsons and F.M. Richter. A relationship between the driving force and geoid
anomaly associated with mid-ocean ridges. Earth Planet. Sci. Lett., 51:445–450,
1980.
[54] B. Parsons and J.G. Sclater. An analysis of the variation of the ocean floor
bathymetry and heat flow with age. J. Geophys. Res., 82:803–827, 1977.
[55] N.K. Pavlis, S.A. Holmes, S.C. Kenyon, and J.K. Factor. The development and
evaluation of the Earth Gravitational Model 2008 (EGM2008). Journal of geo-
physical research: solid earth, 117(B4), 2012.
BIBLIOGRAPHY 219
[56] J. Phipps-Morgan, A. Harding, J. Orcutt, and Y.J. Chen. An observational and

theoretical synthesis of magma chamber geometry and crustal genesis along a
mid-ocean ridge spreading center. In M. P. Ryan, editor, Magmatic Systems.
Academic Press Inc., Cambridge, MA, 1994.
[57] R.H. Rapp and Y. Yi. Role of ocean variability and dynamic topography in the
recovery of the mean sea surface and gravity anomalies from satellite altimeter
data. J. Geodesy, 71:617–629, 1997.
[58] M.L. Renkin and J.G. Sclater. Depth and age in the North Pacific. Journal of
Geophysical Research-Solid Earth and Planets., 93:2919–2935, 1988.
[59] M.P. Ryan, editor. Magmatic Systems. Academic Press, San Diego, 1994.
[60] D.T. Sandwell. Thermal isostasy: response of a moving lithosphere to a dis-
tributed heat source. J. Geophys. Res., 87:1001–1014, 1982.
[61] D.T. Sandwell. A detailed view of the South Pacific from satellite altimetry. J.
Geophys. Res., 89:1089–1104, 1984.
[62] D.T. Sandwell. Thermomechanical evolution of oceanic fracture zones. J. Geo-

phys. Res., 89:11,401–11,413, 1984.
[63] D.T. Sandwell and G. Schubert. Lithospheric flexure at fracture zones. J. Geo-
phys. Res., 87:4657–4667, 1982. doi:10.1029/JB087iB06p04657.
[64] D.T. Sandwell and G. Schubert. Evidence for retrograde lithospheric subduction
on Venus. Science, 257(5071):766–770, 1992.
[65] D.T. Sandwell and W.H.F. Smith. Marine gravity anomaly from Geosat and
ERS-1 satellite altimetry. J. Geophys. Res., 102:10,039–10,054, 1997.
[66] J.C. Savage. Equivalent strike-slip cycles in half-space and lithosphere-

asthenosphere earth models. J. Geophys. Res., 95:4873–4879, 1990.
[67] J.C. Savage and R.O. Burford. Geodetic determination of relative plate motion in
central California. Journal of Geophysical Research, 78(5):832–845, 1973.
[68] C.H. Scholz. The Mechanics of Earthquakes and Faulting. Cambridge University
Press, Cambridge, England, 1990.
[69] H. Schouten and K. McCamy. Filtering marine magnetic anomalies. J. Geophys.
Res., 77:7,089–7,099, 1972.
[70] G. Schubert and D. Sandwell. Crustal volumes of the continents and of oceanic
and continental submarine plateaus. Earth and Planetary Science Letters,
92(2):234–246, 1989.
[71] J.G. Sclater, C. Jaupart, and D. Galson. The heat flow through oceanic and con-
tinental crust and the heat loss of the Earth. Reviews of Geophys. and Space
Physics, 18:269–311, 1980.
BIBLIOGRAPHY 220
[72] T. Seno and Y. Yamanaka. Double seismic zones, compressional deep trench-
outer rise events, and superplumes. In G. E. Bebout, D. Scholl, S. Kirby, and
J. Platt, editors, Subduction Top to Bottom, volume 36, pages 347–355. American
Geophysical Union, Washington, DC, 1996.
[73] L. Siebert and T. Simkin. Volcanoes of the World: an Illustrated Cata-
log of Holocene Volcanoes and their Eruptions., volume GVP-3. Smithso-
nian Institution, Global Volcanism Program, Digital Information Series, 2002.
http://www.volcano.si.edu/world/.
[74] G.M. Smith and S.K. Banerjee. Magnetic structure of the upper kilometer of
the marine crust at DSDP hole 504B, Eastern Pacific Ocean. J. Geophys. Res.,
91:10,337–10,354, 1986.
[75] W.H.F. Smith and D.T. Sandwell. Bathymetric prediction from dense satellite
altimetry and sparse shipboard bathymetry. J. Geophys. Res., 99:21,803–21,824,
1994.
[76] W.H.F. Smith and D.T. Sandwell. Global sea floor topography from satellite al-
timetry and ship depth soundings. Science, 277:1956–1962, 1997.
[77] F.D. Stacey. Physics of the Earth. John Wiley & Sons, 1977.
[78] F.D. Stacey and P.M. Davis. Physics of the Earth. Cambridge University Press,
fourth edition, September 2008.
[79] J. Stewart and A.B. Watts. Gravity anomalies and spatial variations of flexural
rigidity at mountain ranges. J. Geophys. Res., 102:5327–5352, 1997.
[80] D.L. Turcotte and E.R. Oxburgh. Finite amplitude convection cells and continen-
tal drift. J. Fluid Mech., 28:29–42, 1967.
[81] D.L. Turcotte and G. Schubert. Geodynamics. Cambridge University Press, cam-
bridge, UK, third edition, 2014.
[82] D.L. Turcotte, P.H. Tag, and R.F. Cooper. A steady-state model for the distribution
of stress and temperature on the San Andreas fault. J. Geophys. Res., 85:6224–
6230, 1980.
[83] A. B. Watts, C. C. Pierce, J. Collier, R. Dalwood, J. Cannales, and T. J. Henstock.

A seismic study of lithospheric flexure in the vicinity of Tenerife, Canary Islands.
Earth Planet Sci. Lett., 146:431–447, 1997.
[84] A.B. Watts. An analysis of isostasy in the world’s oceans 1. Hawaiian-Emperor
Seamount Chain. Journal of Geophysical Research, 83, January 1978.
[85] A.B. Watts. Isostasy and Flexure of the Lithosphere. Cambridge University Press,
2001.
BIBLIOGRAPHY 221
[86] A.B. Watts and J.R. Cochran. Gravity anomalies and flexure of the lithosphere
along the Hawaiian-Emperor Seamount Chain. Geophysical Journal of the Royal
Astronomical Society, 38:119–141, 1974.
[87] A.B. Watts, G.D. Karner, and M.S. Steckler. Lithospheric flexure and the evo-
lution of sedimentary basins. In The Evolution of Sedimentary Basins, volume
305a, pages 249–281. Phil. Trans. R. Soc. Lond., 1982.
[88] A.B. Watts and C. Marr. Gravity anomalies and the thermal and mechanical
structure of rifted continental margins. In Rifted Ocean-Continent Boundaries,
pages 65–94. Kluwer Academic Publishers, Dordrecht, 1995.
[89] A.B. Watts and U.S. ten Brink. Crustal structure, flexure, and subsidence history
of the Hawaiian Islands. J. Geophys. Res, 94(B8):10,473–10,500, August 1989.
[90] J. Weertman and J. R. Weertman. Elementary dislocation theory. Macmillan,
1966.
[91] D.A. Weins and S. Stein. Intraplate seismicity and stresses in young oceanic
lithosphere. J. Geophys. Res., 89:11,442–11,464, 1984.
[92] J.K. Weissel and G.D. Karner. Flexural uplift of rift flanks due to mechanical
unloading of lithosphere during extension. J. Geophys. Res., 94:13,919–13,950,
1989.
[93] P. Wessel and W.H.F. Smith. New version of the Generic Mapping Tools released.
EOS Trans. AGU, 76:329, 1995. http://gmt.soest.hawaii.edu/.
[94] M.M. Yale, D.T. Sandwell, and W.H.F. Smith. Comparison of along-track resolu-
tion of stacked Geosat, ERS-1 and TOPEX satellite altimeters. J. Geophys. Res.,
100:15,117–15,127, 1995.

Apendice Geodynamics

Uploaded by

Copyright:

Available Formats

Apendice Geodynamics

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Apendice Geodynamics

Uploaded by

Copyright:

Available Formats

An Appendix to the Classic Textbook:

Geodynamics by Turcotte and Schubert

December 26, 2018

1 Observations Related to Plate Tectonics 1

1.1 Global Maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

2 Fourier Transform Methods in Geophysics 10

2.2 Definitions of Fourier Transforms . . . . . . . . . . . . . . . . . . . 11

2.3 Fourier Sine and Cosine Transforms . . . . . . . . . . . . . . . . . . 12

2.4 Examples of Fourier Transforms . . . . . . . . . . . . . . . . . . . . 13

2.5 Properties of Fourier transforms . . . . . . . . . . . . . . . . . . . . 17

2.7 Fourier series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

3.1 Plate Motions on a Flat Earth . . . . . . . . . . . . . . . . . . . . . . 26

3.2 Triple Junction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

3.3 Plate Motions on a Sphere . . . . . . . . . . . . . . . . . . . . . . . 31

3.4 Velocity Azimuth . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

3.5 Recipe for Computing Velocity Magnitude . . . . . . . . . . . . . . . 35

3.6 Triple Junctions on a Sphere . . . . . . . . . . . . . . . . . . . . . . 36

3.7 Hot Spots and Absolute Plate Motions . . . . . . . . . . . . . . . . . 36

4 Marine Magnetic Anomalies 38

4.2 Crustal Magnetization at a Spreading Ridge . . . . . . . . . . . . . . 38

4.3 Uniformly Magnetized Block . . . . . . . . . . . . . . . . . . . . . . 43

4.4 Anomalies in the Earth’s Magnetic Field . . . . . . . . . . . . . . . . 43

4.5 Magnetic Anomalies Due to Seafloor Spreading . . . . . . . . . . . . 44

5 Cooling of the Oceanic Lithosphere 52

5.2 Temperature versus Depth and Age . . . . . . . . . . . . . . . . . . . 56

5.3 Heat Flow versus Age . . . . . . . . . . . . . . . . . . . . . . . . . . 58

5.4 Thermal Subsidence . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

5.5 Buoyancy of the Cooling Lithosphere . . . . . . . . . . . . . . . . . 63

6 Crustal Structure, Isostasy, and Rheology 68

6.2 Oceanic Crustal Structure . . . . . . . . . . . . . . . . . . . . . . . . 69

6.3 Continental Crustal Structure . . . . . . . . . . . . . . . . . . . . . . 71

6.4 Vertical Force Balance - Isostasy . . . . . . . . . . . . . . . . . . . . 72

6.5 Horizontal Force Balance - Swell Push Force . . . . . . . . . . . . . 75

6.6 Rheology of the Lithosphere . . . . . . . . . . . . . . . . . . . . . . 78

7 Brief Review of Elasticity 80

7.3 Stress vs. Strain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

7.4 Invariants and Principal Stress . . . . . . . . . . . . . . . . . . . . . 82

7.5 Principal Stress and Strain . . . . . . . . . . . . . . . . . . . . . . . 83

8 Flexure of the Lithosphere 86

8.1 Constant Flexural Rigidity, Line Load, No End Load . . . . . . . . . 88

8.2 Variable Flexural Rigidity,

8.3 Stability of Thin Elastic Plate Under End Load . . . . . . . . . . . . 95

9.1 Large Seamount . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99

9.2 Seamount Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100

9.3 Seamount Flexure vs. Age . . . . . . . . . . . . . . . . . . . . . . . 101

9.4 Trench Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102

9.5 Trench Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

9.6 Outer Rise Normal Faults . . . . . . . . . . . . . . . . . . . . . . . . 104

9.7 Brittle and Ductile Deformation . . . . . . . . . . . . . . . . . . . . 105

9.8 Trench Flexure and Earthquakes . . . . . . . . . . . . . . . . . . . . 106

9.9 Continental Yield Strength and Earthquakes . . . . . . . . . . . . . . 107

9.10 Fracture Zone Model . . . . . . . . . . . . . . . . . . . . . . . . . . 108

9.11 Fracture Zone Data . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

9.12 Normal Fault Model . . . . . . . . . . . . . . . . . . . . . . . . . . 110

9.13 Normal Fault Data . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

9.14 Thrust Fault . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112

9.15 Continental Rift and Basin Evolution . . . . . . . . . . . . . . . . . . 113

9.16 Passive Margin Model . . . . . . . . . . . . . . . . . . . . . . . . . 114

9.17 Passive Margin Data . . . . . . . . . . . . . . . . . . . . . . . . . . 115

9.18 Venus Trench Data . . . . . . . . . . . . . . . . . . . . . . . . . . . 116