Modeling the Impact of Atmospheric Moisture Transport on Global Ice Volume

Modeling the Impact of Atmospheric Moisture Transport on Global Ice Volume
Modeling the Impact of Atmospheric Moisture Transport on
Global Ice Volume
by
Kerim Hestnes Nisancioglu
Submitted to the Department of Earth, Atmospheric, and Planetary Sciences
in partial fulfillment of the requirements for the degree of
Doctor of Philosophy
at the
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
June 2004
c Massachusetts Institute of Technology 2004. All rights reserved.
Author . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Department of Earth, Atmospheric, and Planetary Sciences
April 30th, 2004
Certified by . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Peter H. Stone
Professor of Climate Dynamics
Thesis Supervisor
Accepted by . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Maria Zuber
Department Head and E.A. Griswold Professor of Geophysics
Modeling the Impact of Atmospheric Moisture Transport on Global Ice
Volume
by
Kerim Hestnes Nisancioglu
Submitted to the Department of Earth, Atmospheric, and Planetary Sciences
on April 30th, 2004, in partial fulfillment of the
requirements for the degree of
Doctor of Philosophy
Abstract
Following Milankovitch’s original hypothesis most model studies of changes in global ice volume
on orbital time scales have focused on the impact of ablation on ice sheet mass balance. In most
cases, poleward moisture flux is fixed and accumulation of snow only depends on local temperature. In this study, a simple coupled atmosphere-ice process model is introduced. An improved
representation of the atmospheric hydrological cycle is included, and accumulation is related to
the meridional flux of moisture by large scale baroclinic eddies. The ice sheets in the Northern
Hemisphere respond to both precession and obliquity frequencies when the model is forced with
seasonal insolation. Obliquity variations are introduced by the impact of earth’s tilt on the meridional temperature gradient and the poleward flux of moisture, whereas precession governs surface
melting by regulating summer temperatures. The response of the ice sheet to obliquity and precession is comparable, and significantly smaller than what is observed in the oxygen isotope record
of the late Pliocene and early Pleistocene (2.7 - 0.8 Ma BP). This suggests that in order to successfully reproduce the strong 41 Ka periodicity observed in the record, other mechanisms must
be involved such as nonlinear self-sustained, or stochastic processes, or alternatively the obliquity
dominated signal originates from Antarctica. In Antarctica the seasonal cycle is damped due to the
large thermal mass of the southern ocean, and surface melt is insignificant. Both of these factors
reduce the influence of precession in regulating ice volume. Instead, the mass balance is dominated
by accumulation and calving, thereby enhancing the role of obliquity in controlling ice volume.
Thesis Supervisor: Peter H. Stone
Title: Professor of Climate Dynamics
2
Contents
1 Introduction
7
2 The Paleoclimate Record
12
2.1
Orbital Parameters and Insolation . . . . . . . . . . . . . . . . . . . . . . . . . .
14
2.2
Ice Volume and Oxygen Isotopes . . . . . . . . . . . . . . . . . . . . . . . . . . .
17
2.3
Milankovitch Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
21
2.3.1
Daily Insolation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
21
2.3.2
Seasonal Insolation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
25
2.4
The 41 Ka Glacial Cycles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
28
2.5
Insolation Gradients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
29
2.6
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
33
3 Coupled Atmosphere and Ice Process Model
3.1
35
Atmosphere . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
35
3.1.1
Insolation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
37
3.1.2
Albedo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
40
3.1.3
Atmospheric Eddy Heat Flux Parameterization . . . . . . . . . . . . . . .
44
3.1.4
Approximate Meridional Temperature Profile . . . . . . . . . . . . . . . .
45
3.2
Mixed Layer Ocean . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
47
3.3
Ice Sheet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
48
3.3.1
50
Ice Sheet Mass Balance . . . . . . . . . . . . . . . . . . . . . . . . . . .
3
3.4
3.3.2
Ablation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
51
3.3.3
Accumulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
54
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
55
4 Model Validation and Climate Sensitivity
4.1
56
Control Experiment with Modern Insolation . . . . . . . . . . . . . . . . . . . . .
57
4.1.1
Atmospheric Meridional Temperature Profile . . . . . . . . . . . . . . . .
57
4.1.2
Atmospheric Eddy Heat Flux . . . . . . . . . . . . . . . . . . . . . . . . .
59
4.2
Climate Sensitivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
63
4.3
Ice Sheet Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
65
4.3.1
Temperature Dependent Albedo . . . . . . . . . . . . . . . . . . . . . . .
67
4.3.2
Inclusion of an Active Ice Sheet . . . . . . . . . . . . . . . . . . . . . . .
70
4.3.3
Parameterized Sea Ice . . . . . . . . . . . . . . . . . . . . . . . . . . . .
72
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
76
4.4
5 Orbital Insolation and Ice Sheet Mass Balance
77
5.1
Obliquity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
77
5.2
Precession . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
80
5.3
Last Interglacial and Glacial Inception . . . . . . . . . . . . . . . . . . . . . . . .
85
5.4
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
87
6 Late Pliocene and Early Pleistocene Glacial Cycles
89
6.1
Origin of Northern Hemisphere Ice Sheets . . . . . . . . . . . . . . . . . . . . . .
90
6.2
Late Pliocene to Early Pleistocene Ice Volume . . . . . . . . . . . . . . . . . . . .
92
6.2.1
Temperature Dependent Accumulation . . . . . . . . . . . . . . . . . . .
96
6.2.2
Cold Climate with Large Ice Sheets . . . . . . . . . . . . . . . . . . . . .
97
6.3
Late Pliocene to Early Pleistocene Snow Cover . . . . . . . . . . . . . . . . . . .
99
6.4
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
4
7 Summary and Discussion
102
7.1
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
7.2
Sea Ice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
7.3
The Topographic Effect of an Ice Sheet . . . . . . . . . . . . . . . . . . . . . . . 109
7.4
The Role of Antarctic Ice Sheets . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
7.5
Further Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
A The 41 Ka World: Milankovitch’s Other Unsolved Mystery
117
B Reorganization of Miocene Deep Water Circulation in Response to the Shoaling of the
Central American Seaway
124
Bibliography
137
5
Acknowledgments
I would like to thank my thesis committee members Ed Boyle, Maureen Raymo, Peter Stone and
Eli Tziperman as well as my office mates Baylor Fox-Kemper, Veronique Bugnion, Peter Huybers
and Jeff Scott for stimulating discussions throughout the course of my studies at MIT. At the same
time, I am grateful to the many people I got to know at MIT and in Cambridge during the past
several years. In particular I would like to thank Mary Elliff, the BAD coop, the Green Monkeys,
the PAOC lunch gang, and the unofficial MIT telemark club. Without you, my stay at MIT would
have been no fun! Last, but not least, I am deeply in debt to Sussie who very patiently and lovingly
supported me and kept me going throughout the hardest parts of these past years.
This work was in part funded by the Paleoclimate Program of the National Science Foundation
with the award "Collaborative Research: Modeling the Role of Obliquity and Insolation Gradients
in Controlling Global Ice Volume".
6
Chapter 1
Introduction
The glacial cycles observed in proxy records of the past ∼ 2.8 Ma, represent some of the largest
and most significant changes in past climate. Explaining these changes would greatly advance
the understanding of the climate system and its future response to man-made forcing. However,
the physical mechanisms driving the cycles in ice volume are not well understood. After data
supporting the early work of Milankovitch (1941) was presented by Hays et al. (1976), the prevalent
theory has been that major fluctuations in global climate, associated with the ice age cycles, are
caused by variations in insolation at critical latitudes and seasons. In particular, ice sheet growth
and retreat is thought to be sensitive to high northern latitude summer insolation.
Variations in summer insolation at high latitudes are dominated by the precession of the equinoxes at periods of 19 Ka and 23 Ka, combined with a minor contribution due to changes in obliquity of the Earth’s axis at a period of 41 Ka. In contrast, changes in global ice volume and deep
sea temperatures are dominated by cycles of 41 Ka and ∼ 100 Ka, with a smaller contribution by
cycles at 19 Ka and 23 Ka. Although Hays et al. (1976) found a statistical association between the
ice volume record and high latitude summer insolation, a physical explanation for the cause of the
glacial cycles is still missing.
Numerous studies have tried to explain the large ∼ 100 Ka cycles dominating the records of
global ice volume during the past ∼ 0.8 Ma. However, relatively few have investigated the origin
7
of the smaller amplitude 41 Ka cycles, which have been found to dominate records older than ∼ 0.8
Ma (Pisias and Moore, 1981; Ruddiman et al., 1986). As for the ∼ 100 Ka ice volume cycles, the
41Ka cycles are left unexplained by the currently accepted Milankovitch theory. At the same time,
no existing climate model has been able to successfully reproduce the 41 Ka cycles in ice volume.
The main objective of this study is to investigate the possible physical mechanisms responsible
for the 41 Ka cycles observed in the ice volume record. A hypothesis is presented where variations
in the insolation gradient between high and low latitudes play an important role in controlling the
poleward flux of heat and moisture by the atmosphere. Under favorable orbital configurations, the
flux of moisture to high latitudes is enhanced, and the accumulation of snow on the continents
increases. Combined with a strong ice-albedo feedback this could lead to growth of ice sheets.
Climate models are valuable tools in the search for physical explanations of the questions posed
by the proxy record. However, existing coupled atmosphere-ocean General Circulation Models
(GCMs) require too much computing power to simulate climate on orbital time scales (> 10 Ka).
Instead, it has been common to use Energy Balance Models (EBMs) to study changes in climate
in response to variations in insolation. These types of models can be grouped into four categories:
1) annual mean atmospheric models; 2) seasonal atmospheric models with a mixed layer ocean; 3)
Northern Hemisphere ice sheet models; and 4) coupled climate-ice sheet models, which in some
cases include a representation of the deep ocean.
Examples of studies with the first type of models include the early work by Budyko (1969);
Sellers (1969); North (1975), who investigated the sensitivity of climate to changes in annual
global mean insolation. However, changes in the Earth’s orbital parameters result in a redistribution of insolation with latitude and time of the year, with a negligible impact on annual global
mean insolation. Therefore, annual mean models are not adequate when investigating the impact
of orbital insolation on climate, as they cannot capture the parts of the insolation variations which
are seasonal and translate them into long term climate change.
The second type of models include a representation of the seasonal cycle, and have been used
to investigate the orbital theory of Milankovitch (e.g. Schneider and Thompson (1979); Suarez and
8
Held (1979)). In this case, the seasonal variations in orbital insolation are resolved. However, as
for the first type of models, past changes in ice cover are assumed to follow the simulated variations
in the extent of perennial snow. This approach assumes that ice cover and the powerful ice-albedo
feedback are governed only by temperature, as the extent of snow in these models is fixed to the
zero degree isotherm. In reality the growth and decay of land based ice sheets is governed by the
balance of accumulation and ablation. Therefore, when investigating changes in ice cover, it is
necessary to include an appropriate representation of the dynamics and mass balance of ice sheets
in the model.
The third type of models improve upon this by focusing on modeling past changes in mass
balance and size of typical Northern Hemisphere ice sheets, such as the Laurentide. This type of
studies was initiated by Weertman (1964, 1976) and later improved by Oerlemans (1980); Birchfield et al. (1981); Hyde and Peltier (1987); Clark and Pollard (1998), who used simple ice sheet
models, forced by a prescribed distribution of accumulation minus ablation, to predict ice thickness
versus latitude. These models do not calculate the atmospheric energy balance in order to estimate
snowfall and surface melt, instead changes to the prescribed distribution of net accumulation follow
variations in mean summer insolation.
The fourth type of models include zonal mean seasonal climate models coupled to the simple
Weertman (1964) type ice sheet model (e.g. Pollard (1978, 1982)), as well as intermediate complexity climate models coupled to a dynamic ice sheet (e.g. Gallee et al. (1991); Deblonde et al.
(1992); Berger et al. (1999)). These models give a more realistic representation of the climate as
compared with the simpler models. However, it should be noted that none of the models described
here take into account the influence of changes in meridional atmospheric moisture flux on accumulation of snow on the ice sheet. Even in the more sophisticated climate models it has been
common to parameterize accumulation by relating it to local temperature, thereby neglecting any
impact by changes in atmospheric moisture transport.
Partly due to the lack of good data on variations in global ice volume older than about halfmillion years, most model studies have focused on understanding the more recent records domin-
9
ated by the ∼ 100 Ka glacial cycles (see reviews in Imbrie et al. (1993) and Paillard (2001)). All
of these models respond with periods close to the precession and obliquity periods of the insolation forcing. However, the amplitude of the response is in most cases significantly smaller than
what is observed in the proxy records. At the same time, the dominant ∼ 100 Ka cycles of the
ice volume record, characterized by rapid deglaciations, are only found when including a time lag
in the response of the model. Such an internal time lag can be produced by taking into account
bedrock depression under the load of the ice (e.g. Oerlemans (1980); Birchfield et al. (1981); Hyde
and Peltier (1987)), or by adding a parameterization of ice calving into proglacial lakes or marine incursions at the margin of the ice sheet (Pollard, 1982). Alternatively, the ∼ 100 Ka cycles
have been explained as free, self-sustained oscillations (e.g. Letreut and Ghil (1983); Saltzman
and Sutera (1984); Ghil (1994); Gildor and Tziperman (2001)), which might be phase-locked to
oscillations in orbital insolation (e.g. Saltzman et al. (1984); Gildor and Tziperman (2000)).
One of the very few model studies which has investigated variations in ice volume before the
∼ 0.8 Ma transition, is the study by Berger et al. (1999). In this study, a zonal mean atmospheric
model coupled to an ocean mixed layer and continental ice sheets (Gallee et al., 1991) is forced
with seasonal insolation for the past 3 Ma. Under conditions of increased atmospheric CO2 , the
simulated ice volume responds to the main periods of obliquity and precession in the interval 2 − 1
Ma, whereas no significant periodicity is observed in the interval 3 − 2 Ma BP. There is no clear
dominance of obliquity as observed in the proxy records ∼ 3 − 0.8 Ma BP, and during most of
the simulation there is a relatively strong eccentricity signal with a period of 400 Ka, which is not
observed in the record. At the same time, during the early part of the simulation (3.0 − 1.0 Ma BP),
ice-free conditions dominate, inter-spaced by brief intervals of ice advance paced by these longer
period oscillations of eccentricity.
As is the case for all the models discussed above, the possible influence of changes in atmospheric moisture flux has not been taken into account, possibly neglecting an important feedback
between changes in insolation and ice volume on orbital time scales. With this concern in mind,
a seasonal atmosphere-ice process model is formulated in chapter 3 which is as simple as pos-
10
sible, while including a dependence of accumulation on atmospheric moisture flux, as well as an
improved representation of the ice-albedo feedback. Keeping the model simple, ensures that the
physical processes and feedbacks involved can be understood, and that multiple experiments forced
with insolation reconstructions covering periods of up to 1 Ma can be carried out in a reasonable
amount of time.
The sensitivity of the model climate to changes in insolation is investigated in chapters 4 and 5.
The perturbations applied will be of three different characters: 1) variations in solar constant, which
applies a uniform change in insolation at all latitudes and seasons similar to the effect of changing
atmospheric CO2 levels; 2) variations in earth’s obliquity, which principally changes the relative
amount of insolation received at high and low latitudes, and 3) variations in orbital precession,
which redistributes the amount of insolation received during each season, without affecting the
latitudinal insolation distribution.
In chapter 6, the atmosphere-ice model is used to investigate the physics governing the oscillations in ice volume observed in the proxy record, with emphasis on the glacial cycles of the Late
Pliocene and Early Pleistocene (∼ 3.0 − 0.8 Ma BP). Finally, a summary of the results is given in
chapter 7. Before introducing the model, the following chapter (chapter 2) gives an introduction to
the paleoclimate record as well as a discussion of the present understanding of the original orbital
theory due to Milankovitch (1941).
11
Chapter 2
The Paleoclimate Record
Over the past years a tremendous amount of data on past climate has been collected from deep sea
sediment cores, ice cores, and terrestrial data. However, at present, several of the most fundamental
questions posed by this data remain unanswered. A few examples are as follows: it is not known
what triggered the onset of enhanced glaciations ∼ 2.8 − 2.6 Ma BP; why the initial oscillations in
ice volume were dominated by a period of 41 Ka; or why larger oscillations in ice volume with a
duration of approximately 100 Ka dominated after ∼ 0.8 Ma BP.
Following the original work of Milankovitch (1941) most scientists would agree that on orbital
time scales, variations in ice volume are related to oscillations in high latitude insolation. However,
most work in the field has focused on the last 0.8 Ma of the record. In this chapter, it will be
argued that when considering the full ice volume record of the past 3 Ma, current understanding of
the relation between insolation changes and glacial cycles is not adequate. Instead, an alternative
hypothesis is tested which relates changes in ice volume to the impact of obliquity on the meridional
flux of moisture to high latitudes. Part of the work presented here has been published by Raymo
and Nisancioglu (2003), and the reader is referred to appendix A for further reading.
12
TODAY
Spring Equinox
Summer Solstice
Sun
ω
Perihelion
Aphelion
Winter Solstice
Earth
Fall Equinox
11 Ka BP
Fall Equinox
Winter Solstice
Sun
ω
Perihelion
Aphelion
Summer Solstice
Earth
Spring Equinox
Figure 2-1: Sketch of the Earth’s orbit around the sun today (0 Ka BP) and at the last glacial
termination (11 Ka BP), showing the positions of the solstices and equinoxes relative to perihelion.
The longitude of perihelion (ω) is measured as the angle between the line to the Earth from the Sun
at spring equinox and the line to the Earth at perihelion.
13
0
0
−80
−60
−4
−20
20
30
20
0
0 0
20
0
−60
−90
0
−20
0
−30
0
20
60
−2
0
Latitude (degrees)
60
0
−4
−2
0
90
0
−4
120
180
240
Mean Orbital Longitude (degrees)
−80
−60
−4
0
300
360
Figure 2-2: Insolation difference in units of W/m2 as a function of latitude and season for a decrease in obliquity from 25◦ to 21◦ in the case of a perfectly circular orbit (e = 0). The start of the
year is fixed at the spring equinox and the year is divided into 360 degrees, where 90◦ is summer
solstice. A detailed description of the calendar adopted in this study is given in section 3.1.1 on
page 37.
2.1 Orbital Parameters and Insolation
The Earth’s orbit around the sun is an ellipse where the degree to which the orbit departs from
a circle is measured by its eccentricity (e). The point on the orbit closest to the sun is called
the perihelion, and the point most distant from the sun the aphelion (figure 2-1). If the distance
from the Earth to the Sun is rp at perihelion, and ra at aphelion, then the eccentricity is defined as
e = (ra − rp )/(ra + rp ). Variations in the eccentricity of the Earth’s orbit follow cycles of 100 Ka
and 400 Ka, giving a change in annual mean insolation on the order of 0.2%, or less. This change
in insolation is believed to be too small to produce any significant effect on climate.
A more significant change in insolation is caused by variations in the seasonal and latitudinal
distribution of insolation due to obliquity. Obliquity (ε) is the angle between Earth’s axis of rotation and the normal to the Earth’s plane around the sun (figure 2-1). This angle is 23.5◦ today, but
varies between values of 22.1◦ and 24.5◦ with a period of 41 Ka. A decrease in obliquity decreases
the seasonal insolation contrast, with the largest impact at high latitudes. At the same time, annual
mean insolation at high latitudes is decreased compared to low latitudes. An example of the effect
14
of obliquity variations on seasonal insolation is shown in figure 2-2. During times when obliquity
is small, high latitude summertime insolation decreases, whereas mid latitude wintertime insolation increases. The magnitude of the change in high latitude summer insolation due to obliquity
variations can be as large as 10%.
The third and last variable affecting insolation is the longitude of perihelion (ω). This parameter
is defined as the angle between the line to the Earth from the sun at spring equinox and the line
to the Earth at perihelion (figure 2-1). It determines the direction of Earth’s rotational axis relative
to the orientation of Earth’s orbit around the sun, thereby giving the position of the seasons on the
orbit relative to perihelion. Changes in the longitude of perihelion result in the Earth being closest
to the sun at different times of the year. Today, the Earth is closest to the sun on the 4th of January,
or very near winter solstice in the Northern Hemisphere. All other things being equal, this will
result in relatively warm winter and cool summer seasons in the Northern Hemisphere, whereas the
opposite is the case in the Southern Hemisphere. At the time of the last deglaciation, 11 Ka BP, the
Earth was closest to the sun at summer solstice, resulting in extra warm summers and cool winters
in the Northern Hemisphere.
If the Earth’s orbit was a circle, the distance to the sun would remain constant at all times of
the year and it would not make any difference where on the orbit the seasons were positioned.
Therefore, the impact of variations in the longitude of perihelion depends on the eccentricity of
the Earth’s orbit and is described by the precession parameter (esinω). The combined effect of
eccentricity and longitude of perihelion can give changes in high latitude summer insolation on the
order of 15% and varies with periods of 19 Ka, and 23 Ka, but is modulated by the longer period
variations of eccentricity. Figure 2-3 shows the variations in obliquity (ε), longitude of perihelion
(ω), eccentricity (e), and the precession parameter (esinω) for 1800 − 1200 Ka BP as computed by
Laskar et al. (1993). Note that the two main periods of variability of precession are not present in
the oscillation of longitude of perihelion (ω), but rather are produced by the splitting of the basic
precession period of about 22 Ka, due to its modulation by eccentricity.
The main features of variations in insolation on orbital time scales are summarized as follows:
15
a)
ε (degrees)
24.0
23.0
22.0
b)
180
ω (degrees)
120
60
0
−60
−120
−180
0.06
c)
e
0.04
0.02
0
0.06
d)
e×sin(ω)
0.03
0
−0.03
−0.06
−1800
−1700
−1600
−1500
Time (Ka)
−1400
−1300
−1200
Figure 2-3: Variations in a) obliquity of the Earth’s axis (ε), b) longitude of perihelion relative to
the spring equinox (ω), c) Earth’s orbital eccentricity (e), and d) precession parameter (esinω) for
1800 − 1200 Ka BP (Laskar et al., 1993).
16
1. Annual mean insolation averaged over the globe, or over one hemisphere varies only as a
result of eccentricity (e). The observed eccentricity range (e < 0.06), produces a change in
insolation of less than ∼ 0.2%.
2. Annual mean insolation at a specific latitude is controlled by obliquity (ε) and eccentricity
(e), and is independent of the longitude of perihelion (ω). The observed range of obliquity
(22.0◦ − 24.5◦ ) can produce ∼ 10% variations in insolation at high latitudes.
3. Seasonal mean insolation at a specific latitude varies with obliquity (ε) as well as the precession parameter (esinω) due to the influence of the latter on the length of the season. The
combined effects of these parameters can cause variations in seasonal insolation as large as
∼ 30% at high latitudes.
4. Daily mean insolation at a specific latitude depends on all the orbital parameters (e, ω, ε).
2.2 Ice Volume and Oxygen Isotopes
Some of the longest continuous records of past climate come from deep sea sediment cores. Ocean
sediments are laid down over time, and by drilling into the sea floor, layered sediment cores can be
extracted containing valuable information about the conditions at the time when the layers where
formed. The time resolution of the core depends on the sedimentation rate at the drill site, as well as
the degree to which the top layer of sediments have been disturbed by bottom dwelling organisms
(bioturbation).
By studying the relative abundance of oxygen isotopes in shells of tiny marine organisms (foraminifera) found in the sediments, it is possible to estimate the amount of water tied up in continental ice sheets and glaciers. This is because water molecules containing the lighter isotope of
oxygen (16 O) are more readily evaporated and transported from the oceans to be deposited as ice
on land (Dansgaard, 1954, 1964). Therefore, during glacial periods ocean water is enriched with
the heavy oxygen isotope (18 O). Deviations of the oxygen isotope ratio relative to a standard is
17
defined as
δ 18 O =
(18 O/16O)sample − (18 O/16 O)SM OW
× 103
(18 O/16 O)SM OW
where SMOW signifies Standard Mean Ocean Water. The δ 18 O value of surrounding sea water is
recorded by foraminifera who build their shells from calcium carbonate (CaCO3 ). However, the
fractionation of the oxygen isotopes when forming CaCO3 depends on water temperature (Urey,
1947; Emiliani, 1955): low water temperatures give high δ 18 O values. To further add to the complexity, the δ 18 O value of sea water is not uniform throughout the ocean due to the pattern of evaporation and precipitation. This problem can be partly avoided by studying benthic foraminifera
living on the sea floor, rather than surface dwelling planktonic foraminifera. At depth, the water is
far more homogeneous than on the surface. At the same time, water at the ocean floor is very cold,
and would not have been much colder during glacial periods, thereby reducing the contribution of
temperature variations to the δ 18 O value.
The first continuous long benthic δ 18 O ice volume records were extracted by Shackleton and
Opdyke (1973) and Hays et al. (1976). These cores extend back to the Brunhes-Matuyama magnetic reversal event (780 Ka BP) making it possible to construct a time scale by assuming linear
accumulation rates (e.g. Shackleton and Opdyke, 1973). Analysis of the data show cycles in ice
volume with periods of about 20 Ka, 40 Ka and 100 Ka; in agreement with the predictions of Milankovitch. Further studies with cores extending past the Brunhes-Matuyama reversal by Pisias and
Moore (1981) and Ruddiman et al. (1986) show that the late Pliocene (3.6 − 1.8 Ma BP) and early
Pleistocene records (1.8 − 0.8 Ma BP) are dominated by smaller amplitude cycles with a period of
41Ka, rather than the large 100 Ka cycles of the late Pleistocene.
Many records generated since this time have confirmed these early observations, namely:
1. the main period of ice volume change from 3.0 to 0.8 Ma, is 41 Ka, which is the dominant
period of orbital obliquity.
2. after ∼ 0.8 Ma, ice sheets vary predominantly with a period of 100 Ka and the amplitude of
oscillations in δ 18 O increases, implying growth of larger ice sheets.
18
100 kyr world
δ18O (‰)
2.0
41 kyr world
3.0
4.0
B M
0
500
J
TOld
1000
1500
M G
2000
2500
3000
Age (kyr)
Figure 2-4: Benthic δ 18 O record from DSDP site 607 (Ruddiman et al., 1989; Raymo et al., 1989)
in the North Atlantic (solid line) plotted to a paleomagnetic time scale. The magnetic field reversals
are marked, as well as the transition from a dominant 41 Ka to a 100 Ka periodicity in ice volume.
B (Brunhes), M (Matuyama), J (Jaramillo), TOld (top of Olduvai), and G (Gauss).
Depth (mcd)
0.0
31.84
40.345
43.965
73.655
111.58
129.50
Age (Ka)
0
780
984
1049
1757
2600
3054
Magnetic Event
top of core
Brunhes/Matuyama
Jaramillo top
Jaramillo bottom
Olduvai top
Matuyama/Gauss
Kaena top
Table 2.1: Age control points used for paleomagnetic time scale at DSDP site 607.
19
41
1
10
23 19
Power Spectrum
Raw data
99% sign.
0
10
−1
10
−2
10
0
0.02
0.04 0.06 0.08
Frequency (1/Ka)
0.1
Figure 2-5: Power spectrum of benthic δ 18 O from site 607 for the time period 2.75 − 0.8 Ma BP
using the same paleomagnetic time scale as in figure 2-4. The spectrum was calculated using the
Multi Taper Method (MTM) (Thomson, 1982; Percival and Walden, 1993; Ghil et al., 2002). The
red dashed line represents the 99% significance level of the data relative to estimated background
noise.
The benthic δ 18 O record from DSDP 607 illustrates both these points (figure 2-4). Note that the
isotope record is plotted with a paleomagnetic time scale (table 2.1) determined by the depth of
magnetic field reversals recorded by ferromagnetic grains in the sediment core (Clement and Kent,
1987). Constant sedimentation rates are assumed between these magnetic reversal events which are
dated by interpolating sea floor magnetic anomalies between fixed calibration points (Cande and
Kent, 1992, 1995). The two calibration points used (B/M and M/G) are independently derived by
both radiometric and astronomic tuning techniques (Berggren et al., 1995).
Using this simple time scale, which is not biased by orbital tuning, one can clearly observe
the 41 Ka periodicity of the late Pliocene and Early Pleistocene (3.0 − 0.8 Ma). The obliquity
periodicity is confirmed by plotting the power spectrum of the record (figure 2-5). Note the near
complete lack of variance at the 19 Ka and 23 Ka precession, and 100 Ka eccentricity periods. The
gradual increase in the strength of the 100 Ka cycle can be seen when filtering the δ 18 O record
at the main orbital frequencies (figure 2-6). In this case the tuned time scale of Shackleton et al.
(1990) is used. Note that the variance at the 41 Ka period remains relatively unchanged, whereas
20
the influence of the precession periods increases toward the younger part of the record.
Because site 607 is located in the subpolar North Atlantic (41◦ N, 23◦ W ), it also contains a
record of ice-rafted detritus (IRD) delivered to the open ocean over the Plio-Pleistocene (figure 2-7).
Over the entire length of the record, the input of IRD co-varies with the δ 18 O record (Raymo et al.,
1989; Ruddiman et al., 1989). The sediment core data thus proves that the variability observed
in benthic δ 18 O must derive in part from the growth and decay of ice sheets bounding the North
Atlantic.
2.3 Milankovitch Theory
Based mainly on climate proxy records covering the last 0.5 Ma a general scientific consensus has
emerged that variations in summer insolation at high northern latitudes are the dominant influence
on climate over tens of thousands of years. This theory is associated with Milankovitch (1936,
1941) and Köppen and Wegener (1924), who based their work on earlier studies by Croll (1875)
and others. The basic idea is that at times of reduced summer insolation, snow and ice can persist
at high latitudes through the summer melt season. At the same time, the cool summer seasons
are proposed to be accompanied by mild winter seasons which could lead to enhanced winter
accumulation of snow. When combined, reduced melting and a slight increase in accumulation,
enhanced by a positive snow albedo feedback, could eventually lead to full glacial conditions. In
relation to the orbital parameters, a minimum in summer insolation at high latitudes is achieved
when obliquity is small, eccentricity is large, and northern summer solstice occurs when the earth
is farthest from the sun (aphelion).
2.3.1 Daily Insolation
In the past, countless research papers have plotted, or even tuned, climate records to June or July
65◦ N insolation, following on Milankovitch’s original idea. Using many of these records, Imbrie
et al. (1992) shows that climate variance at precession and obliquity frequencies appears to be
21
18
Anomaly (permil)
0.4
a) δ O filtered at 41 Ka period
0.2
0
−0.2
−0.4
Anomaly (permil)
0.4
b) δ18O filtered at 19 & 23 Ka periods
0.2
0
−0.2
−0.4
Anomaly (permil)
0.4
18
c) δ O filtered at 100Ka period
0.2
0
−0.2
−0.4
−2500
−2000
−1500
−1000
Time (Ka BP)
−500
0
Figure 2-6: Benthic δ 18 O record from site DSDP 607 filtered at the main orbital periods a) 41 Ka,
b) 19 & 23 Ka, and c) 100 Ka. The data is plotted using the tuned time scale of Shackleton et al.
(1990).
22
60
%CaCO
3
70
80
90
100
−1800
−1700
−1600
−1500
Time (Ka BP)
−1400
−1300
−1200
Figure 2-7: %CaCO3 at site 607 plotted to the same time scale as the δ 18 O data in figure 2-4
(Raymo et al., 1989; Ruddiman et al., 1989). Decreases in %CaCO3 are caused by increases in
lithic fragments (ice rafted detritus) within the sediment.
600
a) Summer Solstice at 65N
W/m
2
550
500
450
400
550
b) Summer Solstice at 25N
W/m
2
500
450
400
−1800
−1700
−1600
−1500
Time (Ka)
−1400
−1300
−1200
Figure 2-8: Summer solstice insolation (ω = 90◦ ) from 1.8 to 1.2 Ma BP at a) 65◦ N, and b) 25◦ N
(Laskar et al., 1993). Today summer solstice occurs on June 21st .
23
41 23 19
6
10
Raw data
99% sign.
Power Spectrum
4
10
2
10
0
10
−2
10
0
0.05
0.1
0.15
Frequency (1/Ka)
0.2
Figure 2-9: Power spectrum of summer solstice insolation at 65◦ N covering the time period from
1.8 to 1.2 Ma BP. The spectrum was calculated using the Multi Taper Method (MTM) (Thomson,
1982; Percival and Walden, 1993; Ghil et al., 2002). The red dashed line represents the 99%
significance level of the data relative to estimated background noise. Vertical dotted lines mark the
main periods of obliquity (41 Ka), and precession (19 Ka, 23 Ka).
linearly forced by and is coherent with northern summer insolation. Only the 100 Ka cycle is
left unexplained by this model and is typically ascribed to non-linear variability arising internally
within the climate system.
In the late Pliocene and early Pleistocene, no significant variance at the 100 Ka period is observed in benthic δ 18 O or IRD records (figures 2-4 and 2-7). Therefore, it may be expected that
global ice volume should vary linearly and coherently with northern high latitude summer insolation, as implied by the standard Milankovitch model. A comparison of the δ 18 O ice volume record
with the standard insolation record (figures 2-8 & 2-9) shows that although the ice volume proxies
have a strong 41 Ka periodicity, summer insolation is dominated by the 19 Ka and 23 Ka periods of precession. As insolation time series at a given time of the year (e.g. June or July) are in
phase across all latitudes of the same hemisphere (Berger et al., 1993) the proxy records could be
compared equally well with insolation from other latitudes than the typical choice of 65◦ N.
Any direct response of climate at high latitudes to summer insolation would require a strong
presence of precession in the geologic record. However, this frequency is barely discernible in
24
the late Pliocene and early Pleistocene ice volume record. It is therefore questionable whether
summer insolation at high northern latitudes exerts a dominant influence on climate over most of
the Northern Hemisphere ice ages.
2.3.2 Seasonal Insolation
An alternative to June or July insolation as the driver of high latitude climate, is seasonal, or annual
mean insolation. It turns out that the influence of obliquity increases when averaging over parts
of the annual cycle; e.g. for the case of annual mean insolation the effects of precession cancel,
leaving only variations due to obliquity. The astronomical summer season, or summer half year,
is defined as the section of the Earth’s orbit between spring and fall equinox (0◦ < λ < 180◦ , see
figure 2-1). Similarly the astronomical winter season is defined as 180◦ < λ < 360◦ , where λ is the
position of the Earth in its orbit relative to spring equinox. Note that the summer season includes
spring, and winter season includes fall.
As was observed by Milankovitch (1936, p. A40), the total insolation received during the
astronomical seasons is independent of precession and varies at all latitudes with obliquity. There
is also a phase reversal of the response at about 44◦ N for annual insolation and 11◦ N for seasonal
insolation. The influence of precession is only included when taking into account the duration of the
seasons. Variations in the duration of the seasons can amount to as much as 33 days (Milankovitch,
1936; Berger, 1978a). Therefore, when calculating the time mean insolation for the astronomical
seasons, the result depends on both obliquity and precession (figure 2-10).
Because the relative duration of the astronomical seasons changes with time, Milankovitch
(1923, 1941) argued for using a different measure of seasonal insolation. Instead of using the
astronomical half years he imposed two restrictions when determining seasonal insolation: 1) the
year should be divided into two time periods of equal duration, and 2) each day of the summer
season should receive more insolation than any day of the winter season. The seasons following
these requirements were defined as the caloric summer and caloric winter half years. Although the
calculation is slightly more time consuming, the half years are of equal duration and the year to
25
420
a) Astronomical Summer Half Year at 65N
W/m
2
400
380
360
340
70
b) Astronomical Winter Half Year at 65N
W/m2
65
60
55
50
218
c) Annual Mean at 65N
W/m
2
216
214
212
210
−1800
−1700
−1600
−1500
Time (Ka)
−1400
−1300
−1200
Figure 2-10: a) Astronomical summer half year (0◦ < λ < 180◦ ), b) winter half year (180◦ < λ <
360◦), and c) annual time mean insolation (J/(m2 s) = W/m2 ) at 65◦ N from 1.8 to 1200 Ka BP
(Laskar et al., 1993).
26
6
a) Caloric Summer
80N
60N
20N
(J/m2)×106
4
2
0
−2
−4
4
b) Caloric Winter
80N
60N
20N
(J/m2)×106
2
0
−2
−4
−6
−1800
−1700
−1600
−1500
Time (Ka BP)
−1400
−1300
−1200
Figure 2-11: Caloric summer and winter half year insolation calculated using the orbital solution
due to Berger (1978b), see also Vernekar (1972) Caloric half years are periods of equal duration
where each day of the summer half year receives more insolation than any day of the winter half
year (Milankovitch, 1923, 1941, p. 286).
year changes in caloric insolation can be compared.
Figure 2-11 shows cumulative caloric summer and winter half year insolation for the period 1.8
to 1.2 Ma BP at three different latitudes. For the caloric summer half year obliquity dominates at
high latitudes (> 65◦ N), whereas climatic precession (esinω) dominates at low latitudes (< 55◦ N),
as was pointed out by Milankovitch (1936, p. A157). In the mid latitudes (∼ 55 − 65◦ N) the
contribution by obliquity and climatic precession are of similar magnitude. The figure only shows
the insolation in the Northern Hemisphere. In the Southern Hemisphere, variations in caloric half
year insolation due to obliquity are in phase with the Northern Hemisphere and could potentially
amplify the global signal, whereas the variations due to climatic precession are out of phase.
Based mainly on geologic data from the last 0.5 Ma BP, and without knowledge of the 41 Ka
cycles of the early Pliocene and late Pleistocene, Berger (1976, 1978a,b) reasoned that climatic
27
precession plays the leading role in past climates. Therefore, it was assumed that the seasonal contrast and the equinoctial, or monthly insolation are the most important parameters to be considered
when modeling glacial-interglacial fluctuations. Following this lead most researchers replaced the
caloric half year insolation as a driver of climate by mid-month, or monthly mean insolation for
e.g. June or July at 65◦ N (figure 2-8).
As was discussed earlier; mid-month, or monthly mean insolation is not a good candidate for
explaining early Pliocene and late Pleistocene glacial cycles which were dominated by 41 Ka periods. The original caloric summer half year insolation of Milankovitch has a stronger obliquity
signal and therefore would be a better candidate for explaining the 41 Ka cycles. Several components of the climate system, such as the oceans and ice sheets, have a long response time and
integrate any external forcing, such as insolation, over time. However, it should be emphasized that
choosing a preferred time period, or latitude for insolation when forcing a climate model will bias
the results and should be avoided. Eventually, experiments with physical models should make it
possible to understand which part of the insolation cycle is most important in driving the 41 Ka
cycles in ice volume.
In the next section an alternative hypothesis is presented to explain the 41 Ka glacial cycles
which does not involve choosing a preferred time period, or latitude for the insolation forcing.
2.4 The 41 Ka Glacial Cycles
While numerous studies have attempted to model the 100 Ka cycles of the late Pleistocene, few have
focused on understanding the 41 Ka cycles. A notable exception is Andre Berger and colleagues
who used the Louvain-la-Neuve intermediate complexity two dimensional climate model (LLN2D) to simulate growth and decay of ice sheets over the past 3 Ma (Berger et al., 1999). While
the obliquity period is present in the modeled ice volume, the precession periods are also strongly
present, and dominate at times. In other words, although the study successfully reproduced the lack
of 100 Ka cycles in ice volume during the early part of the record, it was not able to model an ice
28
sheet that varies only at the obliquity period. This appears to be because the model is very sensitive
to northern high latitude summer insolation, and variations in precession.
It is important to note that the majority of intermediate complexity climate models used for
paleoclimate experiments, such as the LLN model, do not include an adequate representation of
the hydrological cycle. Instead it is common to perturb a modern observed precipitation field
according to changes in temperature. As a result; changes in atmospheric moisture flux have no
direct impact on high latitude precipitation and the accumulation of ice sheets. In the present study
it is hypothesized that atmospheric moisture flux is important in controlling growth and decay
of ice sheets through its influence on accumulation, and will be explicitly included in the model
configuration.
There have been very few alternative theories to explain the 41 Ka cycles in ice volume. Kukla
(1968, in a reply to Emiliani, 1968) proposed that northern latitude winter insolation may drive late
Pliocene/early Pleistocene climate cycles. However, the total insolation received in January is a
factor of 20 less than summer insolation at the same latitude. At the same time, latitudes north of
the Arctic circle are in darkness for several months in the winter. Therefore, high latitude winter
insolation is thought to be of less importance in controlling ice volume.
2.5 Insolation Gradients
Given that local daily, and monthly mean summer insolation has too much precession and winter
insolation appears too weak to explain the 41 Ka cycles in ice volume, an alternative explanation
has to be found. Following Raymo and Nisancioglu (2003, appendix A) it is hypothesized that the
gradient in insolation between high and low latitudes may play an important role in controlling
climate from ∼ 3.0 to 0.8 Ma years ago, through its influence on the poleward flux of moisture.
Figure 2-12 shows the gradient in astronomical summer half-year insolation between 25◦ N and
65◦ N, defined as the time mean insolation received between the spring (λ = 0◦ ) and fall (λ = 180◦)
equinoxes. When calculating the spectra of the insolation gradient it can be seen that it is dominated
29
95
a) Astronomical Summer Half Year Gradient 25N − 65N
W/m
2
90
85
80
75
70
−1800
−1700
−1600
−1500
−1400
−1300
−1200
Figure 2-12: Gradient in astronomical summer half year insolation between 25◦ N and 65◦ N (Laskar et al., 1993).
41 23 19
4
10
Power Spectrum
Raw data
99% sign.
2
10
0
10
−2
10
0
0.05
0.1
0.15
Frequency (1/Ka)
0.2
Figure 2-13: Power spectrum of the gradient in astronomical summer half year insolation between
25◦ N and 65◦ N. The spectrum was calculated using the Multi Taper Method (MTM) (Thomson,
1982; Percival and Walden, 1993; Ghil et al., 2002). The red dashed line represents the 99%
significance level of the data relative to estimated background noise. Vertical dotted lines mark the
main periods of obliquity (41 Ka), and precession (19 Ka, 23 Ka).
30
5
W/m2
90
4
70
−1800
18
80
δ O (per mil)
100
−1700
−1600
−1500
Time (Ka BP)
−1400
−1300
3
−1200
Figure 2-14: Benthic δ 18 O record from site 607 (red solid line) plotted together with astronomical
summer half year insolation gradient between 25◦ N and 65◦ N (black dashed line). The δ 18 O data
is plotted to the same paleomagnetic time scale as in figure 2-4.
by obliquity (figure 2-13). It is this insolation gradient that drives the poleward fluxes of heat and
moisture in the atmosphere. The correlation between the insolation gradient and δ 18 O ice volume
record (figure 2-14) suggests that increased gradients promote ice sheet growth. Note that the δ 18 O
record in figure 2-14 appears to lag the insolation gradient by approximately 3 Ka, which is slightly
shorter than the response to obliquity observed in records from the late Pleistocene (e.g. Imbrie
et al., 1992). The true lag of response after forcing would be almost impossible to determine in
sediments of this age.
Although no one has invoked the influence of insolation gradients as an explanation for the 41
Ka glacial cycles before Raymo and Nisancioglu (2003, appendix A), the notion that insolation
gradients could influence climate on orbital time scales has been proposed in the past: Kutzbach
et al. (1968) argued that the magnitude of orbitally driven changes in insolation gradient is great
enough to have a significant impact on climate; and Young and Bradley (1984) suggested that hemispheric insolation gradients may have contributed to the growth and decay of continental ice sheets
in the past 150 Ka through their modulating influence on the poleward transport of moisture. They,
and previously Berger (1976), suggest that times of rapid ice growth and decay correspond to especially pronounced deviations in latitudinal insolation gradients. Johnson (1991) similarly invokes
a decrease in insolation gradient, rather than a change in direct summer insolation at high latitudes,
as the immediate cause of deglaciation at Termination 2 (127 Ka BP), offering this mechanism as
31
the explanation for paleoclimate data which suggest that deglaciation occurred prior to the increase
in summer insolation. This mismatch in timing between the deglaciation at Termination 2 and the
timing predicted by Milankovitch theory has also been discussed by Winograd et al. (1992) and
more recently by Henderson and Slowey (2000); Gallup et al. (2002), although their findings are
contested by the early warming observed in the SSTs records of Herbert et al. (2001).
It is possible that in the effort to correlate proxy records to high latitude summer insolation,
the influence of meridional fluxes of sensible and latent heat, driven by hemispheric temperature
gradients, has been underestimated. The mass balance of an ice sheet is not governed by the rate
of ablation alone, but by the relative rates of accumulation and ablation. At the same time, the rate
of ablation is not only controlled by local incoming solar radiation, but also by local atmospheric
temperature which is partially set by the strength of atmospheric and oceanic heat fluxes. The rate
of accumulation is controlled by the amount of moisture supplied to the ice sheet and the local
temperature. As atmospheric meridional heat and moisture fluxes strongly influence temperature
and precipitation at high latitudes, it is expected that they will exert a strong influence on ice sheet
mass balance.
It is plausible that long term variations in meridional heat and moisture fluxes, driven by orbital obliquity variations, dominate over the effect of local insolation, driven largely by precession,
and thus imprint the 41 Ka signal on the climate record. The powerful ice-albedo feedback would
enhance the effect of the insolation gradient on the meridional fluxes. As the atmospheric temperature cools at the onset of a glacial period, snow and ice expands into regions previously covered
by surfaces such as forests with relatively low albedo. This increased snow and ice cover raises
high latitude surface albedo significantly, reflecting incoming radiation, and causing a further decrease in temperature. In effect, ice-albedo is a very strong positive feedback mechanism, which
strengthens the meridional temperature gradient, further enhancing the poleward flux of moisture
which feeds the ice sheet.
The link between the insolation gradient and meridional atmospheric moisture flux is illustrated
in a coupled atmosphere-ocean GCM study by Khodri et al. (2001). In this case, a simulation of
32
climate at the last glacial inception (115 Ka BP) is compared with a modern control experiment.
At glacial inception, obliquity was about 1◦ lower than today, and the model shows warming at
low latitudes together with cooling at high latitudes, in agreement with estimates of sea surface
temperatures from sediment core data (Cortijo et al., 1999). This increased equator to pole temperature gradient is amplified by the response of the ocean, and causes a significant increase in mean
annual poleward atmospheric heat transport, including a 6% increase in latent heat transport. The
model does not include a representation of ice sheets. However, the combination of relatively cold
high latitudes and enhanced moisture transport could be beneficial for the growth of ice sheets,
supporting the hypothesis presented here.
2.6 Summary
According to the benthic oxygen isotope record, oscillations in ice volume 3.0 − 0.8 Ma BP are
dominated by a period of 41 Ka, which is the main period of orbital obliquity. Following this
period, the influence of precession increases and there is a switch to a dominance of periods close
to 100 Ka. At the same time, the amplitude of ice volume oscillations increases by as much as 50%.
Following the work of Milankovitch (1941), the present understanding of glacial cycles assumes
that at times of reduced summer insolation, snow and ice can persist at high latitudes through the
summer melt season. Combined with mild winter seasons and enhanced accumulation of snow,
this could lead to growth of ice on land surfaces in the Northern Hemisphere. Based mainly on
data from the last 0.8 Ma, Berger (1976, 1978a,b) concluded that climatic precession plays the
leading role in past climates. Following this lead most researchers plot, or even tune, climate
records to June or July 65◦ N insolation. However, summer insolation at high latitudes is governed
by precession, and is not a good candidate to explain the 41 Ka period cycles dominating the ice
volume record 3.0 − 0.8 Ma BP. Instead, an alternative proposal is discussed, where the obliquity
dominated variations in meridional gradient in insolation influences the poleward flux of moisture
and plays an important role in controlling high latitude ice volume (Raymo and Nisancioglu, 2003,
33
appendix A). Comparing an untuned δ 18 O record with changes in the insolation gradient, suggests
that increased gradients promote ice sheet growth. In the following chapter a simple model is
developed in order to test this hypothesis.
34
Chapter 3
Coupled Atmosphere and Ice Process Model
At present, comprehensive earth system models require too much computing power to simulate
climate on orbital time scales. However, even if the models were capable of simulations on these
time scales, the large number of model parameters involved would significantly increase the number
of possible parameter errors. As a result, the more sophisticated models become less determined
even though the apparent fit of the model results to data is better (Hasselmann, 1981).
The model developed in this chapter is therefore kept as simple as possible to ensure that the
physical processes and feedbacks involved can be tested and understood. Instead of a fully fledged
coupled GCM, the model consists of an atmospheric box model with parameterized heat and moisture fluxes, coupled to a mixed layer ocean and a plastic ice sheet.
3.1 Atmosphere
The model developed here is similar to the atmospheric models of Nakamura et al. (1994), Rivin
and Tziperman (1997), and Gildor and Tziperman (2001). It consists of one hemisphere with
low, mid, and high latitude boxes, each covering 30◦ of latitude (figure 3-1). Since the thermal
adjustment time of the climate system is of order 1 Ka, and the focus is on forcing on orbital
time scales (> 10 Ka), it is assumed that the climate system is in thermal equilibrium with the
forcing. Thus the temperature of each box is calculated by balancing the inputs and outputs of
35
HIN
HIN
HOUT
HIN
HOUT
FSH
T1
FSH
T2
FLH
HOUT
T3
FLH
ICE
0N
30N
60N
F airsea
F airsea
90N
F airsea
Figure 3-1: Sketch of the atmospheric box model. The ice sheet is coupled to the atmospheric
model and extends south from the latitude of the Arctic ocean (75◦ N). The balance of incoming
and outgoing fluxes of heat (Hin , Hout , FSH , FLH ) determines the mean temperature of each box
(T1 , T2 , T3 ).
energy. The important exchanges of energy are: (Hin ) incoming short wave solar radiation at the
top of the atmosphere; (Hout ) outgoing long wave thermal radiation; and (FSH , FLH ) meridional
atmospheric transports of sensible and latent heat (at the latitudes in question the flux of potential
energy is negligible). Integrating over the volume of the individual atmospheric boxes gives the
following expression for atmospheric surface temperature (Tatm ):
Catm
∂Tatm
= Hin − Hout + FSH + FLH + Fairsea
∂t
(3.1)
where Catm = ρw cp ∆zatm Aatm is the effective heat capacity of each atmospheric box taken to be
equivalent to a column of water 2 m deep, Aatm is the surface area of the atmospheric box, and
Fairsea represents the exchange of heat between the atmosphere and a mixed layer ocean (described
in section 3.2).
Incoming solar and outgoing thermal radiation in each box is parameterized following Wang
and Stone (1980)
2
Hin = RE
Z
0
2π
Z
φ2
φ1
36
Q(1 − α)cosφdφdλ
(3.2)
Parameter
∆tatm
ρw
cp
∆zatm
RE
S
Aout
Bout
KSH
KLH
n
Bcc
Value
10
1000
4180
2
6.37 × 106
1350
211
1.7
1.0 × 1026
6.0 × 1038
2.0
5.42 × 103
Units
days
kg/m3
J/(kgK)
m
m
W/m2
W/m2
W/(m2 K)
K
Description
Atmospheric time step
Density of water
Specific heat of water at constant pressure
Equivalent water depth of atmospheric column
Earth’s mean radius
Solar constant
Outgoing radiation constant
Outgoing radiation constant
Sensible heat flux constant
Latent heat flux constant
Eddy heat flux exponent
Clausius Clayperon parameter
Table 3.1: Atmospheric model parameters.
Hout = Aout + Bout Tatm
(3.3)
where RE is the radius of the Earth, Q(φ) is the flux of solar radiation incident at the top of the
atmosphere, α(φ) is the effective albedo of the box, φ is the latitude, λ is the longitude, Tatm is
the area weighted mean surface temperature of each box in degrees Celsius, and (Aout , Bout ) are
constants. The albedo is calculated by accounting for the relative areas covered by snow, ice and
bare land.
3.1.1 Insolation
Because the model will be run for time periods of several 100 Ka, the seasonal insolation is only
calculated at 1 Ka intervals and interpolated to find the seasonal cycle for the current model year.
In order to accurately represent the seasonal cycle and to compare the seasonal output from year to
year, it is important to choose an appropriate yearly calendar (Joussaume and Braconnot, 1997).
In most studies of paleoclimate the conventional 365 day calendar with months and days will
not be appropriate: it is specific to the current orbital configuration with March 21st at the spring
equinox. As was discussed in section 2.1; the position of a conventional calendar date with respect
to the equinox changes with time (figure 2-1). This is alleviated by fixing the spring equinox to
37
TODAY
Spring Equinox
93.6 days
89.5 days
Summer Solstice
Sun
ω
Perihelion
Aphelion
Winter Solstice
Earth
92.7 days
89.0 days
Fall Equinox
126 Ka BP
Fall Equinox
97.3 days
88.5 days
Winter Solstice
Sun
ω
Perihelion
Aphelion
Summer Solstice
Earth
85.4 days
93.8 days
Spring Equinox
Figure 3-2: Sketch of Earth’s orbit around the Sun today (0 Ka BP: ε = 23.4◦ , ω = −78.7◦ ,
e = 0.017), and at Termination 2 (126 Ka BP: ε = 24.0◦, ω = −111.2◦ , e = 0.041), with the
length of each season indicated in number of days as given by Joussaume and Braconnot (1997).
The duration of the seasons is proportional to the area covered between the astronomical positions,
and can vary significantly with time.
38
Figure 3-3: Sketch of the Earth’s orbit around the Sun. The dots represent the position of the planet
after equal intervals of time, and the shaded regions all have equal areas.
March 21st. However, because the number of days in each season varies with time, the solstices and
fall equinox will occur on different dates (figure 3-2). For example, the time between equinoxes and
solstices has varied from 82.5 to 100 days in the last 1 Ma (Berger and Loutre, 1994). Therefore, it
is misleading to use conventional calendar days to express an orbital position in the past.
Due to the problems with the modern calendar it is better to specify the Earth’s position on its
orbit in degrees relative to the spring equinox. In this case there are two options: 1) to use the
true longitude of the Earth in its orbit, or 2) the mean longitude of the Earth with respect to spring
equinox. Here mean longitude (option 2) is defined as the fictive angle described by the Earth if
it were traveling around the sun at a constant velocity (Laskar et al., 1993). In this case, mean
longitude is proportional to time, and each ”degree” on the orbit is equal to one day out of a 360
day year.
Figure 3-3 shows the path of a planet, such as the Earth, in its orbit around the sun. The line
from the Sun to the planet sweeps out equal areas in equal times, following Kepler’s second law.
When using the mean longitude calendar (option 2) to specify the position of the planet on this
orbit, the time interval between each point on the orbit is equal.
39
The true longitude calendar (option 1) is appropriate for plotting an insolation time series for a
fixed angle on the orbit, such as the solstices or equinoxes. However, if this option is chosen when
defining a calendar in a climate model, the model time step would be varying during the course of
a year. Therefore, the mean longitude calendar (option 2), which has a constant time step, will be
used in this study.
3.1.2 Albedo
Snow and ice albedo are important feedbacks in the climate system: when temperatures cool, snow
and ice expand into regions previously covered by surfaces such as forests that have a relatively
low albedo. This increased snow and ice cover raises the surface albedo dramatically, reflecting
incoming radiation, and causing a further decrease in local temperature. In most of the calculations
presented here the potential effects of sea ice are neglected; only variations in snow and ice cover
over land contribute to changes in albedo.
The effective albedo at the top of the atmosphere for ice- and snow-free surfaces is calculated
using the parameterization of Wang and Stone (1980);
αland = α0 +
α2
(3sin2 φ − 1)
2
(3.4)
where the constants α0 = 0.136, and α2 = 0.146 are found using the annual mean zonal albedo and
incident solar radiation distribution given by Ellis and Vonder Haar (1976). This parameterization
takes into account the zenith angle effect which results in an increase in albedo at high latitudes
where the solar zenith angle is large.
The extent of the snow line is controlled by the position of the zero degree isotherm: land areas
with a surface temperature below freezing are assumed to be snow covered and have an albedo
of αsnow = αland + δαsnow . In most cases, part of the area with temperatures below freezing is
already covered by the ice sheet and will therefore have an albedo of αice = αland + δαice . When
calculating the effective albedo the relative areas covered by bare land, snow (fsnow ), and ice (fice )
40
Land fraction (%)
100
80
60
40
20
0
−90
−70
−50
−30
−10
10
Latitude (degrees)
30
50
70
90
Figure 3-4: Latitudinal distribution of land masses (Birchfield et al., 1982). The mean land fraction
in the Northern Hemisphere is approximately 40%. However, at high northern latitudes south of
the Arctic ocean the land fraction is closer to 60%.
are taken into account;
αef f ective = αland + fsnow δαsnow + fice δαice
(3.5)
For simplicity, the albedo of bare land and ocean surfaces are taken to be equal (αland ). However,
only 40% of each latitude band can be covered by snow and ice ((fsnow + fice ) < 0.40), which is
roughly equal to the average land fraction in the Northern Hemisphere (figure 3-4). At latitudes
north of 75◦ N, in the Arctic ocean, the surface is assumed to be permanently covered by sea ice
with an albedo fixed at αsnow .
The strength of the snow and ice albedo feedback is strongly dependent on the values chosen
for δαsnow and δαice . Using results from a 1-D radiative convective model Wang and Stone (1980)
estimated that the change in zonal earth-atmosphere albedo from ice-free to ice covered conditions
is approximately δαice = 0.186. A similar value for δαice is obtained when using the empirical
relations of Cess (1976) and Coakley (1979) to convert surface albedo into effective top of the
atmosphere albedo taking into account the effect of clouds and solar zenith angle (Lian and Cess,
1977). Figure 3-5 shows latitudinal profiles of zonal and annual mean top of the atmosphere albedo
over land surfaces with ice-free conditions with an albedo of 0.25, versus completely ice covered
conditions with an albedo of 0.65. Note that the value of δαice increases slightly at lower latitudes,
41
TOA albedo over land
0.7
a)
0.6
0.5
0.4
ice−free
ice−covered
0.3
0.2
b)
δαice
0.19
0.18
0.17
0
10
20
30
40
50
Latitude (degrees)
60
70
80
90
Figure 3-5: a) Zonal and annual mean top of the atmosphere (TOA) albedo over land surfaces
where the ice-free surface albedo is 0.25 and ice-covered surface albedo is 0.65 estimated using
the empirical relations described in Coakley (1979). b) Difference in TOA albedo for ice-free and
ice-covered conditions. The cloud cover fraction is assumed to be constant at 50% at all latitudes:
the observed zonal annual mean cloud cover in the Northern Hemisphere varies from about 40% to
65% (Cess, 1976).
suggesting that the strength of the ice-albedo feedback increases as snow and ice moves toward the
equator. However, this effect is quite small, and δαsnow and δαice are both kept fixed at a value of
0.186 in the model.
When calculating global mean planetary albedo, local albedo is weighted by insolation;
αp =
R π/2
−π/2
Q(φ)αef f ective (φ)dφ
R π/2
−π/2 Q(φ)dφ
(3.6)
where Q(φ) is the mean annual distribution of radiation reaching the top of the atmosphere, and
αef f ective (φ) is mean top of the atmosphere albedo at latitude φ. The latitudinal profile of the
42
70
Albedo (%)
60
50
40
DJF
JJA
Annual
Model
30
20
0
10
20
30
40
50
Latitude
60
70
80
90
Figure 3-6: Observed top of the atmosphere albedo in percent from the Earth Radiation Budget
Experiment (ERBE) averaged over the period from November 1984 to February 1990, for mean
summer (June, July, August: JJA), winter (December, January, February; DJF), and annual conditions (Li and Leighton, 1993). The dash-dotted line is the albedo profile estimated by the model,
assuming all latitudes above 75◦ N are permanently covered by sea ice.
Earth’s albedo for summer, winter, and annual mean conditions from the Earth Radiation Budget
Experiment (ERBE) is plotted in figure 3-6. Observed annual mean planetary albedo from ERBE
is 0.297 (Li and Leighton, 1993). Albedo estimated from equation 3.4, with latitudes above 75◦ N
assumed to be permanently covered by sea ice, is plotted in the same figure. In this case the
planetary albedo is αp = 0.30, which agrees well with the ERBE data.
Absorbed incoming radiation in each box of the atmospheric model is calculated by approximating the latitudinal insolation distribution with a polynomial expansion;
Q(φ) = Q0 +
Q4
Q2
(3 sin2 φ − 1) +
(35 sin4 φ − 30 sin2 φ + 3)
2
8
(3.7)
where the polynomial coefficients (Q0 , Q2 , Q4 ) are found by fitting the insolation profile (Q(φ)) to
insolation calculated at the mid-points of the three atmospheric boxes as described in the previous
section.
43
3.1.3 Atmospheric Eddy Heat Flux Parameterization
In the model it is assumed that meridional atmospheric transport is dominated by baroclinic transient eddy transport. This is a good approximation at latitudes at and above about 30◦ N, where the
contribution to atmospheric transport by stationary components is relatively small (Trenberth and
Stepaniak, 2003). From the theory of baroclinic eddies, meridional atmospheric eddy sensible heat
flux (FSH ) can be parameterized as (Green, 1970; Stone, 1972)
FSH (φ) = KSH
∂Tatm (φ)
∂φ
!n
(3.8)
where Tatm (φ) is the zonal mean surface temperature at latitude φ, and KSH is chosen such that the
heat fluxes between the atmospheric boxes are close to the modern observed values. The exponent
is kept constant at all latitudes at a value of n = 2.0, which is at the lower range of the theoretical
and empirical estimates: Held (1978) found that the heat flux should depend on the meridional
temperature gradient to a power between 2.0 and 5.0, with the higher powers being favored at low
latitudes where the β effect is strong. By using seasonal data Stone and Miller (1980) estimated
that the power varies from 1.6 ± 0.5 at 60◦ N to 3.4 ± 0.8 at 30◦N, which is consistent with the
results of Held (1978). The equation for the heat flux can also be derived from more sophisticated parameterizations, which have been tested against GCM simulations (Stone and Yao, 1990).
However, it should be noted that the parameterization has not been tested for climates significantly
different from the modern, such as during the glacial periods.
In addition to sensible heat flux, a parameterization for meridional atmospheric moisture flux
(Fw ) is required. The moisture flux is particularly important because of its effect on accumulation
of snow on the ice sheets. A common procedure is to relate the meridional eddy flux of latent heat
to the flux of sensible heat as follows (Leovy, 1973; Stone and Yao, 1990)
FLH (φ) =
Lv ∂qs (Tatm (φ))
qr
FSH
cp
∂T
(3.9)
where Lv is the latent heat of vaporization, and qr is the relative humidity. The saturation specific
44
30
3 boxes
2 boxes
NCEP
ο
Temperature ( C)
20
10
0
−10
−20
0
10
20
30
40
50
Latitude (degrees)
60
70
80
90
Figure 3-7: Approximate annual and zonal mean temperature profiles calculated using even order
Legendre Polynomials compared with NCEP 1000 mb temperatures (Kalnay et al., 1996). The first
two and three even order polynomial coefficients are retained in a two and three box version of the
model, respectively.
humidity qs (Tatm ) depends on the local atmospheric temperature at latitude φ, as described by the
Clausius-Clapeyron relation;
qs (Tatm ) =
Acc −Bcc /Tatm
e
Ps
(3.10)
where Ps is the surface pressure, and (Acc , Bcc ) are constants. Substituting into equation 3.9 gives
e−B/Tatm
FLH (φ) = KLH
2
Tatm
∂Tatm (φ)
∂φ
!n
(3.11)
where the relative humidity (qr ) is assumed constant, and incorporated into KLH .
3.1.4 Approximate Meridional Temperature Profile
The fluxes of sensible and latent heat between the boxes depend on the meridional surface temperature gradient at the interfaces. At the same time, the flux of latent heat depends on local temperature
through the Clausius-Clapeyron relationship. To estimate these quantities at specific latitudes, the
meridional profile of zonal mean sea level temperature (Tatm (φ)) is approximated using even order
45
Legendre Polynomials as follows;
Tatm (φ) = P0 +
P4
P2
(3 sin2 φ − 1) + (35 sin4 φ − 30 sin2 φ + 3)
2
8
(3.12)
where the coefficients (P0 , P2 , P4 ) are determined by fitting the temperature profile (Tatm (φ)) to the
mean temperatures of the atmospheric boxes as given by the energy balance (equation 3.1). Thus,
P0 , P2 , and P4 become the primary dependent variables of the model, and are determined by the
heat balance equations for each box. North (1975) tested the performance of the Legendre Polynomial approximation in an energy balance model and found a good fit to the observed meridional
temperature profile when retaining the first two even polynomial coefficients (P0 & P2 ), as well as
a small improvement when including three coefficients (equation 3.12). A plot of the two approximate modern temperature profiles as well as the zonal and annual mean 1000 mb temperatures
from NCEP are shown in figure 3-7.
Similarly, the temperature gradient at specific latitudes is found by differentiating equation 3.12:
P2
P4
∂Tatm (φ)
=
(3 sin2 2φ) +
cos φ(35 sin3 φ − 15 sin φ)
∂φ
2
2
(3.13)
The resulting meridional profile of atmospheric temperature gradient is used together with local
temperature to estimate the fluxes of heat at the box interfaces, as well as the flux of latent heat at
the edge of the ice sheet which controls the accumulation of new ice.
It should be noted that by only including even order terms in the polynomial expansions, it is
implicitly assumed that the temperature gradient at the equator is zero. To account for seasonal
fluxes across the equator in an inter hemispheric model, it is necessary to include the odd order
terms in the expansion.
46
Parameter
∆zmix
fland
τairsea
Value
75
0.40
2.5 × 107
Units
m
Description
Mixed layer depth
Fraction of each latitude covered by land
Restoring time scale for air-sea flux
s
Table 3.2: Mixed layer ocean parameters.
3.2 Mixed Layer Ocean
As the model does not include a representation of the ocean, it is necessary to couple the atmospheric boxes to a mixed layer ocean in order to correctly resolve the seasonal cycle (North and
Coakley, 1979).
Here, the energy balance of the mixed layer is given by
Cmix
∂Tmix
= −Fairsea
∂t
(3.14)
where Cmix = ρw cp ∆zmix Aatm (1 − fland ) is the effective heat capacity equivalent to a layer of
water ∆zmix = 75 m deep, and Tmix is the mean temperature of the mixed layer. The flux of
heat between the mixed layer boxes and atmospheric boxes (Fairsea ) is defined as a restoring flux
following the work of Haney (1971)
Fairsea =
Cmix
(Tmix − Tatm )
τairsea
(3.15)
where τairsea is a restoring time scale, and Tatm is the temperature of the overlying atmospheric box.
fland is the fraction of each latitude belt covered by land. Although this factor depends on latitude
it will be kept constant at 40% in the following experiments. The time scale for the interaction
between the atmosphere and mixed layer (τairsea ) is adjusted in order to give the correct amplitude
and phase of the modern observed seasonal cycle for the specified mixed layer depth.
47
h
Ice Sheet
x
x=L
x=0
o
Bedrock
75 N
Arctic Ocean
North
Figure 3-8: Sketch of the ice sheet configuration. The ice sheet extends south from the Arctic ocean
at 75◦ N with a total meridional extent of 2L.
3.3 Ice Sheet
A common procedure when investigating the sensitivity of Earth’s climate to changes in size of
the polar ice sheets is to fix the ice extent to the latitude of the zero degree isotherm (e.g. Budyko
(1969); Sellers (1969); North (1975); Schneider and Thompson (1979); North and Coakley (1979)).
Although this is a valuable exercise, it assumes that the glacial growth and retreat of the major ice
sheets is controlled exclusively by local surface temperature. Here, a different approach is taken
where the extent of the ice sheet is calculated with a simplified ice sheet model, taking into account
the balance of snowfall and melt (figure 3-8).
The ice in the model is treated as a perfectly plastic material (Orowan and Perutz, 1949; Nye,
1951; Reeh, 1982). In this case the deformation is negligible when the applied stress is below some
critical value, defined as the yield stress (τ0 ). When the stress is larger than the yield stress, the
material is assumed to deform instantly to relieve the applied stress. Although this may not be
realistic, it gives a reasonable approximation for reconstructing the large ice sheets found during
glacial periods (Weertman, 1976, 1964).
As the stress in the ice cannot exceed the yield stress, the drag at the base of the ice sheet (basal
drag) must be equal to τ0 , such that
−ρice gh
∂h
= τ0
∂x
48
(3.16)
4000
h(x) (m)
3000
10 kPa
50 kPa
100 kPa
2000
1000
0
1500
1000
500
0
x (km)
Figure 3-9: Height profiles of a plastic ice sheet with a constant volume per unit width of 3.0 ×
109 m3 , but with varying values for the yield stress τ0 . A small value of τ0 reduces the ice thickness.
where ρice is the density of ice, g is the acceleration due to gravity, x is the distance from the center
of the ice sheet, and h is the elevation of the ice surface. Assuming that the ice sheet is resting on a
flat bed, this equation can be integrated to give
h2 (x) − h2 (0) =
2τ0
(x − x0 )
ρice g
(3.17)
By applying the condition that the ice thickness has to be zero at the edges, and prescribing the half
length of the ice sheet by L it follows that the profile is parabolic;
h2 (x) =
2τ0
(L − x)
ρice g
(3.18)
The value of the plastic yield stress determines the height to length ratio of the ice sheet: a large
value of τ0 corresponds to a thick ice sheet; whereas a small value of τ0 reduces the ice thickness,
which could result from larger ice velocities (figure 3-9). The thickness at the center of the ice
sheet (x = 0) is H = (2τ0 L/ρg)1/2 .
Based mostly on measurements from Alpine valley glaciers, the yield stress is estimated to
vary in the range τ0 = 50 − 150kP a (1kP a = 103 P a) (Patterson, 1994, p. 240). Using a
value of τ0 = 100kP a, together with the observed size of the ice sheet in central West Greenland
(L = 500km) gives a maximum ice thickness of 3200m, which compares well with the observed
49
Parameter
φarctic
ρice
τ0
αabl
βabl
βatm
Lv
δαsnow , δαice
Value
75N
920
5 × 104
0
3.0 × 10−8
7.0 × 10−3
2.5 × 106
0.186
Units
degrees
kg/m3
Pa
m/s
m/(s × K)
K/m
J/(kg × K)
Description
Latitude of edge of Arctic ocean
Density of ice
Yield stress of ice
Ablation constant
Ablation constant for temperature
Atmospheric lapse rate
Latent heat of vaporization
Albedo anomaly for snow and ice
Table 3.3: Ice sheet model parameters.
value of 3150m. However, in most cases τ0 = 100kP a is too high a value for present observed
ice sheets, where the yield stress is observed to vary from 0 to 100kP a with a mean of ∼ 50kP a
(Patterson, 1994, p. 242).
The Laurentide ice sheet is believed to have been considerably flatter than modern ice sheets
due to an underlying bed of deformable sediment with a low yield stress (Clark et al., 1999). This
deformable sediment layer is suggested to have been underlying the entire Laurentide ice sheet
prior to the middle Pleistocene (Clark and Pollard, 1998). As a result, the ice sheet is thought to
have been as extensive as in the late Pleistocene, while only containing about one third to one half
as much ice (figure 2-4). For this to be possible the basal yield stress would have been lower than
what is observed in Greenland and Antarctica today, resulting in a relatively thin ice sheet. Based
on these assumptions a yield stress of τ0 = 10kP a is chosen: In this case, an ice sheet with about
half the estimated LGM volume of the Laurentide will have the same meridional extent as observed
for the LGM.
3.3.1 Ice Sheet Mass Balance
When using the plastic ice sheet model described in the previous section to calculate fluctuations in
size of the ice sheet it has to be coupled to the surface mass balance. Assuming that the ice sheet is
symmetric and integrating the parabolic height profile (equation 3.18) over the southern half (x = 0
50
to x = L), the volume per unit width of half the ice sheet is
V =
Z
L
0
2
h(x)dx =
3
2τ0
ρice g
!1/2
L2/3
(3.19)
Change in ice volume over time is governed by the relative rates of snowfall and melting;
∂V
=
∂t
Z
L
(acc − abl)dx
(3.20)
0
where acc is accumulation (snowfall) and abl is ablation (melting). Using this equation, together
with the parabolic height profile, the volume and area covered by the ice sheet can be calculated,
giving a value for the albedo. Note that it is implicitly assumed that any changes in the mass balance
of the northern half of the ice sheet is balanced by calving into the Arctic ocean (Weertman, 1964).
3.3.2 Ablation
There are several options for estimating melting of ice and snow on the ice sheet. The first two are
based on the model surface temperature as the only input, whereas the third involves calculating
the energy balance at the surface of the ice.
(1) The simplest option is based on a linear relation between temperature at the ice surface and
ablation. From observations in Greenland, Ohmura et al. (1996) found that the total annual ablation
can be expressed as a linear function of mean summer temperature:
abl = 0.930 + 0.514 × TJJA when TJJA > −1.8◦ C
(3.21)
where ablation (abl) is given in units of (m water equivalent)/yr and TJJA is the mean surface temperature (in ◦ C) during the three summer months (June, July, & August). Because this relationship
is based on the observed annual loss of ice mass it takes into account the effects due to re-freezing
of melt water at the surface of the ice sheet.
Pollard (1980) suggests including a similar dependency of ablation on net radiation. However,
51
melting already depends to a large degree on incoming long wave radiation, which is strongly
influenced by the air temperature above the ice (Ohmura, 2001). At the same time, observations
show that incoming short wave solar radiation is poorly correlated with ablation (Braithwaite and
Olesen, 1989). Therefore, including a term relating insolation to ablation does not improve the
ablation parameterization.
(2) A more common procedure used to calculate ablation is based on the Positive Degree Day
(PDD) method. It was first used to estimate ablation on glaciers in the Alps, and was later tested
for conditions on Greenland (Braithwaite and Olesen, 1989; Reeh, 1989; Huybrechts et al., 1991;
Braithwaite, 1995). In this case, daily ablation rate (m/day) at any location on the ice sheet is
related to local air temperature when the temperature is above the melting point;
abl(m/day) = αabl + βabl Ts
(3.22)
where αabl and βabl are empirical constants. However, the observed correlation between ablation
and temperature is found to be higher when averaged over periods of months, rather than days
(Braithwaite and Olesen, 1989). Therefore, ablation is often given as the total annual ablation
(m/yr), and related to the number of days with temperatures above the melting point (N) as well
as a positive degree day sum (P DD)
abl(m/yr) = αabl N + βabl P DD
(3.23)
where the positive degree day sum is defined as the yearly sum of daily temperatures above the
melting point:
P DD =
Z
0
year
Ts dt where Ts > 0◦ C
(3.24)
Physically αabl represents the rate of melting when the air temperature is at 0◦ C, while βabl , often
called the degree-day factor, describes the increase of melt rate with temperature. The relation
between ablation and temperature has been tested on several glaciers in Greenland and their cor-
52
relation has been found to be fairly strong: for the glacier Qamanârssûp sermia in West Greenland
the correlation between monthly temperature and ablation is r = 0.93 with a degree-day factor
βabl = 0.08m/(day ◦C) (Braithwaite and Olesen, 1989). However, the value for αabl (rate of melting when Ts = 0◦ C) is not found to be significantly different from zero. The data also indicates
that the degree-day factor is reasonably constant throughout the year, with no marked differences
between summer and winter.
It should be noted that the degree-day factor quoted above has been found for ice surfaces
with little or no snow cover throughout the year, and observations suggest that the positive-degree
day factor for snow is on the order of βabl = 0.03m/(day ◦C), i.e. less than half that for ice
(Braithwaite, 1995; Braithwaite and Zhang, 2000). This indicates that the rate of ablation could
decrease significantly during times of net accumulation, resulting in different ablation rates for
growing and shrinking glaciers and ice sheets.
(3) The third and perhaps most sophisticated method to estimate ablation involves calculating
the energy balance at the ice sheet surface. In this case the latent heat of melt is found as a residual,
after accounting for all the processes involved in supplying and removing heat to the layer of ice
and snow at the surface of the ice sheet. The variables necessary to perform this calculation are
difficult to measure, and in a model the different processes involved in the surface energy balance
are not well understood. Therefore, the two previous options have been the ones most frequently
used in the literature.
In the model presented here, the two first ablation parameterizations are tested. At the same
time, the difference between using daily, or seasonal mean temperatures when calculating ablation
with the degree day method is investigated. The temperature at the surface of the ice sheet is found
using the Legendre Polynomial expansion (equation 3.12). Whereas accumulation occurs over the
whole area of the ice sheet, melting only takes place in a relatively narrow zone on the steep margins
of the ice sheet. Therefore, it is necessary to take into account the atmospheric lapse rate when
calculating the local temperature for ablation. Here a constant lapse rate of βatm = 6.0 × 10−3 K/m
is chosen.
53
3.3.3 Accumulation
Partly because Milankovitch theory is based on the assumption that ablation controls the growth
and decay of ice sheets on orbital time scales, the treatment of accumulation in conceptual models
has been very crude. In the simplest models the snow line is fixed to the latitude of the zero degree
isotherm, thereby completely ignoring any contribution by the advection of moisture (e.g. Budyko
(1969); Sellers (1969); Schneider and Thompson (1979); North et al. (1983)). Suarez and Held
(1979) improve this simple representation by specifying a fixed rate of accumulation with latitude
whenever temperatures fall below freezing, and find that the sensitivity of the model is sharply
reduced. Similarly, in a model including a physical representation of ice sheets, Pollard et al. (1980)
estimate accumulation by simply applying a modern observed precipitation pattern wherever the
model compute temperatures below freezing. Even in more sophisticated GCMs a similar approach
is often used (e.g. Huybrechts (1990); Gallee et al. (1991); Deblonde et al. (1992); Tarasov and
Peltier (1997); Huybrechts and de Wolde (1999)), where a precipitation field is taken from present
day climatology, but perturbed by taking into account the deviation of the simulated temperature
field from present values. This means that only the rate of accumulation can change, but not its
pattern. At the same time the change in the rate is only a function of local temperature with no
influence by atmospheric transport.
As opposed to these previous model studies, the effect of atmospheric transport is explicitly
included in the calculation of accumulation in the present model. In this case, net precipitation
(precipitation - evaporation) on the ice sheet is set by the meridional moisture flux at the southern
edge of the ice (φicedge )
acc = FLH (φicedge )/(Lv ρice )
(3.25)
where FLH (φicedge ) is the latent heat flux at the ice edge from equation 3.11 on page 45, Lv is the
latent heat of vaporization, and ρice is the density of ice. All the precipitation is assumed to fall as
snow, contributing to accumulation of additional ice. This is not considered a serious constraint,
since only a small fraction of yearly precipitation on an ice sheet falls as rain, and a large part of
54
this rainfall is expected to re-freeze into superimposed ice (Huybrechts et al., 1991).
3.4 Summary
An atmosphere-ice process model is developed which is efficient enough to perform multiple experiments covering long time periods (> 1 Ma). At the same time, the model is kept simple,
in order to be able to understand the underlying physics. There are a few important difference
between the new model and existing similar models. An explicit parameterization is included for
calculating atmospheric moisture flux and accumulation on the ice sheet, instead of assuming that
accumulation is only related to local temperature. At the same time, the ice-albedo feedback is
improved compared to the early one dimensional energy balance models, and found to agree well
with empirical data on the change in albedo from ice free to ice covered surfaces. These two factors
will be important in the following chapters when investigating the impact of insolation changes on
changes in the mass balance and albedo of high latitude ice sheets.
55
Chapter 4
Model Validation and Climate Sensitivity
Climate sensitivity is most often measured as the response of global mean surface temperature to
changes in insolation or atmospheric CO2 . In the following chapter the sensitivity of the model
climate is evaluated and compared with previous model studies. Sensitivity studies with energy
balance models often use a simplified parameterization of the ice-albedo feedback (e.g. Budyko
(1969); Sellers (1969); North (1975); North and Coakley (1979); Schneider and Thompson (1979)).
In this case, ice, or snow extent is fixed to the zero degree isotherm. In reality, the extent of an ice
sheet is set by the relative rates of accumulation and ablation. The strength of the ice-albedo
feedback and the sensitivity of climate when using these two different approaches in modeling ice
extent will be discussed.
Before investigating climate sensitivity it is important to ensure that the model control climate yields surface temperatures close to the observed climate. This increases the likelihood that
the important feedbacks in the system that depend on temperature will have strengths which are
comparable to the current climate. Therefore, in the following section the control experiment is
described, and the simulated modern climate is compared to observations.
56
b)
−10
−2
0
50
5
−5
0
60
5
40
30
0
Latitude (degrees)
70
0
5
−1
−1
−15
80
a)
−5
90
−10
−5
10
10
15
15
0
5
20
20
10
25
15
20
30
60
120
180
240
Mean Orbital Longitude (degrees)
25
30
0
0
20
25
10
300
DJF
JJA
YEAR
360 −40 −20
0
20 40
Temperature (οC)
Figure 4-1: Modern atmospheric meridional temperature as calculated by the model (◦ C): a) daily
mean values as a function of latitude; b) winter (DJF), summer (JJA) and annual (YEAR) mean as
a function of latitude. The start of the year is fixed at spring equinox and the year is divided into
360 degrees, where 90◦ is summer solstice (see section 3.1.1 on page 37).
4.1 Control Experiment with Modern Insolation
The control experiments described here are all forced with modern seasonal insolation and the
standard parameters given in table 3.1. An equilibrium solution with a stable seasonal cycle is
attained by integrating the equations in section 3.1 forward using a leapfrog scheme with a timestep of 10 days. This process takes about 100 years.
4.1.1 Atmospheric Meridional Temperature Profile
The equilibrium surface temperature structure is plotted in figure 4-1a as a function of time of
year and latitude. Figure 4-1b shows summer (June, July, & August), winter (December, January,
February), and annual mean temperature as a function of latitude.
The model temperatures agree reasonably well with observed 1000 mb atmospheric temperature
57
90
a)
80
Latitude (degrees)
70
b)
0
−1
−5 0
− −1
−5 10 5
−25
−20
5 0
−15
−10
60
5
50
40
10
15
20
30
25
10
−5
0
15
5
20
10
15
25
20
20
25
10
0
0
60
120
180
240
Mean Orbital Longitude (degrees)
300
DJF
JJA
YEAR
360 −40 −20
0
20 40
Temperature (οC)
Figure 4-2: Modern atmospheric meridional temperature climatology at the 1000 mb level from
NCEP data covering the period 1979-2001: a) Daily values as a function of latitude; b) winter
(DJF), summer (JJA) and annual (YEAR) mean as a function of latitude.
climatology from NCEP reanalysis data covering the period 1979-2001 shown in figure 4-2. The
phase of the model seasonal cycle is comparable to observations, but the increase in amplitude
of the seasonal cycle with latitude is smaller. Disagreement between the model and observations
is mainly found at low latitudes. This is due to the inability of the heat flux parameterization to
capture the meridional heat flux due to the Hadley cell, as well as the neglect of cross-equatorial
heat transports.
The fit of the amplitude of the seasonal cycle is accomplished by adjusting the relaxation time
of the atmospheric and mixed layer temperatures (∆tairsea in equation 3.15): increasing ∆tairsea
increases the amplitude of the seasonal cycle. Similarly, the meridional temperature structure is
fit to observations by adjusting the strength of the meridional heat fluxes (equations 3.8 & 3.11):
increasing the constants KSH and KLH decreases the meridional temperature gradient.
58
7
Sensible
Latent
Total
TC2000
Heat Flux (PW)
6
5
4
3
2
1
0
0
10
20
30
40
50
Latitude (degrees)
60
70
80
90
Figure 4-3: Modern meridional profiles of annual and zonal mean atmospheric sensible, latent, and
total heat fluxes as calculated by the model (1 P W = 1015 W ). Also shown is estimated total
atmospheric heat flux from NCEP (Trenberth and Caron, 2001).
4.1.2 Atmospheric Eddy Heat Flux
The model profiles of annual mean atmospheric meridional sensible and latent heat fluxes are shown
in figure 4-3 together with observed total atmospheric heat flux estimated from NCEP data by
Trenberth and Caron (2001). The total heat flux calculated by the model is slightly larger than
the observed value. However, any contribution by ocean heat fluxes is not included in the model.
Therefore, the atmospheric heat flux is forced to compensate in order to obtain a reasonable meridional temperature structure, in particular at low latitudes. The relative contribution by sensible
and latent heat fluxes is comparable to observations, where the maximum latent heat flux has a
magnitude about half that of the sensible heat flux. The components of the observed meridional
heat flux are shown in figure 4-4: at latitudes poleward of 30◦ sensible and latent heat fluxes dominate, whereas at lower latitudes the influence by the flux of potential energy becomes important.
As this study is mostly concerned with high latitude processes related to the growth and decay of
ice sheets, the inadequate representation of low latitude heat fluxes is not thought to be a problem.
At equilibrium, the atmospheric heat flux calculated by the model agrees reasonably well with
observations. However, a more important question is whether the sensitivity of the flux to changes
59
Heat Flux (PW)
10
5
0
Sensible
Latent
Kinetic
Potential
Total
−5
−10
−80
−60
−40
−20
0
20
Latitude (degrees)
40
60
80
Figure 4-4: Modern meridional profiles of annual and zonal mean atmospheric sensible, latent,
kinetic, and geopotential heat fluxes from NCEP climatology based on data covering the period
1979-2001 (PW) (Trenberth and Stepaniak, 2003).
in meridional temperature structure implied by the parameterizations discussed in section 3.1.3
is realistic. An empirical study by Stone and Miller (1980) based on mean seasonal data finds a
high correlation (90%) between total meridional atmospheric sensible heat flux and the 1000 mb
meridional temperature gradient. When comparing the temperature gradient to fluxes due to only
transient and stationary eddies, the correlation is increased to 97%. Similar high correlations are
found when comparing the observed flux of latent heat by transient eddies with those calculated
using the observed meridional temperature structure and equation 3.11 on page 45 (Scott, 1995).
Figure 4-5 taken from Scott (1995) shows the log-log correlation between latent heat flux as
estimated by the parameterization in equation 3.11 on page 45 with the observed meridional flux
of latent heat in the atmosphere by all eddies, transient eddies alone, and stationary eddies alone.
The data used are 10 years of zonally averaged monthly data for the period 1963 − 1973 from
Oort (1983). The temperature data is vertically averaged, and the gradient is computed as the
temperature difference between two latitude belts, separated by a distance of 30◦ . This distance
is large enough to ensure that primarily planetary scale forced variations are diagnosed (Lorenz,
1979; Stone and Miller, 1980). The dashed line in the figure represents the 99.9% confidence level
according to the Student’s t-test, as discussed in Stone and Miller (1980).
60
Correlation Coefficient
1
0.5
0
-0.5
-1
10
20
30
40
50
60
70
Latitude (degrees)
Figure 4-5: Log-log correlation of eddy latent heat flux estimated using the parameterization in
equation 3.11 on page 45 with the observed meridional flux of latent heat in the atmosphere by
all eddies (◦), transient eddies alone (×), and stationary eddies alone (∗) as a function of latitude
(Scott, 1995). The dashed line represents the 99.9% confidence level.
Figure 4-5 shows a very high correlation for both total and transient eddy fluxes of latent heat,
comparable to the results of Stone and Miller (1980). It makes very little difference whether the
correlation with the mean flux is calculated using vertically averaged temperature (figure 4-5), or
surface temperature (not shown). The correlation for the stationary eddy flux is relatively low,
and there is a negative correlation near 30◦ N due to the summer Indian monsoon (Scott, 1995).
However, the contribution by stationary eddy fluxes to the total eddy heat flux is relatively small,
in particular at mid and high latitudes.
The same study also investigated the values of the exponent n in equation 3.11 on page 45
which gives the best power law fits to the eddy flux of latent heat in the atmosphere. Figure 4-6
shows that the best power law fits for latitudes north of 30◦ N have exponents in the range n = 1−2,
with the highest values found for the total latent eddy heat flux.
During glacial times the meridional temperature structure would have been significantly different from today. At the same time, the presence of large ice sheets at high latitudes of the Northern Hemisphere could have influenced atmospheric circulation. Experiments with an atmospheric
GCM by Kageyama et al. (1999) with specified modern sea surface temperatures and insolation
61
2.5
Exponent
2
1.5
1
0.5
0
10
20
30
40
Latitude
50
60
70
Figure 4-6: Exponent n in equation 3.11 on page 45 giving the best fit for the zonal mean atmospheric flux of latent heat by all eddies (◦), and transient eddies only (×) as a function of latitude
(Scott, 1995).
fixed to values appropriate for the last glacial inception (∼ 115 Ka BP) indicate an increase in
storm activity and precipitation at high latitudes. These changes are a direct result of an increase
in the meridional temperature gradient forced by the change in insolation. In similar experiments
with last glacial maximum (LGM) boundary conditions, the change in weather patterns is more
severe. There is a significant increase in storm activity, however this is only partially translated into
precipitation changes due to much colder temperatures at high latitudes.
Similar results for the last glacial inception are found by Khodri et al. (2001) in experiments
with a coupled atmosphere-ocean GCM: poleward moisture flux is strengthened in response to a
stronger meridional temperature gradient. The strong response of the atmosphere is partly a result
of a weakened meridional overturning circulation and heat transport by the ocean. These model
studies suggest that in the past, there is a strong connection between atmospheric heat fluxes and
the temperature gradient, similar to what is observed today (Stone and Miller, 1980).
62
Sensitivity parameter β (◦ )
∆T when S + 2% (◦ C)
∆T when S − 2% (◦ C)
no snow active snow no snow
/no ice
/no ice
/active ice
138
166
219
+2.8
+3.2
+3.4
−2.8
−3.4
−4.1
Table 4.1: Sensitivity (β) of different model configurations with and without active snow and ice
covers. In all three experiments the land fraction is set to 40%.
4.2 Climate Sensitivity
When comparing different climate models it is useful to measure their sensitivities. Climate sensitivity (β) is defined as the change in global mean sea level temperature (Tmean ) in response to a
small change in solar constant (S)
β=S
∂Tmean
∂S
(4.1)
Another common measure for sensitivity is the equilibrium response of global surface temperature
to a doubling of atmospheric CO2 . This is approximately equal to a 2% increase in solar constant.
According to the Intergovernmental Panel on Climate Change (IPCC) sensitivity of climate to a
doubling of CO2 is likely to be in the range 1.5 − 4.5◦ C (Cubasch and Meehl, 2001, p. 527). These
estimates are given by experiments with coupled ocean atmosphere GCMs.
The response of the simple process model to changes in solar constant can be compared with
sensitivity estimates from the IPCC as well as other climate models. The sensitivity parameter (β)
and the change in temperature due to a 2% change in solar constant in model experiments with and
without interactive snow and ice components are given in table 4.1.
The sensitivity of the most basic version of the model without an active snow- or ice-albedo
feedback is β = 138◦ , which implies a 1.38◦ C global mean temperature decrease for a 1% decrease
in solar constant. This only accounts for the effect of the negative thermal radiation feedback, which
regulates the amount of outgoing radiation. The strength of this feedback is set by the value chosen
for the parameter B in equation 3.3, and implicitly accounts for water vapor and cloud feedbacks.
When including an active snow-albedo feedback, where the snow line over land follows the
63
Model
Budyko (1969)
Sellers (1969)
North (1975)
Lian and Cess (1977)
Coakley (1979)
Wetherald and Manabe (1975)
Wang and Stone (1980)
β no snow β with snow
155◦
400◦
150◦
326◦
140◦
400◦
◦
147
184◦
152◦
207◦
146◦
185◦
◦
135
188◦
Table 4.2: Climate sensitivity (β) found in model studies with and without an active snow-albedo
feedback. In the studies where the value of β is not given, it is evaluated using equation 4.1 with
the appropriate change in S as considered in the model experiments.
0◦ C isotherm, the sensitivity is increased to β = 166◦. This is lower than the values found in
the energy balance model studies of Budyko (1969); Sellers (1969); North (1975), mostly due to
different albedo parameterizations: in these early model studies the change in albedo as the surface
changes from ice-free to ice-covered (δαsnow ) is significantly larger than the value used in this
study. On the other hand, the sensitivity found here is similar to that obtained with the improved
albedo parameterizations used in the studies of Lian and Cess (1977) and Coakley (1979), as well
as the GCM experiments of Wetherald and Manabe (1975) (table 4.2). The albedo parameterization
introduced by Lian and Cess (1977) and Coakley (1979) takes into account the effect of clouds and
solar zenith angle (see section 3.1.2 on page 40).
Note that the strength of the albedo feedback is also influenced by the negative dynamical flux
feedback: as snow cover increases, high latitudes cool, and the meridional temperature gradient
increases. Thus, the flux of heat to high latitudes increases, hindering the advance of the snow
line. The strength of the dynamical feedback depends on the exponent in the parameterization
of atmospheric heat flux (equations 3.8 & 3.11). In most previous studies with energy balance
models, heat transport is represented by linear diffusion (n = 1). As discussed in section 3.1.3,
the empirical value is closer to n = 2. This increases the strength of the dynamical feedback, and
reduces the sensitivity in models with an active snow-albedo feedback.
Replacing snow cover with an active ice sheet results in a sensitivity of β = 219◦ . This is
significantly larger than the sensitivity found with the simplified snow-albedo feedback. In the case
64
of the ice sheet, growth is not limited as strongly by the zero degree isotherm. Instead, surface melt
increases as the margin moves toward the equator and temperatures increase. The strength of this
negative feedback is controlled by the melt factor (βabl ) in the ablation parameterization (equation
3.22). If the ice margin is to be constrained to follow the zero degree isotherm, the melt factor has
to be increased significantly beyond its empirical value. At the same time as ablation increases with
the advance of the ice margin, accumulation increases. This is because the meridional moisture flux
is larger at mid latitudes as shown in figure 4-3, and results in a positive feedback which further
enhances the sensitivity of the model. When both an active snow cover and ice sheet are included
(not shown) the results are similar to the case with only an active ice sheet. This is because the
surface area sufficiently cold for there to be snow in summer is covered by the ice sheet already,
and the change in winter snow has a minor influence on the sensitivity.
Changes in global mean surface temperature due to a 2% increase in solar constant (table 4.1),
are within the range given by the IPCC (1.5 − 4.5◦ C). Note that a 2% increase in solar constant
is equivalent to a radiative forcing of 3.4W/m2 , which is very close to the radiative forcing of
3.7W/m2 expected from a doubling of atmospheric CO2 (Myhre et al., 1998). In the experiment
with an active ice sheet, all the ice melts in response to a 2% increase in solar constant. Once
the ice has completely melted, the sensitivity of the system is reduced. Because of this, there is
a significant difference between the response of climate to an increase, versus a decrease in solar
constant. In effect, modern climate is most sensitive to decreases in solar constant, or CO2 if the
perturbation is sufficiently large.
4.3 Ice Sheet Stability
Since the early model studies by Budyko (1969) and Sellers (1969), ice-albedo has been recognized
as one of the most important feedback mechanisms determining the sensitivity of climate. At
present, perennial ice with a high albedo exists at both poles. If the climate is cooled, these ice
areas will grow, leading to a higher planetary albedo and a further cooling. This strong positive
65
feedback mechanism enhances climate sensitivity, and has the potential to cause multiple climate
equilibria. According to North et al. (1981) these multiple equilibria are associated with two types
of instabilities:
1. Small Ice Cap Instability (SICI); where if an ice cap decreases sufficiently in size it becomes
unstable and rapidly melts.
2. Ice Covered Earth Instability (ICEI); in this case an instability arises when the ice cap grows
too large, and the climate moves to a state where the entire globe is covered with ice.
In studies with simple one dimensional energy balance models where the ice-albedo feedback is
parameterized by fixing the ice limit to the zero degree isotherm, it has been shown that there are
three possible solutions for the present level of solar forcing: one with a small ice cap, one with a
large ice cap, and one completely ice covered. This type of multiple equilibria, except for the small
ice cap instability, has even been shown to exist in some GCM experiments (Lee and North, 1995).
The small ice cap instability has been proposed as a mechanism for the initiation and growth of
the Greenland and Antarctic ice sheets and for the oscillations in size of the Laurentide ice sheet
(North and Crowley, 1985). However, the small ice cap instability only exists in energy balance
models where the meridional heat flux is represented by linear diffusion (North, 1984). In this
case, the model has no stable ice cap smaller than a certain finite size. In the model presented here,
meridional atmospheric heat flux is nonlinear, and the small ice cap instability is not expected to
occur.
The ice covered earth instability is not only found in simple energy balance models, but also
in radiative convective equilibrium models (Wang and Stone, 1980), as well as in an atmospheric
GCM (Lee and North, 1995). In this case there are two possible solutions for the same solar forcing:
one corresponding to an ice covered earth, and the other with a small ice cap. A small perturbation
in insolation from the present value can cause the system to rapidly move to the ice covered state.
However, once ice covered, the solar constant has to be increased to as much as 40% above its
present value before ice starts to melt (Crowley and North, 1991, p. 17). This observation is the
66
origin of the “faint young sun paradox”: over the past several billion years the sun’s luminosity
has been steadily increasing, therefore the question is raised why the earth is not completely ice
covered. One explanation often cited is that the concentration of CO2 in the atmosphere was much
greater in the past and therefore prevented the initial build up of ice (Owen et al., 1979).
4.3.1 Temperature Dependent Albedo
To test the stability and possible existence of multiple equilibria in the atmosphere-ice process
model, experiments are constructed where the solar constant is increased from 80% to 120% of
today’s value. The first set of experiments are carried out without an active ice sheet. Instead, the
ice-albedo feedback is represented by a varying snow line over land which follows the latitude of
the zero degree isotherm. Figure 4-7 shows the annual mean equilibrium latitude of the snow line
as a function of normalized solar constant (S/S0 ). For large values of the solar constant (15% above
modern) there is no snow in the model. As S decreases, snow gradually covers the surface until it
is completely snow covered. For values of S smaller than about 0.86S0, the earth is covered with
snow and the global mean temperature is below −20◦ C. At the present value of S, annual mean
snow cover reaches about 60◦ N, with a global mean temperature of 14◦ C, and a planetary albedo
of about 0.30.
Once in the completely snow covered state, increasing the solar constant (analogous to increasing atmospheric CO2 ) starts melting snow when S reaches a value of about 0.86S0 . Thus, for the
range of S tested there is only one equilibrium solution, and there is no hysteresis loop as found
in similar experiments with one dimensional energy balance models (Budyko, 1969; Sellers, 1969;
North, 1975). There is no evidence of an ice covered earth (ICEI), or small ice cap instability
(SICI), and a relatively small value of S is required to escape from the completely ice covered
situation.
The seasonal range in snow cover is also sensitive to changes in solar constant, as can be seen in
figure 4-8. Here, winter and summer extremes of snow cover are shown as functions of normalized
solar constant for the experiment where insolation is gradually reduced. For a solar constant larger
67
S decreasing
S increasing
0.85
0.9
0.95
b)
50
40
30
20
10
0
−10
−20
−30
−40
−50
0.8 0.85 0.9 0.95 1 1.05 1.1 1.15 1.2
S/S0
1
S/S0
1.05
0.5
Global Mean Albedo
°
Snowedge Latitude ( )
°
Mean Temperature ( C)
a)
90
80
70
60
50
40
30
20
10
0
0.8
1.1
1.15
1.2
c)
0.4
0.3
0.2
0.8 0.85 0.9 0.95 1 1.05 1.1 1.15 1.2
S/S0
Figure 4-7: a) Annual mean snow edge latitude (degrees), b) global annual mean temperature (◦ C),
and c) global annual mean planetary albedo as a function of normalized solar constant (S/S0 ) in an
experiment without an active ice sheet and where δαsnow = 0.186. Two experiments are plotted:
one where S/S0 is increasing, and one where S/S0 is decreasing. However, the results from the
two cases are indistinguishable.
68
Latitude (degrees)
90
80
70
60
50
40
30
20
10
0
0.8
Winter Maximum
Summer Minimum
0.85
0.9
0.95
1
S/S
1.05
1.1
1.15
1.2
0
Figure 4-8: Seasonal snow cover limits as a function of normalized solar constant (S/S0 ) in an
experiment with a temperature dependent snow cover (SNOW).
than 0.99S0 there is no snow in the summer months, whereas insolation has to be increased to about
1.16S0 for all months of the year to be free from snow. At present, snow cover varies from about
90◦ N in summer to 40◦ N in winter. The latitude of permanent, or summer minimum snow cover,
is more sensitive to changes in solar constant than the maximum winter snow cover. In summer,
the meridional surface temperature gradient is weaker than in winter. Therefore, a change in global
temperature can cause a large change in the meridional extent of summer snow cover, whereas the
change in winter snow cover is relatively small. This agrees with the results of Held and Suarez
(1974), who find that for an annual mean climate model, the sensitivity of the snow line is inversely
proportional to the temperature gradient near the snow line.
In the one dimensional energy balance model of North et al. (1981) an ice covered earth instability (ICEI) occurs if the solar constant is lowered by just a few percent from its present value,
and does not recover again until S is increased to about 40% above its current value. However,
the change in albedo from snow-free to snow-covered conditions (δαsnow = 0.32) is significantly
larger than what is used here (δαsnow = 0.186).
Figure 4-9 shows the result of increasing the strength of the snow-albedo feedback to a value
of δαsnow = 0.40. An ice covered earth instability sets in when S is lowered to about 0.94S0 .
For smaller values of the solar constant, all land surfaces are covered with ice. Once in the snow
covered state, increasing the solar constant does not result in melting of snow until S reaches a value
69
of about 0.99S0 . Thus, there are two equilibrium states when the solar constant is between 0.94S0
and 0.99S0: one with a relatively small snow cover, and one where the planet is completely snow
covered. The resulting hysteresis loop is similar to that found in experiments with one dimensional
energy balance models, however there is no small ice cap instability (SICI). At the same time, the
snow cap is more stable to decreases in S and does not require a very large increase in S in order
to escape from the completely ice covered state.
Figure 4-9b shows global mean sea level temperature versus normalized solar constant for the
experiment with δαsnow = 0.40. This illustrates that climate sensitivity is greater for the state with
relatively small snow cover (upper branch) compared to the completely snow covered state (lower
branch). The sensitivity β, is approximately 290◦ for the state with partial snow cover, whereas
it is only 110◦ in the completely snow covered state. Sensitivity in the ice covered state is lower
because the ice albedo feedback becomes insignificant when the area of ice remains unchanged.
4.3.2 Inclusion of an Active Ice Sheet
The simple albedo feedback mechanism adopted in the preceding section and in most previous
studies with energy balance models is far from realistic: ice caps are treated as snow caps with
instantaneous response to temperature, ignoring any effects of accumulation and ablation rates. In
the following experiment the temperature dependent snow line is replaced by an active ice sheet
which responds to changes in the balance of accumulation and ablation.
Figure 4-10 shows the response of the ice sheet to changes in solar constant. There is no ice
present for values of S above 1.01S0 . As S decreases, the ice sheet gradually expands, and for
values of S below 0.80S0 nearly all land areas are covered by ice. However, as ice approaches the
equator, rates of accumulation and ablation are extremely small and the growth of the ice sheet is
very slow. Once in the ice covered state, increasing the solar constant does not result in a retreat of
the ice sheet until S reaches a value of about 0.87S0 . Thus, there are two equilibrium states when
S is sufficiently small. There is no rapid switch to a completely ice covered state as implied by the
ice covered earth instability. However, it should be kept in mind that accumulation is given by the
70
S decreasing
S increasing
0.85
0.9
0.95
b)
50
40
30
20
10
0
−10
−20
−30
−40
−50
0.8 0.85 0.9 0.95 1 1.05 1.1 1.15 1.2
S/S
1
S/S0
1.05
0.5
Global Mean Albedo
°
Snowedge Latitude ( )
°
Mean Temperature ( C)
a)
90
80
70
60
50
40
30
20
10
0
0.8
1.1
1.15
1.2
c)
0.4
0.3
0.2
0.8 0.85 0.9 0.95 1 1.05 1.1 1.15 1.2
S/S
0
0
Figure 4-9: a) Annual mean snow edge latitude (degrees), b) global annual mean temperature (◦ C),
and c) global annual mean planetary albedo as a function of normalized solar constant (S/S0 ) in an
experiment without an active ice sheet and where δαsnow = 0.40.
71
b)
40
30
20
10
0
−10
−20
−30
−40
0.8
S decreasing
S increasing
0.85
0.9
0.95
S/S
1
0
0.5
Planetary Albedo
°
Latitude of Ice edge ( )
°
Global Mean Temperature ( C)
a)
90
80
70
60
50
40
30
20
10
0
0.8
0.9
S/S0
1
1.1
c)
0.4
0.3
0.2
0.8
1.1
1.05
0.9
S/S0
1
1.1
Figure 4-10: a) Annual mean ice edge latitude (degrees), b) global annual mean temperature (◦ C),
and c) global annual mean planetary albedo as a function of normalized solar constant (S/S0 ) in an
experiment with an active ice sheet.
parameterized moisture flux which is not accurate at low latitudes.
As the solar constant increases, the ice sheet retreats until it is completely melted for values of S
above 1.01S0 . As in the experiments with a temperature dependent snow line, there is no evidence
suggesting that a small ice sheet is unstable. There is a very small difference between the increasing
and decreasing branches of the experiment when the ice sheet is small (figure 4-10a). However,
this difference diminishes as the rate of change in solar constant is reduced, and is thought to be
insignificant. In the experiment shown, the rate of change in S is fixed at 1% every 500 Ka.
4.3.3 Parameterized Sea Ice
The experiments in the previous sections only take into account the effects of an ice sheet and
temperature dependent snow cover over land surfaces. The effect of an interactive sea ice cover
72
no snow active seaice
/no ice
/no ice
◦
Sensitivity parameter β ( )
138
187
◦
∆T when S + 2% ( C)
+2.8
+3.7
∆T when S − 2% (◦ C)
−2.8
−3.8
active seaice
/active snow
282
+4.6
−5.8
Table 4.3: Sensitivity (β) of different model configurations with and without active snow, ice, and
sea ice cover. In all four experiments the land fraction is set to 40%.
is not considered, as it requires an appropriate representation of the thermal mass and circulation
in the ocean. Nevertheless, with these caveats in mind, an experiment with parameterized sea
ice is briefly discussed here. In this simplified case, latitudes where the seasonal mixed layer
temperatures are below freezing are considered to be covered by sea ice. The model does not
account for any variations in the thickness of the sea ice, and as for snow and ice cover over land,
the albedo of surfaces covered by sea ice is increased by δαseaice = 0.186.
Table 4.3 gives a summary of climate sensitivity for different model configurations when including an active sea ice cover. The experiment where both sea ice and snow cover is active is
closest to previous studies with energy balance models without an interactive ice sheet (table 4.2).
As expected, the sensitivity is increased when sea ice is included, as compared to the experiment
with only an active snow cover in table 4.1. However, the sensitivity in the case with both an active
sea ice and snow cover (β = 282◦ ) is comparable to the sensitivity found when increasing the snow
albedo feedback to δαsnow = 0.40 (β = 290◦; section 4.3.1).
To test the effect of including both a temperature dependent sea ice and snow cover on the
stability of the model, the experiment in figure 4-7 on page 68 is rerun. Again, as can be seen
in figure 4-11, the result is similar to that obtained when increasing the strength of the ice albedo
feedback to δαsnow = 0.40 (figure 4-9 on page 71), although the hysteresis loop is slightly wider.
Investigating the seasonal snow and sea ice cover as a function of normalized solar constant
(figure 4-12), in the case when the solar constant is decreasing, shows that the seasonal cycle of
sea ice is too small. At the same time, the sea ice cover extends to about 55◦ N in summer for the
current climate. This is partly due to the simple mixed layer ocean used in the model, and indicates
73
b)
90
80
70
60
50
40
30
20
10
0
0.8
S decreasing
S increasing
0.85
0.9
0.95
1
S/S
1.05
1.1
1.15
1
S/S0
1.05
1.1
1.15
0
0.85
0.9
0.95
c)
50
40
30
20
10
0
−10
−20
−30
−40
−50
0.8 0.85 0.9 0.95 1 1.05 1.1 1.15 1.2
S/S0
0.6
Global Mean Albedo
°
Snow edge Latitude ( )
°
Sea Ice Latitude ( )
°
Mean Temperature ( C)
a)
90
80
70
60
50
40
30
20
10
0
0.8
1.2
d)
0.5
0.4
0.3
0.2
0.8 0.85 0.9 0.95 1 1.05 1.1 1.15 1.2
S/S0
Figure 4-11: a) Annual mean snow edge latitude (degrees), b) annual mean sea ice latitude (degrees), c) global annual mean temperature (◦ C), and d) global annual mean planetary albedo as a
function of normalized solar constant (S/S0 ). The experiment is run without an active ice sheet, but
with snow cover over land and sea ice over ocean surfaces, both with δαsnow = δαseaice = 0.186.
74
°
Snow edge Latitude ( )
°
Sea Ice Latitude ( )
a)
90
80
70
60
50
40
30
20
10
0
0.8
b)
90
80
70
60
50
40
30
20
10
0
0.8
Winter Maximum
Summer Minimum
0.85
0.9
0.95
1
S/S0
1.05
1.1
1.15
1.2
Winter Maximum
Summer Minimum
0.85
0.9
0.95
1
S/S
1.05
1.1
1.15
1.2
0
Figure 4-12: Seasonal a) snow, and b) sea ice cover limits as a function of normalized solar constant
(S/S0 ). The experiment is run without an active ice sheet, but with snow cover over land and sea
ice over ocean surfaces, both with δαsnow = δαseaice = 0.186.
75
that an appropriate representation of the deep ocean as well as an interactive ocean heat transport
is required in order to accurately model sea ice.
4.4 Summary
A comparison of simulated and observed climate shows that the model does a reasonable job at
estimating the present seasonal cycle and atmospheric fluxes of sensible and latent heat. However,
a good apparent fit between simulated and observed climate is no guarantee that the response of the
model to changes in external forcing, such as insolation, is realistic. Therefore, the sensitivity of the
model to changes in solar constant is calculated and found to lie within the range β = 138 − 219◦ .
The lowest sensitivity is found in an experiment without an active ice-albedo feedback, and the
largest is found in an experiment with an active ice sheet. When parameterizing the ice-albedo
feedback by fixing the snow line to the latitude of the 0◦ C isotherm, the sensitivity is β = 166◦ .
This is significantly lower than values found in past studies with energy balance models, partly
due to the improved albedo parameterization used in this study, and the increased strength of the
negative dynamical flux feedback. When varying the solar constant within a range from 80% to
120% of today’s value, the ice sheet is found to be relatively stable without any indications of
the existence of multiple equilibria. However, when the strength of the snow-albedo feedback is
increased, or when a simplified representation of sea ice is included, the sensitivity of the model
is increased and multiple equilibria exist, in agreement with previous studies with energy balance
models.
76
Chapter 5
Orbital Insolation and Ice Sheet Mass
Balance
Changes in the solar constant result in a global change in insolation comparable at all latitudes
and seasons. The only orbital parameter which gives a similar uniform change in insolation is
eccentricity. However, the magnitude of the change in insolation due to variations in eccentricity
is negligible. On the other hand, variations in obliquity and precession have a significant impact
on the spatial and temporal distribution of insolation. Obliquity controls the relative amount of
insolation received at low and high latitudes, and the main effect of precession is to redistribute
insolation between summer and winter. As a result, the sensitivity of climate to changes in the
orbital parameters is significantly different from changes in the solar constant and will be discussed
in the following sections.
5.1 Obliquity
To test the sensitivity of the model climate to obliquity variations, the orbit is kept perfectly circular
with eccentricity e = 0, thereby excluding the influence of variations in precession. Obliquity is
varied within the range 21◦ to 25◦ , where today’s value is about 23.5◦. The change in seasonal
insolation due to a 4◦ reduction in obliquity (21◦ − 25◦ ) is shown in figure 2-2 on page 14: high
77
Temperature Anomaly (K)
1
0
−1
−2
−3
0
CNTR
SNOW
ICE
10
20
30
40
50
Latitude (degrees)
60
70
80
90
Figure 5-1: Change in annual mean meridional temperature structure of the model produced by a
4◦ decrease in obliquity (21◦ − 25◦ ) for the case without snow or ice (CNTR), with a temperature
dependent snow albedo (SNOW), and with an active ice sheet, but no snow (ICE).
latitude summertime insolation decreases, whereas mid latitude wintertime insolation increases.
To test the impact of changing obliquity on climate, three different model experiments are
constructed as follows:
(CNTR)
snow line is fixed at the latitude of the Arctic ocean (75◦ N), and there is no ice sheet;
(SNOW)
seasonal snow cover over land follows the latitude of the 0◦ C isotherm;
(ICE)
snow line is fixed at the latitude of the Arctic ocean, and an active ice sheet is included.
The difference in annual mean meridional temperature structure produced by a 4◦ decrease in
obliquity from 25◦ to 21◦ is shown in figure 5-1 for the three different model experiments. Annual
and zonal mean temperature is increased at low latitudes and reduced at high latitudes in the control
experiment (CNTR) in response to the reduction in obliquity. This result does not change much
when including an active snow cover (SNOW). However, in the experiment with an active ice
sheet (ICE), the increase in meridional temperature gradient is amplified. In this case, tropical
temperatures remain unchanged, whereas temperatures at high latitudes cool significantly when
obliquity is reduced.
Figures 5-2a and 5-2b show the impact of obliquity variations on seasonal snow limits and ice
cover over land in experiments SNOW and ICE, respectively. As found earlier when testing the
78
a) SNOW
90
Summer Minimum
Winter Maximum
80
Ice Latitude (°)
°
Snow Latitude ( )
90
70
60
50
40
30
21
b) ICE
80
70
60
50
40
22
23
°
Obliquity ( )
24
30
21
25
22
23
°
Obliquity ( )
24
25
Figure 5-2: a) Seasonal snow cover limits as a function of obliquity in the experiment without
an active ice sheet (SNOW). b) Latitude of the ice sheet margin as a function of obliquity for the
experiment with an active ice sheet, but no snow (ICE).
response to changes in solar constant (section 4.3.1), the sensitivity of the minimum extent of snow
in summer, is larger than the sensitivity of maximum snow extent in winter. Because minimum
summer snow latitude represents the position of perennial, or permanent snow cover, it is often
used as an indication of the position of the ice margin in climate models without an active ice
sheet(e.g. Budyko (1969); Sellers (1969); North (1975); Suarez and Held (1979); Schneider and
Thompson (1979)). In experiment SNOW, the permanent snow line retreats from 80◦ N to 90◦ N
as obliquity is increased from 21◦ to 25◦ . The maximum winter snow limit remains relatively
unchanged close to 40◦ N, with a slight retreat when obliquity is low.
The response of the ice sheet is larger than that observed for the snow line. In experiment ICE,
the margin of the ice sheet retreats from 58◦ N to 75◦ N, as obliquity is increased. This suggests
that an active ice sheet with an explicit formulation of the surface mass balance is more sensitive to
changes in obliquity than a temperature dependent snow line.
The response of climate to obliquity variations can be further understood by investigating the
change in seasonal temperature structure. Figure 5-3 shows the change in zonal mean temperature over one seasonal cycle in response to a 4◦ decrease in obliquity. The results of experiments
CNTR and SNOW are similar, with a decrease in high latitude summer temperatures and increase
in low latitude winter temperatures. Maximum cooling (∼ 3◦ C) occurs about 20 days after sum-
79
mer solstice. When including an active ice sheet, the temperature change is increased significantly:
summer temperatures at high latitudes are reduced by about 6◦ C, whereas the change at low latitudes is less pronounced.
There are several feedbacks influencing the sensitivity of the system to obliquity changes. In
experiment SNOW, a decrease in obliquity cools high latitude temperatures, resulting in the snow
line moving toward the equator. This increases land surface albedo and enhances the initial cooling.
However, a low obliquity also results in a steep meridional temperature gradient. This constrains
the movement of the snow line and stabilizes the system. The experiments with and without a
temperature dependent snow-albedo feedback (CNTR and SNOW) give similar results, suggesting
that the sensitivity of the temperature dependent snow line to obliquity changes is small.
In the case with an active ice sheet, the increase in temperature gradient when obliquity is low
enhances the transport of moisture to the ice sheet and therefore accumulation. At the same time,
low obliquity cools high latitudes, reducing ablation. As a result, the ice sheet grows and cools high
latitudes further. However, the expansion of the ice sheet is restricted by an increase in ablation
as the margin of ice moves toward the equator. This constraint on ice growth is weaker than that
imposed on the extent of the snow cover, and the ice sheet can expand beyond the zero degree
isotherm. As a result, the change in high latitude temperature when including an active ice sheet
is twice as large as when only including an active snow cover. This suggests that climate is more
sensitive to changes in obliquity when including an explicit formulation of ice sheet mass balance.
5.2 Precession
The following set of experiments are designed to test the sensitivity of the model climate to changes
in precession. Obliquity is kept constant at today’s value of ǫ = 23.439◦. Eccentricity is fixed at
a value of e = 0.030, which is about twice the current value, thereby increasing the amplitude of
variations in precession. Figure 5-4 shows the change in annual mean meridional temperature due
to a 180◦ change in longitude of perihelion (ω = −90◦ minus ω = +90◦ ) for the three different
80
−2
0
1
−1
2
0
b) SNOW
−2
−1
−3
0
−2
0
1
−1
2
−4
−6
−5
−4
0
−3
−1
−2
0
−3
−2
c) ICE
90
80
70
60
50
40
30
20
10
0
0
−1
0
−2
Latitude (degrees)
−1
0
Latitude (degrees)
1
1
Latitude (degrees)
−3
−2
0
90
80
70
60
50
40
30
20
10
0
a) CNTR
1
90
80
70
60
50
40
30
20
10
0
−1
60
120
180
240
Mean Orbital Longitude (degrees)
1
300
360
Figure 5-3: Change in zonal mean temperature of the model in response to a 4◦ decrease in obliquity
(21◦ − 25◦ ) for experiments a) without varying snow or ice covers (CNTR), b) with temperature
dependent snow albedo (SNOW), and c) with an active ice sheet, but no snow (ICE). The start of
the year is fixed at spring equinox and the year is divided into 360 degrees, where 90◦ is summer
solstice.
81
Temperature Anomaly (K)
1
0
−1
−2
−3
0
CNTR
SNOW
ICE
10
20
30
40
50
Latitude (degrees)
60
70
80
90
Figure 5-4: Change in annual mean meridional temperature structure of the model produced by a
180◦ change in the longitude of perihelion (ω = +90◦ minus ω = −90◦ ) for the case without snow
or ice (CNTR), with a temperature dependent snow albedo (SNOW), and with an active ice sheet,
but no snow (ICE).
model experiments described in the previous section (CNTR, SNOW, & ICE). At ω = −90◦ , perihelion coincides with winter solstice, giving warm winter and cool summer seasons. Conversely,
at ω = +90◦, perihelion coincides with summer solstice, giving cool winter and warm summer
seasons (figure 2.1 on page 14).
Figures 5-5a and 5-5b show the impact of changes in longitude of perihelion on seasonal snow
limits and ice cover over land in experiments SNOW and ICE, respectively. As opposed to the
previous set of experiments (figure 5-4), which gives the response to a fixed change in longitude
of perihelion, these two experiments give the equilibrium response of snow and ice cover at all
possible longitudes of perihelion.
As can be seen in figure 5-4, annual and zonal mean temperature remains unchanged in experiment CNTR regardless of the change in longitude of perihelion. In this experiment there is no
active snow- or ice-albedo feedback, and the effects of changes in longitude of perihelion on insolation cancel when averaged over an annual cycle. Therefore, the annual mean response is expected
to be negligible. On the other hand, seasonal variations in snow, or ice cover can result in an annual
mean temperature change. When including a temperature dependent albedo (SNOW) this response
is very small: there is a slight warming of mid-latitudes due to a small retreat of the winter snow
line and a slight cooling at high latitudes as the summer snow line advances (figure 5-5a).
82
90
Winter
Summer
70
60
50
40
30
−180 −120 −60
0
60 120
°
Longitude of Perihelion ( )
b) ICE
80
°
80
a) SNOW
Ice Latitude ( )
Snow Latitude (°)
90
70
60
50
40
30
−180 −120 −60
0
60 120
°
Longitude of Perihelion ( )
180
180
Figure 5-5: a) Seasonal snow cover limits as a function of longitude of perihelion (ω) in the experiment without an active ice sheet (SNOW). b) Latitude of the ice sheet margin as a function of ω
for the experiment with an active ice sheet, but no snow (ICE).
In the experiment with an active ice sheet (ICE) the response of annual mean temperature to
a change in longitude of perihelion is significantly larger, and increases with latitude (figure 54). This is caused by an expansion of the ice sheet from 75◦ N to about 60◦N when ω = −90◦ ,
and perihelion coincides with winter solstice (figure 5-5b). On the other hand, when perihelion
coincides with summer solstice (ω = +90◦ ), the ice sheet is completely melted.
Figure 5-6 gives the change in zonal mean meridional temperature over one seasonal cycle in
response to a 180◦ change in longitude of perihelion, going from an orbital configuration where
perihelion coincides with summer solstice (ω = +90◦ ) to a situation where it coincides with winter
solstice (ω = −90◦ ). The results for experiments CNTR and SNOW are similar, with a decrease
in summer temperatures at all latitudes, and a slight increase in temperatures during the rest of the
year. When including an active ice sheet, cooling is enhanced at high latitudes during summer. The
maximum cooling is about the same magnitude as obtained in the experiment with a 4◦ change in
obliquity. However, here the expansion of the ice sheet is dominated by a reduction in melting,
rather than an increase in accumulation as observed when reducing obliquity.
83
c) ICE
90
80
70
60
50
40
30
20
10
0
0
−2
Latitude (degrees)
Latitude (degrees)
90
80
70
60
50
40
30
20
10
0
Latitude (degrees)
a) CNTR
90
80
70
60
50
40
30
20
10
0
1
1
−2
2
0
2
3
2
−3
0
−1
−1
−3
1
1
2
−2
−3
−3
1
−1
0
2
0
2
−1
−4
−2
b) SNOW
1
−5
3
−1
−2
−5
−4
−2
−6
1
0
0
−3
0
−3
1
−4
−1
2
60
120
180
240
Mean Orbital Longitude (degrees)
300
360
Figure 5-6: Change in zonal mean temperature of the model in response to a 180◦ change in longitude of perihelion (ω = −90◦ minus ω = +90◦ ) for experiments a) without varying snow or ice
covers (CNTR), b) with temperature dependent snow albedo (SNOW), and c) with an active ice
sheet, but no snow (ICE). The start of the year is fixed at spring equinox and the year is divided
into 360◦ , where 90◦ is the summer solstice.
84
Obliquity (ε) Perihelion (ω) Eccentricity (e)
Glacial Inception
low
−90◦
high
Deglaciation
high
+90◦
high
Table 5.1: Optimal orbital configurations for growth (glacial inception) or decay (deglaciation) of
ice sheets in the Northern Hemisphere.
5.3 Last Interglacial and Glacial Inception
Combining the results of the previous two sections, it is possible to predict the orbital configuration
which is most conducive to growth, or decay of ice sheets in the Northern Hemisphere. The optimal
conditions are listed in table 5.1: a combination of low obliquity, perihelion at winter solstice, and
high eccentricity is preferential for the growth of ice. Low obliquity enhances poleward transport
of moisture during all seasons, and increases accumulation on the ice sheets. When perihelion
coincides with winter solstice the Earth is closest to the Sun in winter, resulting in warm winter and
cool summer seasons, with relatively low melting in summer. However, the impact of changes in
longitude of perihelion is only significant if eccentricity is high.
Optimal conditions for deglaciation is a combination of high obliquity, perihelion at summer
solstice, and high eccentricity. High obliquity reduces accumulation by weakening the meridional
temperature gradient. Perihelion at summer solstice results in cool winter and warm summer seasons, with relatively high rates of summer melt. Note that in the case of obliquity and eccentricity,
the same criteria for glacial growth or decay are valid in the Southern Hemisphere. However, the
effect of longitude of perihelion is opposite in the Southern Hemisphere. This means that variations in ice sheet mass dominated by obliquity, or eccentricity would be synchronous in both
hemispheres, whereas ice sheets dominated by precession would be of opposite phase in the two
hemispheres.
Two time periods which are often discussed in the literature are the last interglacial at about
126 Ka BP, and the last glacial inception at about 115 Ka BP (Gallimore et al., 1995; Khodri et al.,
2001). As shown in table 5.2 these two time periods have orbital configurations close to the optimal
conditions for growth and decay of ice. Eccentricity is almost at its maximum, and perihelion is
85
Time (Ka BP)
0
115
126
Obliquity (ε) Perihelion (ω) Eccentricity (e)
23.44◦
−78.67◦
0.017
22.45◦
−68.99◦
0.044
23.99◦
111.24◦
0.041
−4
−6
−5
0
−7
−9
−7
−5
−4
−8
0
−1
−2
−3
2
1
2
−3
−2
−1
−6
0
1
3
2
90
80
70
60
50
40
30
20
10
0
0
1
Latitude (degrees)
Table 5.2: Values of obliquity, longitude of perihelion, and eccentricity at the last glacial inception
(∼ 115 Ka BP) and last interglacial (∼ 126 Ka BP), in addition to the modern values (0 Ka BP).
60
120
180
240
Mean Orbital Longitude (degrees)
300
360
Figure 5-7: Difference in zonal mean temperature between the last interglacial (126 Ka BP) and
glacial inception (115 Ka BP) as a function of mean orbital longitude in experiments with an active
ice sheet, but no snow.
close to either winter or summer solstice. The difference in obliquity between the two time periods
is not optimal, but it is relatively low at glacial inception and high during the interglacial.
When applying the two orbital configurations in experiments with an active ice sheet but no
snow cover, the ice extends to 55◦ N at 115 Ka BP, whereas it is completely melted at 126 Ka BP.
The difference in zonal mean temperature between 126 and 115 Ka BP as a function of time of the
year is shown in figure 5-7. High latitude summer temperatures at 115 Ka BP are −9◦ C colder
than at 126 Ka BP, mostly in response to the difference in longitude of perihelion and the presence
of a relatively large ice sheet. The magnitude of the response is significantly larger than the −6◦ C
cooling found in the previous two sections where obliquity and precession are varied individually
(figures 5-3 & 5-6). However, it should be noted that the change in obliquity between the last
glacial inception and last interglacial is only about 1.5◦ . This is smaller than the possible range of
obliquity variations, suggesting that the response of the climate system can be significantly larger.
The seasonal variation in accumulation and ablation during the two time periods is shown in
86
0.08
Accumulation 115Ka
Accumulation 126Ka
Ablation 115Ka
Ablation 126Ka
m2/s
0.06
0.04
0.02
0
0
60
120
180
240
Mean Orbital Longitude (degrees)
300
360
Figure 5-8: Seasonal variation in accumulation and ablation at the last interglacial (∼ 126 Ka BP)
and glacial inception (∼ 115 Ka BP) as estimated by the model with an active ice sheet.
figure 5-8. At glacial inception (115 Ka BP) obliquity is low, and the meridional temperature
gradient is enhanced, resulting in an increase in accumulation. In contrast to precession, a change
in obliquity does not have a strong seasonal preference, and accumulation is increased at all times
of the year. At the same time, the earth is furthest from the sun at summer solstice giving relatively
cool summer and warm winter seasons. As a result, the maximum rate of ablation in summer is
reduced, whereas the length of the melt season is slightly increased.
During the interglacial period (126 Ka BP), obliquity is large and the meridional temperature
gradient is weak, thereby reducing accumulation. Perihelion is at a longitude of 111.2◦ , meaning
that the earth is close to the sun at summer solstice. In effect, the summer season is relatively warm
and ablation is high, whereas the melt season is short.
5.4 Summary
At equilibrium, changes in obliquity and precession have a relatively large impact on ice cover and
high latitude climate. However, the response of snow cover, as represented by the 0◦ C isotherm, is
small when averaged over an annual cycle. The change in snow cover is only significant in summer
when the temperature gradient is relatively weak.
In the case of the ice sheet, optimal conditions for ice growth are attained when: 1) obliquity
is low; 2) eccentricity is high; and 3) perihelion coincides with winter solstice. Low obliquity
87
strengthens the meridional temperature gradient, which in turn increases the flux of moisture to
high latitudes and the accumulation of ice. At the same time, low obliquity cools high latitudes,
reducing surface melt in summer. When perihelion coincides with winter solstice, the Earth is
closest to the Sun in winter, resulting in relatively warm winter and cool summer seasons. As a
result, summer surface melt on the ice sheet is reduced. This effect is particularly strong when
eccentricity is large.
It should be kept in mind that the experiments described here are all run to equilibrium. In this
case, the response can be relatively large, and it is not certain that the response to a time varying
forcing will be of similar magnitude. Although, the time scale for changes in orbital insolation is
longer than the response time of the atmosphere-ocean system, it is comparable to the response
time of large continental ice sheets and the underlying bedrock. However, the three criteria for ice
growth summarized above are thought to be valid for the time varying problem, and will be used in
the following chapter in at attempt to understand what physical mechanisms caused the observed
variations in ice volume in the late Pliocene and early Pleistocene.
88
Chapter 6
Late Pliocene and Early Pleistocene Glacial
Cycles
One of the strengths of the model developed in this study is its ability to perform long time integrations covering multiple cycles in precession, obliquity, and eccentricity. This chapter will
investigate the physics governing the oscillations in ice volume observed in the proxy record, with
emphasis on the glacial cycles of the late Pliocene and early Pleistocene. It should be kept in mind
that the model experiments described are not attempts at simulating past climate, rather they are
intended to test the sensitivity of the ice-atmosphere system to changes in orbital insolation. This
is also why the model is kept simple, such that it is possible to understand the underlying physics.
Before describing the model experiments, the following section briefly discusses the origin of
the large continental ice sheets believed to have played such a dominant role in the climate of the
past 3 Ma. This topic was the motivation of a study on the influence of the Central American
Seaway on changes in ocean circulation in the middle Miocene and early Pliocene (Nisancioglu
et al., 2003). A few of the conclusions from this study are discussed below, however the reader is
referred to appendix B for further details.
89
6.1 Origin of Northern Hemisphere Ice Sheets
The δ 18 O records of benthic and planktonic foraminifera indicate that temperatures in the ocean
decreased by about 15◦ C during the Cenozoic (67 − 0 Ma BP) (Miller and Fairbanks., 1985;
Crowley and North, 1991, p. 183). The early Pliocene (5.3 − 3.6 Ma BP), frequently called the
Pliocene warm period, can be seen as a pause in the general cooling trend, with high latitude sea
surface temperatures ∼ 5 − 8◦ C warmer than today (Dowsett et al., 1996; Crowley, 1996). The
general cooling trend continued in the late Pliocene (3.6 − 1.8 Ma BP) from about 3 Ma BP, and at
about 2.8 − 2.6 Ma BP there was a significant expansion of ice cover, in particular at high latitudes
of the Northern Hemisphere (Shackleton et al., 1984; Jansen et al., 1988; Raymo, 1994).
So far, no satisfactory explanation has been found for the cause of expansion of ice cover in
the late Pliocene, although it is one of the largest and most important changes in past climate (see
review by Raymo (1994)). Two recent hypotheses are that the gradual Cenozoic cooling trend (Philander and Fedorov, 2003), or the closure of the Indonesian seaway 4 − 3 Ma BP Cane and Molnar
(2001) cooled the east equatorial Pacific and reduced atmospheric heat transport from the tropics to
higher latitudes, stimulating global cooling and the eventual growth of ice sheets. Another school
of thought is that the expansion of ice is related to the closure of the Central American Seaway
(CAS) about 5 − 3 Ma BP (Kaneps, 1979; Keigwin, 1982; Stanley, 1995; Haug and Tiedemann,
1998; Driscoll and Haug, 1998). Before this time, the CAS was open, connecting the Pacific and
Atlantic basins with a seaway at about 10◦ N. This allowed for flow of relatively low salinity surface
water from the Pacific to the Atlantic, freshening the surface of the North Atlantic, and inhibiting
the production of deep water.
According to Haug and Tiedemann (1998), the closure of the CAS strengthened the Gulf Stream
and the transport of warm saline water to high latitudes of the North Atlantic. This increased the
production of North Atlantic Deep Water (NADW), leading to greater evaporative cooling of surface waters at high latitudes, and increased atmospheric moisture content. Combined with favorable orbital obliquity, the enhanced moisture content would facilitate a buildup of ice sheets in
the Northern Hemisphere. Driscoll and Haug (1998) propose a similar mechanism, involving en90
hanced freshwater delivery to the Arctic via Siberian rivers and formation of sea ice. An increase
in sea ice cover would have increased surface albedo, and insulated the high heat capacity of the
ocean from the atmosphere. Thus, the insulating effect of sea ice would have reduced the warming
expected at high latitudes from the increase in poleward heat transport by the ocean.
Early experiments with an ocean GCM suggest that the Atlantic meridional overturning circulation is severely reduced, or non existent when the CAS is open (Maier-Reimer et al., 1990).
As the CAS closes, the flow of relatively fresh water from the Pacific to Atlantic is cut-off, and
the modern overturning circulation is established with formation of NADW at rates comparable to
present. Similar experiments with the MIT ocean GCM indicate that significant amounts of NADW
are formed even when the CAS is open, and that the modern overturning circulation is established
at the time the CAS shoals to a depth of about 1000 m in the middle Miocene (Nisancioglu et al.,
2003; appendix B). At the time of the final closure of the CAS, the meridional overturning circulation is strengthened. As a result, total global poleward ocean heat transport is enhanced by about
10%, due mostly to a 30% increase in heat transport by the North Atlantic.
The predicted increase in heat transport by the North Atlantic at the time of the final closure
of the CAS is in agreement with the studies of Haug and Tiedemann (1998) and Driscoll and
Haug (1998). However, whether this event is related to enhanced growth of ice in the Northern
Hemisphere is not clear. The hypotheses of Haug and Tiedemann (1998) and Driscoll and Haug
(1998) imply that an increase in sea surface temperatures in the northern latitudes of the Atlantic
basin, produce an increase in precipitation and accumulation of snow on high latitude land surfaces.
As has been shown in this study, accumulation is not simply related to local temperature. Instead,
accumulation is governed by the poleward flux of moisture by the atmosphere, which is related to
both the local temperature and the strength of the meridional temperature gradient.
Further, if the climate system is at equilibrium, and planetary albedo and orbital configuration
remains unchanged, an increase in poleward heat transport by the ocean would imply a reduction
in poleward heat transport by the atmosphere (Stone, 1978). If this is the case, an increase in heat
transport by the North Atlantic as the CAS closed, would cause a reduction in heat transport by the
91
atmosphere (Nakamura et al., 1994). As a result, transport of moisture by the atmosphere would
decrease, thereby reducing accumulation, and the potential for growth of ice sheets.
Following these arguments it is unlikely that the closure of the CAS caused the onset of enhanced glaciation. As suggested by Berger and Wefer (1996), the predicted increase in poleward
ocean heat transport may have delayed the onset of Northern Hemisphere glaciation by several million years. In effect, the shoaling of the CAS probably had a profound effect on ocean circulation in
the Miocene, and the eventual closure could have been the cause of the early Pliocene warm period
(∼ 5 − 3 Ma BP). However, another explanation has to be found for the cause of the observed
expansion of ice cover ∼ 2.8 − 2.6 Ma BP.
6.2 Late Pliocene to Early Pleistocene Ice Volume
From the beginning of the period with enhanced glaciation in the late Pliocene, oscillations in
benthic δ 18 O followed cycles with a period of 41 Ka, and an amplitude about one third to one half
of the ∼ 100 Ka cycles observed in the late Pleistocene (figure 2-4 on page 19). Numerous studies
have tried to understand the large ∼ 100 Ka cycles in global ice volume (e.g. Weertman (1976);
Oerlemans (1980); Pollard (1982); Hyde and Peltier (1987); Gildor and Tziperman (2001)). However, very few have investigated the dynamics of the 41 Ka cycles discussed in this study. When
constructing model experiments which are to be run on orbital time scales (> 10 Ka) GCMs are
too slow, and it is necessary to parameterize several of the physical processes involved. In most
studies including a representation of ice sheets, the hydrological cycle is very crude and variations
in accumulation are only influenced by local temperature (e.g. Gallee et al. (1992); Deblonde et al.
(1992); Berger et al. (1999); Huybrechts and de Wolde (1999)). The model used in this study is
highly simplified, making it extremely efficient and capable of running experiments spanning several million years. At the same time, a physical representation of the dependence of accumulation
on atmospheric moisture transport is included.
In the following experiment, named ICE600 (table 6.1), the coupled atmosphere-ice model is
92
Name
ICE600
ACC600
COLD600
SNOW600
Description
Control experiment with an active ice sheet and no snow
Temperature dependent accumulation
Cold climate with a large ice sheet
Temperature dependent snow cover and no ice
Table 6.1: List of model experiments.
Ice Extent (degrees)
75
70
65
−1800
−1700
−1600
−1500
Time (Ka BP)
−1400
−1300
−1200
Figure 6-1: 600 Ka of ice sheet extent (1.8 − 1.2 Ma BP) calculated by using the orbital solution of
Laskar et al. (1993).
run for 600 Ka forced with insolation for the period from 1.8 to 1.2 Ma BP, calculated using the
orbital solution of Laskar et al. (1993). This period is taken to represent typical conditions of the
late Pliocene to early Pleistocene, and insolation reconstructions for any other time period within
the last 3 Ma will give similar results. As before, the area north of 75◦ N is permanently covered
by sea ice, and the ice sheet is restricted to land areas south of 75◦ N. The standard parameters for
the atmosphere and ice models are given in tables 3.1 (page 37) and 3.3 (page 50), respectively.
Figure 6-1 shows the change in meridional extent of the ice sheet with time. The latitude of
the ice margin varies from about 66◦ N at its maximum to 75◦ N, when it is completely melted.
The model does not calculate the zonal extent of the ice sheet. However, at its maximum extent
its meridional cross section is about 7 × 108 m2 . If, for simplicity, it is assumed that the ice sheet
has a fixed width of 6000 km, it’s volume is approximately 4 × 1015 m3 . Using an ice density
of ρice = 920 kg/m3 and assuming that the total area of the Earth covered by ocean is about
3.6 × 1014 m2 (71% of the Earth’s surface), the maximum ice sheet extent calculated by the model
93
80
60
40
20
0
0
100
0.02 0.04 0.06 0.08
Frequency (1/Ka)
b) Accumulation
80
60
40
20
0
0
150
Power Density
a) Ice Volume
Power Density
Power Density
100
0.02 0.04 0.06 0.08
Frequency (1/Ka)
b) Ablation
100
50
0
0
0.02 0.04 0.06 0.08
Frequency (1/Ka)
Figure 6-2: Power spectra of mean annual a) ice volume, b) accumulation, and c) ablation for the
period 1.8 − 1.2 Ma BP.
corresponds to a sea level lowering of about 10 m.
According to the δ 18 O record (figure 2-4 on page 19), oscillations in late Pliocene to early
Pleistocene ice volume could have contributed to a sea level change as large as 40 to 60 m. This is
significantly larger than the amplitude of changes in ice volume estimated by the model. However,
the model is highly simplified, and there are several important positive feedbacks in the climate
system, such as CO2 , sea ice, and vegetation, which have not been included. It is conceivable that
the amplitude of the oscillations in ice volume will be amplified in a model where these processes
have been included.
At the same time, it is quite possible that the large amplitude oscillations in global ice volume
observed in the proxy record cannot be explained as a direct response to variations in orbital insolation. Instead, it is likely that at least part of the oscillations in ice volume are a result of nonlinear
self-sustained, or stochastic processes. Both of these options have been suggested as explanations
for the larger amplitude ∼ 100 Ka glacial cycles observed in the last ∼ 0.8 Ma of the proxy record
(e.g. Kallen et al. (1979); Kominz and Pisias (1979)).
However, it should be kept in mind that it is not possible to ascertain from the δ 18 O ice volume
record what portion of the variations in δ 18 O observed in the period ∼ 2.8 − 0.8 Ma BP are due
to changes in temperature (see section 2.2 on page 17), or what portion is due to changes in ice
volume on Antarctica. This issue will be discussed later in section 7.4.
94
Power spectra of mean annual ice volume, accumulation, and ablation calculated by the model
for the period 1.8 − 1.2 Ma BP are given in figure 6-2. As can be seen for ice volume, the strongest
peak in the power spectrum is at the obliquity frequency (1/41 Ka). However, the difference
between the strength of the obliquity and precession signals in the record is insignificant, and the
response of ice volume to the two is approximately equal. As expected, there is no response of ice
volume at the 100 Ka eccentricity period, due to the negligible influence of eccentricity variations
on insolation.
In the case of ablation, there is a clear dominance of precession frequencies (1/19 Ka, 1/23 Ka),
with a negligible contribution by obliquity. This is due to the strong dependence of surface melt on
summer temperatures: melt only occurs during a few months in summer when temperatures on the
surface of the ice are above freezing. Such a preference for a particular season rectifies the annual
cycle and enhances the response to precession (Schneider and Thompson, 1979; Kim et al., 1998;
Huybers and Wunsch, 2003). If melt occurred at all times of the year, the response to precession
would be small, as it’s effect would cancel when averaged over an annual cycle.
The power spectrum of mean annual accumulation shows a slight dominance of obliquity, although precession frequencies are relatively strong. The influence of obliquity on accumulation
is due to the control of the meridional temperature gradient on transport of moisture to the ice
sheet. The precession signal in accumulation is introduced by the dependence of moisture transport on local temperature at the margin of the ice. Variations in local temperature are mostly due
to changes in precession, amplified by the ice-albedo feedback. However, temperature changes at
the ice margin are enhanced by the movement of the ice sheet. Therefore, any changes in latitude
of the ice margin due to variations in ablation will have an impact on accumulation, and strengthen
the precession signal.
In the experiments presented here the effect of an interactive sea ice cover is not considered, as
it requires a representation of the thermal mass and circulation of the ocean. However, as a first
test of the effect of sea ice, the parameterized sea ice cover described in section 4.3.3 on page 72 is
adopted, and experiment ICE600 is rerun. In this case, the results of figure 6-1 and the amplitude of
95
80
60
40
20
0
0
120
b) Accumulation
100
150
Power Density
a) Ice Volume
Power Density
Power Density
100
80
60
40
b) Ablation
100
50
20
0.02 0.04 0.06 0.08
Frequency (1/Ka)
0
0
0.02 0.04 0.06 0.08
Frequency (1/Ka)
0
0
0.02 0.04 0.06 0.08
Frequency (1/Ka)
Figure 6-3: Power spectra of mean annual a) ice volume, b) accumulation, and c) ablation for the
period 1.8 − 1.2 Ma BP in an experiment where accumulation only depends on local temperature.
the response of the ice sheet remains unchanged. This indicates that an increased climate sensitivity,
as defined by equation 4.1 on page 63, does not simply translate into an increased sensitivity of land
based ice sheet extent in response to varying orbital insolation.
6.2.1 Temperature Dependent Accumulation
The results of the previous experiment (ICE600) suggest that the impact of obliquity on variations
in ice volume are mostly a consequence of the dependence of accumulation on the meridional
temperature gradient. In order to test this observation, the experiment is repeated with the relation
between accumulation and temperature gradient removed. In this experiment, named ACC600,
accumulation only depends on local temperature at the ice sheet margin, and the gradient term in
equation 3.11 on page 45 is omitted. Except for the change in the parameterization of accumulation,
the model configuration is unchanged.
Power spectra of mean annual ice volume, accumulation, and ablation for the period 1.8 − 1.2
Ma BP are given in figure 6-3. The power spectrum of ablation is nearly identical to that found in
experiment ICE600 (figure 6-2). However, in the case of accumulation, the influence of obliquity
is significantly reduced. As a consequence, the impact of obliquity on variations in ice volume is
also reduced.
Although the obliquity signal in the power spectrum of accumulation is weakened, it is still
96
significant. This suggests that there is an additional source of variability at the obliquity period in
the accumulation record other than the meridional temperature gradient. This is most likely a result
of accumulation occurring at all times of the year. In this case, when averaging over a full annual
cycle, the response to precession cancels, enhancing the relative role of obliquity. Therefore, the
strong precession signal in the power spectrum of accumulation is a result of variations in the ice
margin imposed by changes in ablation.
6.2.2 Cold Climate with Large Ice Sheets
In the past 3 Ma, ice volume increases slowly in response to the general cooling trend dominating
throughout most of the Cenozoic, culminating at about 800 Ka BP when large ∼ 100 Ka oscillations
in ice volume become dominant (figure 2-4 on page 19). The dynamics of these large ice sheets
is likely to have been different from the smaller ice sheets present in the late Pliocene and early
Pleistocene. In particular, deglaciations in the last part of the record are relatively rapid, suggesting
that some strong non linear processes are involved in destabilizing an ice sheet when it has reached
a certain maximum size.
Although it is not within the scope of this study to understand the dynamics of the large ∼ 100
Ka oscillations in ice volume, it would be useful to understand how the surface mass balance depends on the size of the ice sheet. To investigate this, an experiment (COLD600) is constructed
where the climate is colder, and the mean size of the ice sheet is comparable to reconstructions of
the last glaciation maximum. To cool the climate, the value of the constant A in the parameterization of outgoing thermal radiation (equation 3.3 on page 37) is increased to A = 216 W/m2. As
before, the model is run for 600 Ka using orbital insolation from Laskar et al. (1993) for the period
1.8 to 1.2 Ka BP.
Figure 6-4 shows the latitude of the ice margin with time. The ice sheet extends to a latitude of
about 48◦ N, as compared to 66◦ N in experiment ICE600 (figure 6-1), and the annual mean global
temperature is reduced by about 5◦ C. The amplitude of the oscillations in ice extent is relatively
unchanged. However, the relative contribution by obliquity and precession is different. Power
97
Ice Extent (degrees)
60
55
50
45
−1800
−1700
−1600
−1500
Time (Ka BP)
−1400
−1300
−1200
Figure 6-4: 600 Ka of ice sheet extent (1.8 − 1.2 Ka BP) in an experiment with a colder climate
and a larger mean ice sheet extent.
80
60
40
20
0
0
100
0.02 0.04 0.06 0.08
Frequency (1/Ka)
b) Accumulation
80
60
40
20
0
0
150
Power Density
a) Ice Volume
Power Density
Power Density
100
0.02 0.04 0.06 0.08
Frequency (1/Ka)
b) Ablation
100
50
0
0
0.02 0.04 0.06 0.08
Frequency (1/Ka)
Figure 6-5: Power spectra of mean annual a) ice volume, b) accumulation, and c) ablation for the
period 1.8 − 1.2 Ma BP in an experiment with a colder climate and a larger mean ice sheet extent.
spectra of annual mean ice volume, accumulation, and ablation are given in figure 6-5. The relative
strength of the obliquity signal in the power spectrum of accumulation is slightly reduced, whereas
it is significantly reduced in ice volume.
One reason for the dominance of precession when the mean ice sheet extent is greater, is that
temperature at the ice edge is higher. The mean annual temperature at the ice edge is −2.9◦ C, as
compared to −8.7◦ C for the small ice sheet in experiment ICE600. This results in a significantly
larger ablation zone, and larger ablation oscillations dominating ice extent. Because ablation is
dominated by precession, the precession signal in ice volume is enhanced relative to obliquity.
98
Snow Extent (degrees)
90
80
70
Summer
Winter
60
50
40
30
−1800
−1700
−1600
−1500
−1400
−1300
−1200
Figure 6-6: 600 Ka of seasonal snow cover limits (1.8 − 1.2 Ma BP) calculated using the orbital
solution of Laskar et al. (1993).
6.3 Late Pliocene to Early Pleistocene Snow Cover
As has been seen in chapters 4 and 5, the response of snow and ice cover to insolation variations
differs significantly. In the following experiment (SNOW600), a version of the model without an
active ice sheet, but with active snow cover is forced with insolation variations for the period 1.8
to 1.2 Ma BP. Climate is cooled slightly to allow for snow lasting through the summer season by
setting A = 212 W/m2 in equation 3.3 on page 37, otherwise the standard parameters are used as
in experiment ICE600 and given in table 3.1 on page 37.
Figure 6-6 shows the change in latitudinal extent of winter, and summer snow cover as estimated
by the model. Winter snow extends to about 40◦ N and is less sensitive to insolation variations than
summer snow, as discussed in section 4.3.1 on page 67. The extent of summer, or perennial snow
cover varies between 90◦ N and 70◦ N.
Variations in both winter and summer snow extents are dominated by precession, with very
little influence by obliquity (figure 6-7). This is because snow cover depends on local seasonal
temperature, which is governed by precession. The influence of obliquity only becomes important
when averaging over a large part of the annual cycle, or when the system responds to latitudinal
contrasts in insolation, as is the case when an interactive accumulation is included.
99
a) Winter
100
50
0
0
150
Power Density
Power Density
150
a) Summer
100
0.02 0.04 0.06 0.08
Frequency (1/Ka)
50
0
0
0.02 0.04 0.06 0.08
Frequency (1/Ka)
Figure 6-7: Power spectra of a) winter maximum, and b) summer minimum snow cover for the
period 1.8 − 1.2 Ma BP.
6.4 Summary
When investigating the mechanisms controlling the cycles in ice volume of the late Pliocene and
early Pleistocene it would be useful to understand what caused the onset of enhanced glaciations
2.8−2.6 Ma BP. One theory, often discussed in the literature, is the closure of the Central American
Seaway which was completed ∼ 3 Ma BP. However, here it is argued that this event is more
likely to have caused the early Pliocene warm period, perhaps even delaying the onset of enhanced
glaciations.
In a model experiment forced with orbital insolation variations for the period 1.8 − 1.2 Ma
BP, oscillations in ice volume have a similar contribution by the main periods of precession and
obliquity. Nearly all variability at the obliquity frequency is due to the influence of the meridional
insolation gradient on accumulation, whereas the influence of precession is due to it’s control of
summer temperatures and surface melt.
All the experiments give an amplitude of oscillations in ice volume which is significantly lower
than what is estimated from the benthic δ 18 O record. However, there are several important positive
feedbacks in the climate system which have not been included in the model, such as CO2 , vegetation, and sea ice. Further, it is likely that at least part of the variability observed in the ice volume
record is a result of nonlinear self-sustained, or stochastic processes, which are not represented in
100
the model.
A model experiment where the ice sheet is removed and the ice-albedo feedback is represented
by a temperature dependent snow line illustrates that the two approaches are significantly different.
Assuming that the latitude of perennial snow cover is an indicator of changes in ice cover, as has
been done in previous model studies (e.g. Budyko (1969); Sellers (1969); North (1975); Schneider
and Thompson (1979); Suarez and Held (1979); Short et al. (1991)), is misleading. In such a
simplified approach, perennial snow cover only depends on summer insolation and precession.
Whereas in reality, snow and ice are influenced by accumulation of snow, which responds to the
latitudinal insolation contrast and obliquity.
101
Chapter 7
Summary and Discussion
Interest in simple Energy Balance Models (EBMs) was invigorated by the early work of Budyko
(1969) and Sellers (1969), who independently showed that the temperature-albedo feedback mechanism has the potential to greatly enhance the sensitivity of Earth’s climate. According to their
models, a decrease in the solar constant by only a few percent could trigger a transition to a completely ice covered planet. These early model results, combined with the concern for possible
human induced climate changes, have led to a large research effort focused on understanding past
as well as future climates.
However, some of the largest and most fundamental changes observed in the climate record are
still not understood. A few examples are as follows: it is not known what triggered the onset of
enhanced glaciations ∼ 2.8 − 2.6 Ma BP; why the initial oscillations in ice volume were dominated
by a period of 41 Ka; or why larger oscillations in ice volume with a duration of approximately
100 Ka dominated after ∼ 0.8 Ma BP. Knowledge about these past events is crucial in the effort
to understand the importance of different forcing and feedback mechanisms in determining the
stability of the climate system, as well as its future response to man made forcing.
In this work EBMs have several advantages over comprehensive coupled General Circulation
Models (GCMs): they are extremely efficient and can run multiple sensitivity experiments covering
long time periods; and they are simple, making it possible to isolate individual feedback mechan-
102
isms and understand the underlying physics. In particular, EBMs are well suited to study problems
where the spatial and temporal scales are large, compared with the spatial and temporal scales of
weather systems. In this context, the response to variations in orbital insolation is a good candidate.
As in most other studies with simple EBMs it is important to test the conclusions with a hierarchy of
climate models with varying degrees of complexity, including Earth Models of Intermediate Complexity (EMICs) and GCMs. With this in mind, the results presented here will serve as a valuable
guide for future experiments with more realistic models.
7.1 Summary
A summary of the main topics and results discussed in this study are as follows:
• Oscillations in ice volume before ∼ 0.8 Ma BP are dominated by a period of 41 Ka, which is
the main period of orbital obliquity. After this time, the amplitude of ice volume oscillations
increases by as much as 50%, the influence of precession increases, and there is a switch to a
dominance of periods close to 100 Ka.
• Based mainly on paleoclimate data from the last 0.8 Ka, most researchers compare, or even
tune, climate records to June or July 65◦ N insolation. However, summer insolation at 65◦ N
is governed by precession, and is not a good candidate to explain the 41 Ka period cycles
dominating ice volume before ∼ 0.8 Ma BP.
• As an alternative to current orbital theory, a hypothesis is tested, where obliquity dominated
variations in the meridional insolation gradient influences the poleward flux of moisture and
plays an important role in controlling high latitude ice volume before ∼ 0.8 Ma BP.
• A simple atmosphere-ice process model is developed to test this hypothesis. Two of the most
important differences between the new model and existing similar models is the inclusion of
an explicit parameterization for calculating atmospheric moisture flux and accumulation on
103
the ice sheet, and the adjustment of the ice-albedo feedback for the influence of solar zenith
angle and clouds.
• The model does a reasonable job at estimating the current seasonal cycle and atmospheric
fluxes of sensible and latent heat. The sensitivity (β) to changes in the solar constant is 138◦
with no snow or ice, 166◦ with a temperature dependent snow cover, and 219◦ with an active
ice sheet. The sensitivity in the experiment with a temperature dependent snow cover is
significantly lower than values found in the early EBM studies (Budyko, 1969; Sellers, 1969;
North, 1975), due to the improved albedo parameterization used, as well as the increased
strength of the negative dynamical flux feedback.
• Similarly, the multiple equilibria, observed in these previous model studies in response to
a varying solar constant, are only observed when the strength of the snow-albedo feedback
is increased to an unrealistically large value. However, it should be kept in mind that the
experiments presented here only allow for snow or ice cover over land, taking no account for
changes in sea ice.
• At equilibrium, changes in obliquity and precession have a relatively large impact on ice
cover and high latitude climate. Between the extreme values of obliquity and precession, the
ice sheet extent changes by as much as 15◦ of latitude, and the change in high latitude summer
temperatures is close to 6◦ C. In contrast, the response of a temperature dependent snow cover
is small, and only significant in summer when the temperature gradient is relatively weak.
• In the model version with an active ice sheet, optimal conditions for ice growth are attained
when: 1) obliquity is low; 2) eccentricity is high; and 3) perihelion coincides with winter
solstice. Low obliquity strengthens the meridional temperature gradient, which in turn increases the flux of moisture to high latitudes and the accumulation of ice. At the same time,
low obliquity cools high latitudes, reducing surface melt in summer. When perihelion coincides with winter solstice, the Earth is closest to the Sun in winter, resulting in relatively
warm winter and cool summer seasons, in particular when eccentricity is large. As a result,
104
summer surface melt is reduced.
• In a model experiment forced with orbital insolation variations for the period 1.8 − 1.2 Ma
BP, oscillations in ice volume have a similar contribution by the main periods of precession
and obliquity. Nearly all variability at the obliquity frequency is due to the influence of the
meridional insolation gradient on accumulation, whereas the influence of precession is due
to its control of summer temperatures and surface melt.
• The experiments forced with orbital insolation give an amplitude of ice volume oscillations
(∼ 10 m of mean sea level) which is significantly lower than what is estimated from the
benthic δ 18 O record of the late Pliocene to early Pleistocene (∼ 40 −60 m of mean sea level).
However, there are several important positive feedbacks which have not been included in the
model, such as CO2 , vegetation, and sea ice. At the same time, it is likely that at least part
of the variability observed in the ice volume record is a result of nonlinear self-sustained
oscillations, or stochastic processes.
• Assuming that variations in the latitude of perennial snow cover is an indicator of changes in
ice cover is misleading. In such a simplified approach, perennial snow cover only depends
on summer insolation and precession. Whereas in reality, snow and ice are influenced by
accumulation of snow, which responds to the latitudinal insolation contrast and obliquity.
7.2 Sea Ice
The main focus of this study is to understand processes related to atmospheric heat and moisture
transports, as well as snow and ice cover on land surfaces. To achieve this, the model is kept as
simple as possible, neglecting any influence by an interactive ocean. This implies that there is no
active sea ice included in the model. Instead, it is assumed that the Arctic ocean is permanently
covered by sea ice north of 75◦ N, whereas all other ocean surfaces are free of sea ice.
In order to add an interactive sea ice component to the model, it is necessary to include an
105
appropriate representation of circulation and heat transports by the ocean. One alternative is to
improve the mixed layer ocean, already existing in the model, by including a so-called Q-flux.
This is a fixed meridional ocean heat transport, with a magnitude adjusted such that the model
climate resembles the current observed climate. An alternative would be to use an ocean general
circulation model where the thermal mass of the deep layers as well as ocean circulation are taken
into account. However, it should be kept in mind that this would increase the computational cost of
the model significantly.
Adding sea ice to the system will introduce an additional strong positive feedback mechanism.
As high latitudes cool, sea ice expands, covering the surface ocean which has a relatively low
albedo, and causes a further cooling. In effect, sea ice amplifies the response of the system to
changes in high latitude insolation. The increased sensitivity of the system to a change in mean
annual insolation, or to the solar constant S, is expected to be similar to the effect of increasing the
strength of the snow-albedo feedback parameter (δαice , see section 4.3.1 on page 67). However,
there are several important differences between sea ice and snow cover. As opposed to snow cover,
sea ice is affected by ocean circulation and atmospheric wind patterns. In turn, sea ice has an impact
on surface salinity and the transfer of heat and water vapor between the atmosphere and ocean.
In the case of orbital forcing, there is another important distinction between sea ice and snow
cover. This is due to the different seasonal cycles of sea surface temperature, as opposed to temperature over land. Because of the large thermal mass of the ocean, the seasonal cycle of sea surface
temperature is significantly weaker than for surfaces temperatures on land. This implies that the
seasonal variation in sea ice extent is smaller than the seasonal variation in snow cover. As has been
discussed in the two previous chapters, a mean annual response to precession requires a strong seasonal response of a component in the climate system, such as surface albedo. However, the weak
seasonal cycle in sea ice suggests that its response to precession should be weaker than that of snow
cover over land.
Another important aspect of sea ice, is its effect on the atmospheric hydrological cycle. About
20% of precipitation on Greenland today comes from local sources such as the Norwegian and
106
Greenland seas (Charles et al., 1994), suggesting that a large change in sea ice cover could have
an impact on the supply of moisture to the ice sheet. This link between sea ice cover and moisture
supply to high latitude land surfaces is the basis of the so-called precipitation-temperature feedback
(Kallen et al., 1979; Letreut and Ghil, 1983; Gildor and Tziperman, 2001). The precipitationtemperature feedback implies that reduced sea ice cover and warm temperatures at high latitudes
increase the amount of evaporation from the surface ocean, as well as the moisture holding capacity
of the atmosphere. From this it is inferred that accumulation on land will increase.
The parameterization of atmospheric moisture transport used in this study takes into account
the effect of variations in local temperature as given by the Clausius-Clapeyron relation (equation 3.11 on page 45). Therefore, any variations in high latitude temperature will have an impact
on atmospheric moisture content. However, as has been shown in this study and discussed by Nakamura et al. (1994) and Scott (1995), changes in meridional atmospheric temperature gradient
dominate in controlling the atmospheric moisture transport and accumulation on land. Further, the
empirical study by Scott (1995) (figure 4-5 on page 61), shows that the correlation between the
observed seasonal atmospheric moisture transport and calculated transport using the parameterization adopted in this study is very good. Suggesting that for the range of temperature changes
experienced in a seasonal cycle (∼ 15◦ C), the assumption that the temperature gradient dominates
in controlling moisture transport is valid. For reference, the estimated warming from average last
glacial conditions to the Holocene is approximately 15◦ C (Cuffey et al., 1995; Cuffey and Clow,
1997), whereas during the late Pliocene, early Pleistocene the amplitude of the glacial cycles is
smaller.
It should be noted, that an important assumption made in deriving the parameterized moisture
flux is that relative humidity (qr ) remains constant. It is conceivable that large changes in sea ice, as
might have occurred during the glacial cycles of the past, could have resulted in a change in relative
humidity. Therefore, the possibility that accumulation on the Laurentide and Fennoscandian ice
sheets is influenced by changes in sea ice cover should be thoroughly tested with an appropriate
coupled atmosphere-ocean GCM with an interactive sea ice component. Although, existing model
107
studies indicate that the main moisture source regions for Greenland and Antarctica are tropical
and mid-latitude ocean regions below ∼ 50◦ of latitude (Charles et al., 1994; Werner et al., 2001),
where there is thought to have been little sea ice even at the last glacial maximum (Sarnthein et al.,
2003).
There is very little data on accumulation rates in the past. However, a study of ice cores from
central Greenland by Cuffey and Clow (1997) find that long-term averages (0.5 − 1.0 Ka) of estimated accumulation rates and temperature are negatively correlated during the past 7 Ka of the
Holocene. Whereas, on shorter time scales (0.1 Ka), accumulation rates and temperature are essentially uncorrelated. Based on these observations, the authors conclude that accumulation rate is
not a reliable proxy for temperature, and that there is no evidence supporting the inference that a
climatic warming results in increased precipitation over the Greenland ice sheet.
A study by Kaspner et al. (1995) find that atmospheric dynamics, not temperature, appears to
have been the primary control on snow accumulation in central Greenland over the past 18 Ka.
The sensitivity of accumulation to temperature changes during both warm (Holocene) and cold
(Younger Dryas, Last Glacial Maximum) periods is less than expected if the accumulation rate is
only controlled by the ability of warmer air to deliver more moisture to the ice sheet. At the same
time, accumulation during the transitions between these warm and cold periods varies more than
can be explained thermodynamically.
Similarly, a multivariate regression analysis of Greenland accumulation rates by van der Veen
et al. (2001) finds that about 80% of the total variance in longer-term averaged accumulation can
be explained by the large scale atmospheric circulation and its interaction with the geometry of the
ice sheet. Further, Bromwich and Robasky (1993) show that accumulation in Antarctica is largely
dominated by atmospheric circulation, with little influence of local temperature variations.
Interestingly, in the study of Kaspner et al. (1995) the sensitivity to temperature is found to
be largest during transitions from cold to warm periods. This supports the suggestion that the
sensitivity of surface melt (βabl ) to temperature is increased during times of ice sheet retreat (section 3.3.2 on page 51), due to the lower albedo of the snow free ice surface (Braithwaite, 1995;
108
Braithwaite and Zhang, 2000). A different ablation rate for growing and shrinking ice sheets could
lead to asymmetric glacial cycles, as is observed for the ∼ 100 Ka cycles of the past ∼ 0.8 Ma.
To summarize, there is no historical evidence for a strong link between accumulation of snow
and atmospheric temperature. Rather, the data suggests that accumulation is dominated by large
scale atmospheric circulation. Hence, the assumption that accumulation is controlled solely by
temperature, as used in model studies of changes in past, as well as future ice sheet mass balance
(e.g. Pollard (1980); Huybrechts (1990); Gallee et al. (1992); Deblonde et al. (1992); Tarasov
and Peltier (1997); Huybrechts and de Wolde (1999)), is questionable. However, it is possible that
large variations in sea ice could have influenced atmospheric relative humidity and the supply of
moisture to the ice sheets.
7.3 The Topographic Effect of an Ice Sheet
In the experiments described in this study the lower boundary of the atmosphere has been assumed
to be flat. However, in reality the continental ice sheets represent a significant topographic feature
once they advance to their maximum size. As an example, at the last glacial maximum, the dome
of the Laurentide ice sheet reached a maximum altitude of about 3000 m above sea level (Peltier,
1994). Although ice volume during the late Pliocene, early Pleistocene was significantly lower than
at the last glacial maximum (figure 2-4 on page 19), the presence of the ice sheet could potentially
have had an important effect on the thermal structure and circulation of the atmosphere.
Two possible mechanisms where the topography of an ice sheet might interact with the dynamics of atmospheric heat transport will be briefly discussed here. One mechanism involves the
impact of a sloping lower boundary on the stability of baroclinic eddies, and the other relates to the
topographic forcing of stationary eddies.
Under certain circumstances a sloping lower boundary can be compared to the β-effect which
stabilizes the longer wavelengths (Charney, 1947). According to Pedlosky (1964a,b), the necessary
condition for instability demands that at the lower boundary, the northward slope of the surfaces of
109
constant potential temperature exceed the northward slope of the boundary. Therefore, a sufficiently
strong topographic slope can have a stabilizing effect. The relevance of this mechanism to the ice
sheets of the late Pliocene, early Pleistocene requires that the surface slope is relatively steep, and
that the meridional extent is large enough to have an affect on the baroclinic eddies.
To estimate the surface slope of a typical late Pliocene, early Pleistocene ice sheet, the maximum volume of ice contained in the Laurentide ice sheet is taken to be about 50% of that estimated
for the last glacial maximum. This gives a maximum ice thickness of ∼ 1800 m (figure 3-9). Assuming that the density ratio of the ice and Earth’s crust is about 2/3 (Weertman, 1976), bedrock
deflection under the load of the ice sheet results in a maximum altitude of ∼ 1500 m above mean
sea level. Together with a meridional extent (L) from the margin to the dome of the ice sheet on
the order of 1500 km, this gives a mean surface slope of approximately 0.001 (1 : 1000).
In order for the northward slope of the ice sheet surface to have a stabilizing affect, it has to
be larger than the slope of the surfaces of constant potential temperature. In the high latitudes of
the Northern Hemisphere a typical northward slope of surfaces of constant potential temperature
is ∼ 0.001 (Peixoto and Oort, 1992), which is comparable to the estimated mean surface slope of
the Laurentide ice sheet in the early Pliocene, late Pleistocene. This indicates that the surface slope
of the ice sheet might have a weak stabilizing effect if it covers a large enough area to affect the
baroclinic eddies.
However, it should be kept in mind that in the case of most ice sheets, such as the Laurentide, the
increase in surface elevation with latitude is not constant. Instead, most of the increase in surface
elevation is concentrated at its southern margin. When considering only the large interior surface of
the ice sheet, the slope is smaller than the mean slope. According to a study by Orlanski (1969), the
stabilizing effect of a steep marginal slope of a topographical feature is not great enough to oppose
the increased instability caused by a weakly sloping interior. In the case of Antarctica, which is the
closest modern analogue to the Laurentide ice sheet, the different surface slopes at the margin and
in the vast interior result in a system which becomes more unstable overall (Orlanski, 1969).
Another important question is whether the horizontal extent of the ice sheet is large enough
110
to have an effect on the eddies. The ice sheet can only have a stabilizing effect if it grows to a
meridional extent comparable to the scale of the baroclinic eddies. In the atmosphere, baroclinic
eddies that transport heat have a typical north-south extent of about 3000 km (Stone et al., 1982).
Therefore, if the horizontal extent of the ice sheet is small (∼ 1000 km or less) compared to the
scale of the eddies, the stabilizing effect would be of less importance. As discussed above, the
maximum horizontal extent of the ice sheets in the late Pliocene, early Pleistocene is smaller than
the scale of typical baroclinic eddies. Indicating that their possible stabilizing effect on the growth
of instabilities is relatively small.
Finally, even in the case when the size of the ice sheet is large compared to the scale of the
eddies, and the arguments above suggest that the surface slope could have a stabilizing effect, it
should be noted that in the fully equilibrated eddy regime, when the zonal mean flow is stable,
the baroclinic eddies are still present (Solomon and Stone, 2001). In effect, the eddy heat transport
parameterization used in this study does not depend critically on the nature and source of the eddies.
Rather, the parameterization is based on scaling arguments such as the statement that: the radius
of deformation which maximizes the release of eddy available potential energy is the characteristic
meridional eddy scale; and that the eddy kinetic energy is comparable to the mean kinetic energy.
The first statement is supported by the work of Pedlosky (1975), and the second statement is found
to be robust in all seasons and throughout mid-latitudes for the current climate (Lin, 1980a,b).
Further, a sloping bottom boundary impacts the transport by baroclinic eddies by modifying the
height of the most unstable wave, as is the case for the β̀-effect (Held, 1978). However, in the
atmosphere much of the heat is transported by the longer planetary waves which extend throughout
the troposphere and are not subject to this effect (Stone and Miller, 1980).
The second mechanism, involving the impact of ice sheet topography on atmospheric stationary waves is thought to be of more significance. As suggested by (Roe and Lindzen, 2001a,b) the
presence of large continental ice sheets in the northern hemisphere is expected to have a significant
impact on atmospheric circulation by forcing atmospheric stationary eddies. In the current climate
these stationary eddies contribute significantly to the poleward transport of heat in Northern Hemi-
111
sphere winter, whereas the impact in the summer and in the Southern Hemisphere is negligible
(Peixoto and Oort, 1992). The parameterization for heat transport used in this study does not explicitly include the effects of these stationary eddies. It is not known how a change to the stationary
eddies, caused by the presence of the ice sheets, might have altered the meridional transport of heat
and moisture. However, the studies of Stone (1978) as well as Stone and Miller (1980), indicate
that there is a strong negative feedback between stationary and transient eddy fluxes of heat. Thus,
a change in either transient, or stationary eddy heat flux might be compensated by the other, so as
to maintain a constant total eddy heat transport.
7.4 The Role of Antarctic Ice Sheets
From the results of the model experiments, it is clear that surface melt on large land based ice sheets
in the Northern Hemisphere is highly sensitive to variations in precession. As a consequence,
simulated ice volume has a relatively strong precession signal, which is significantly larger than
observed in the δ 18 O ice volume record of the late Pliocene, early Pleistocene.
In Antarctica the situation is different. At present, the surface of the Antarctic ice sheet is too
cold for any melt to occur, even in the summer season. Instead, ice volume is set by a balance
of accumulation in the interior, and calving of ice into the Southern Ocean. If, as in the Northern
Hemisphere, accumulation is governed by poleward transport of moisture by large scale baroclinic
storm systems, long time period variations in accumulation will be dominated by obliquity. As
surface melt is insignificant, the influence of precession on the mass balance will be significantly
reduced.
Although calving accounts for almost all ablation in Antarctica, and roughly 40% in Greenland,
it is not well understood. Empirical data suggest that there is a weak correlation between calving
rate and water depth at the terminus of the ice sheet (Brown et al., 1982). However, the correlation
breaks down during rapid retreat of the ice, and there is no physical mechanism to explain why
there should be any dependency of the calving rate on water depth (Van der Veen, 2002). On the
112
other hand, there is a good correlation between calving rate and flow speed of the ice (Van der
Veen, 1996). Flow speed is governed by the dynamics and surface mass balance of the ice sheet. If
the surface mass balance is dominated by accumulation, as is the case in Antarctica, variations in
moisture flux will regulate the calving rate. In effect, this suggests that any long period oscillations
in volume of the Antarctic ice sheet would be controlled by changes in obliquity.
From the benthic δ 18 O record alone it is not possible to ascertain whether the observed variations in ice volume originate from the Northern Hemisphere, or from Antarctica. However, when
comparing the δ 18 O record with data on influx of IRD at core sites in the North Atlantic (figure
2-7), it is clear that there were continental ice sheets in the Northern Hemisphere with melting icebergs in the North Atlantic Ocean from at least ∼ 2.8 − 2.6 Ma BP (Jansen et al., 1988; Raymo
et al., 1989; Jansen and Sjoholm, 1991). The evolution of the Antarctic ice sheet is less clear. In
the middle Pliocene (∼ 3 Ma BP), benthic δ 18 O values are lower than in the Holocene, suggesting
that a major deglaciation of Antarctica could have taken place. Alternatively, the low δ 18 O values
can be explained by warmer deep ocean temperatures.
A significant deglaciation of Antarctica in the middle Pliocene is supported by evidence of sea
level being between 20 and 40 m higher than today (Wardlaw and Quinn, 1991; Krantz, 1991;
Dowsett et al., 1994). At present, the total volume of the Antarctic ice sheet corresponds to 61
m of sea level equivalent (Huybrechts et al., 2000); implying that more than half the current ice
sheet could have deglaciated in the middle Pliocene. Further, there is evidence from the terrestrial
record of marine diatoms deposited at several locations on the Trans-antarctic Mountains (Webb
and Harwood, 1991; Barrett et al., 1992), suggesting the presence of marine conditions in the
interior of East Antarctica in the past. The cyclic stratigraphy of these deposits indicates that the
ice sheets on Antarctica could have undergone multiple glacial oscillations. Such major dynamic
changes in ice volume during a period only slightly warmer than present, suggests that the Antarctic
ice sheets are relatively unstable.
However, the evidence for a dynamic East Antarctic ice sheet that has undergone major deglaciations, is not universally accepted by the scientific community. According to the original theory
113
of Kennett (1977), the East Antarctic ice sheet formed in the middle Miocene (∼ 14 Ma BP) in
response to a cooling of Antarctica. The cooling was caused by a progressive isolation of the
Antarctic continent due to the opening of the Drake passage, and a strengthening of the Antarctic
Circumpolar Current. Following the middle Miocene, polar climate cooled and the East Antarctic
ice sheet is believed to have remained stable (Denton et al., 1984; Clapperton and Sugden, 1990;
Kennett and Hodell, 1993). On the other hand, the West Antarctic ice sheet, which is considerably
smaller (∼ 7 m of sea level) and marine-based, is considered to be more vulnerable to changes in
climate.
Although the glacial history of Antarctica is unclear and needs to be investigated further, the
possibility of a dynamic Antarctic ice sheet capable of oscillating in size by more than 50% has
serious implications for the understanding of early Pliocene and late Pleistocene glacial cycles.
Presently, the Antarctic ice sheets cover an area about eight times the size of Greenland, and any
significant change in their volume will have an impact on albedo. In the case of the marine-based
West Antarctic ice sheet, a relatively small reduction in ice volume will result in a large decrease
in area. Changes in surface albedo of the Antarctic continent will have an effect on global climate,
and possibly influence the growth and decay of ice in the Northern Hemisphere.
The impact of changes in obliquity is synchronous between the two hemispheres, whereas the
impact of changes in precession is asynchronous. As a result, changes in ice volume due to variations in obliquity will be in phase in both hemispheres, and could amplify the global ice volume
signal. On the other hand, changes in ice volume due to variations in precession will be out of
phase, and could cancel when calculating the global mean. In effect, obliquity dominated oscillations in Antarctic ice volume in the late Pliocene and early Pleistocene, could have been amplified
by oscillations in ice volume of small versions of the Laurentide and Fennoscandian ice sheets.
The hypothesis of the 41 Ka period glacial cycles being dominated by oscillations in Antarctic
ice volume relies on the ice sheets in the Northern Hemisphere being relatively small. The benthic
δ 18 O record indicates that maximum global ice volume in the late Pliocene and early Pleistocene
is about one third to one half that of the late Pleistocene. However, the gradual cooling of climate
114
from the Pliocene to the late Pleistocene is accompanied by a slow increase in global ice volume
(figure 2-4 on page 19). The maximum size of the Antarctic ice sheets is constrained by the size
of the continent. A similar constraint does not apply to the Laurentide ice sheet. Instead, the
glacial size of the Laurentide ice sheet would have increased gradually until it became unstable.
At this point, the rapid nonlinear deglaciations characteristic of the 100 Ka glacial cycles would
have commenced. The processes responsible for these rapid deglaciations are still unclear, but
suggestions include: warm summers and high surface melt amplified by feedbacks related to ocean
circulation (Ruddiman and McIntyre, 1981); catastrophic melt back of ice streams (Hughes et al.,
1977; Hughes, 1987); and enhanced calving into proglacial lakes formed in the bedrock depression
at the margin of a retreating ice sheet (Andrews, 1973; Pollard, 1982).
In effect, the switch from the dominance of 41 Ka to ∼ 100 Ka period glacial cycles about 0.8
Ma BP could have been the result of the emergence of large ice sheets in the Northern Hemisphere.
After this time, the climate system switched from being controlled by the Southern Hemisphere
to the Northern Hemisphere. During the glacial phase of the ∼ 100 Ka cycles, ice volume is
influenced by both obliquity and precession. The increase in the influence of precession at the time
of the emergence of ∼ 100 Ka ice volume cycles can be seen in figure 2-6 on page 22, and is
most likely a result of an increased influence of surface melt on the large ice sheets of the Northern
Hemisphere.
In a warmer climate, such as the early Pliocene, there could have been some surface melt in
Antarctica. As seen in the model experiments presented in this study, surface melt is dominated by
precession. However, the climate of the model is tuned to fit Northern Hemisphere climate. Due
to the vicinity of the large thermal mass of the Southern Ocean, the seasonal cycle on Antarctica is
significantly weaker than in the north. As the impact of precession relies on a strong preference for
a particular season, a weak seasonal cycle reduces the influence of precession.
115
7.5 Further Work
As pointed out in the previous sections there are several aspects of the topics covered in this study
which inspire further research. A few suggestions for questions which should be addressed in the
future are as follows:
• How will an interactive sea ice component change the sensitivity of the model to variations
in the solar constant, as well as orbital insolation?
• What is the impact of large changes in high latitude sea ice on relative humidity and the
accumulation of ice on continental ice sheets such as the Laurentide and Fennoscandian?
• How will the inclusion of an active ocean circulation as well as the thermal mass of the deep
ocean change the model results?
• What portion of the oscillations in benthic δ 18 O in the late Pliocene, early Pleistocene is due
to changes in temperature and Antarctic ice volume?
• How will the addition of an Antarctic ice sheet change the response of global ice volume to
variations in orbital insolation?
116
Appendix A
The 41 Ka World: Milankovitch’s Other
Unsolved Mystery
117
PALEOCEANOGRAPHY, VOL. 18, NO. 1, 1011, doi:10.1029/2002PA000791, 2003
The 41 kyr world: Milankovitch’s other unsolved mystery
Maureen E. Raymo
Department of Earth Sciences, Boston University, Boston, Massachusetts, USA
Kerim Nisancioglu
Department of Earth, Atmospheric, and Planetary Sciences, Massachusetts Institute of Technology, Cambridge,
Massachusetts, USA
Received 26 March 2002; revised 5 September 2002; accepted 19 September 2002; published 6 March 2003.
[1] For most of the Northern Hemisphere Ice Ages, from 3.0 to 0.8 m.y., global ice volume varied
predominantly at the 41,000 year period of Earth’s orbital obliquity. However, summer (or summer caloric half
year) insolation at high latitudes, which is widely believed to be the major influence on high-latitude climate and
ice volume, is dominated by the 23,000 year precessional period. Thus the geologic record poses a challenge to
our understanding of climate dynamics. Here we propose that variations in the insolation gradient between high
and low latitudes control high-latitude climate and ice volume during the late Pliocene and early Pleistocene.
The differential heating between high and low latitudes, driven by obliquity, controls the atmospheric meridional
flux of heat, moisture, and latent energy, which may exert the dominant control on high-latitude climate on
Milankovitch timescales. In the two-dimensional zonal energy balance models typically used to study the longterm evolution of climate, the meridional atmospheric moisture flux is usually kept fixed. The hypothesis that
insolation gradients control the poleward energy fluxes, precipitation, and ice volume at high latitudes has never
been directly examined within the context of an ice sheet model. In light of what we know about modern energy
INDEX TERMS:
fluxes and their relative influence on high-latitude climate, this possibility should be examined.
4267 Oceanography: General: Paleoceanography; 1620 Global Change: Climate dynamics (3309); 3344 Meteorology and Atmospheric
Dynamics: Paleoclimatology; 3359 Meteorology and Atmospheric Dynamics: Radiative processes; KEYWORDS: Milankovitch, orbital
variations, ice ages, Pleistocene, obliquity, paleoclimate
Citation: Raymo, M. E., and K. Nisancioglu, The 41 kyr world: Milankovitch’s other unsolved mystery, Paleoceanography, 18(1),
1011, doi:10.1029/2002PA000791, 2003.
1. Introduction
[2] All serious students of Earth’s climate history have
heard of the ‘‘100 kyr problem’’ of Milankovitch orbital
theory, namely the lack of an obvious explanation of the
dominant 100 kyr periodicity in climate records of the last
800,000 years. However, few have considered an equally
perplexing characteristic of Earth’s climate, one that similarly defies simple physical explanation yet dominates the
Earth’s recent geologic record. We call this the ‘‘Milankovitch 41 kyr problem.’’ For the time interval extending back
to the Brunhes-Matuyama boundary (0.78 Ma), an interval
in Earth’s climate history dominated by the large (and
largely unexplained) 100,000 year periodicity, Imbrie et
al. [1992] definitively showed that the obliquity (41,000
year) and precessional (23,000 year) frequencies observed
in climate records were direct linear responses, with physically appropriate lags, to high-latitude summer insolation
forcing. However, during the previous two million years of
Northern Hemisphere ice sheet growth, from 3 million
years ago to about 0.8 million years ago, global ice volume
varied almost exclusively at the 41,000 year obliquity
period. Because high-latitude summer insolation is always
Copyright 2003 by the American Geophysical Union.
0883-8305/03/2002PA000791$12.00
dominated by precession, we argue that these earlier climate
variations cannot be understood within the current framework of the Milankovitch Hypothesis. Finding an explanation for late Pliocene to early Pleistocene global climate
variations represents one of the most interesting and challenging problems facing climate modelers today.
2. Ice Age Record
[3] The outlines of Earth’s climate history for the last 3
million years have been known for nearly two decades.
With an extremely low sedimentation rate piston core and
then with longer Deep Sea Drilling Project cores, Nicholas
Shackleton, and later William Ruddiman and others, measured oxygen isotopes in benthic foraminifera to derive a
proxy for global ice volume over the last 3 million years.
Many records generated since this time have confirmed their
early observations, namely: (1) the main frequency of ice
volume change from 3.0 to 0.8 m.y. was 41,000 years, the
primary obliquity period; (2) after 0.8 Ma, ice sheets
varied predominately at the 100,000 year period and the
amplitude of d18O variability increased implying growth of
larger ice sheets.
[4] The double-cored and spliced benthic d18O record from
DSDP607 nicely illustrates both these points (Figure 1). Note
that the isotope record is plotted with a paleomagnetic time-
11 - 1
118
11 - 2
RAYMO AND NISANCIOGLU: THE 41 KYR WORLD
Figure 1. Benthic d18O record from DSDP Site 607 in the North Atlantic (solid line) plotted to a
paleomagnetic timescale. The magnetic field reversals are marked, as well as the transition from a
dominant 41 kyr to a 100 kyr world. B, Brunhes; M, Matuyama; J, Jaramillo; TOld, top of Olduvai; G,
Gauss. Also shown is orbital obliquity (red dashed line).
scale (Table 1) determined by the depth of magnetic field
reversals recorded by ferromagnetic grains in the sediment
core [Clement and Kent, 1986]. Constant sedimentation rates
are assumed between these magnetic reversal events (shown
on Figure 1) which are dated by interpolating seafloor
magnetic anomalies between fixed calibration points [Cande
and Kent, 1992, 1995]. The two calibration points used in the
post-3.0 Ma section of the record are independently derived
by both radiometric and astronomic tuning techniques
[Berggren et al., 1995].
[5] Using this simple timescale, which is not biased by
orbital ‘‘tuning,’’ one can clearly observe the dominant
41,000 year periodicity of the Matuyama and Gauss intervals [see also Imbrie et al., 1993a; Tiedeman et al., 1994].
The obliquity periodicity can be further illustrated by
statistically filtering the data at 41,000 years or by Fourier
analysis (Figure 2a). Note the near complete lack of
variance at the 23,000 year precessional and 100,000 year
eccentricity frequencies. Indeed, over long parts of the
record the d18O curve looks almost sinusoidal. Nearly
identical results are seen in many other deep sea isotope
records including benthic d18O records from the Pacific
plotted either to paleomagnetic or orbitally tuned timescales
(e.g., Figure 2b).
[6] Because Site 607 is located in the subpolar North
Atlantic (41N, 33W, 3427 mbsl), it also contains a record
of ice-rafted detritus (IRD) delivered to the open ocean over
the Plio-Pleistocene. Over the entire length of the glacial
record (>125 m), the input of IRD covaries with d18O
[Raymo et al., 1989; Ruddiman et al., 1989]. The sedimentological data thus demonstrates that variability observed in
benthic d18O must derive in part from the waxing and
waning of ice sheets bounding the North Atlantic.
3. Current Milankovitch Theory
[7] Based mainly on climate proxy records of the last 0.5
Ma, a general scientific consensus has emerged that variations in summer insolation at high northern latitudes are the
dominant influence on climate over tens of thousands of
years. The logic behind nearly a century’s worth of thought
on this topic is that times of reduced summer insolation
could allow some snow and ice to persist from year to year,
lasting through the ‘‘meltback’’ season. A slight increase in
accumulation from year to year, enhanced by a positive
snow-albedo feedback, would eventually lead to full glacial
conditions. At the same time, the cool summers are proposed to be accompanied by mild winters which, through
the temperature-moisture feedback [Kallen et al., 1979],
would lead to enhanced winter accumulation of snow. Both
effects, reduced spring-to-fall snowmelt and greater winter
accumulation, seem to provide a logical and physically
sound explanation for the waxing and waning of the ice
sheets as high-latitude insolation changes [e.g., see Hartmann, 1994, p. 310]. However, in this model, the seasonal
contrast, which is controlled by obliquity and varies systematically at the 41,000 year period, can dominate only if the
precessional effects on insolation are assumed to cancel out
over the course of the annual cycle. This assumption has
generally not been made due to the presence of a strong
precessional signal in late Pleistocene records, hence the
greater relative importance accorded summer insolation
versus seasonal contrast in controlling past climate.
[8] Over the last two decades, countless research papers
have plotted (or tuned) climate records to June 65N or July
65N insolation. Using many of these records, Imbrie et al.
[1992] showed that climate variance at precessional and
obliquity frequencies appeared to be linearly forced by and
was coherent with northern summer insolation. Only the
100,000 year cycle is left unexplained by this model (the
familiar ‘‘100,000 year problem’’) and it, typically, is
ascribed to non-linear variability arising internally within
the climate system. A comprehensive summary of work on
this subject is given by Imbrie et al. [1992, 1993b] [see also
Peltier and Marshall, 1995; Gildor and Tziperman, 2000;
Muller and MacDonald, 2000].
[9] In the late Pliocene and early Pleistocene, no significant variance at the 100 kyr period is observed in benthic
Table 1. Age Control Points Used for Paleomagnetic Timescale at
DSDP Site 607
Depth, mcd
Age, kyr
Magnetic Event
0.0
31.84
40.345
43.965
73.655
111.58
129.50
0
780
984
1049
1757
2600
3054
top of core
Brunhes/Matuyama
Jaramillo top
Jaramillo bottom
Olduvai top
Matuyama/Gauss
Kaena top
119
RAYMO AND NISANCIOGLU: THE 41 KYR WORLD
11 - 3
Figure 2. (a) Power spectra of Site 607 d18O plotted using timescale discussed in text (Table 1). The
concentration of variance around 75 –80 kyr is likely due to harmonics arising from the main obliquity
frequency; (b) power spectra of the benthic d18O record of ODP Site 846 [Mix et al., 1995] plotted to
orbitally tuned timescale of Shackleton et al. [1990]. Bandwidths (bw) and confidence intervals (ci)
shown on figure.
d18O records. Hence one might expect that the de facto
implication of the ‘‘standard Milankovitch model’’ would
be that global ice volume should vary linearly and coherently with high northern summer insolation. However, a
comparison of d18O (ice volume; Figure 1) with various
insolation records (Figure 3) clearly shows that while the ice
volume proxies are dominated nearly exclusively by the
41,000 year obliquity periodicity, summer insolation is
dominated, at nearly every latitude, by the 23,000 year
period of precession. Any linear response to summer (or
summer half year) insolation by high-latitude climate would
require the strong presence of precession in the geologic
record. In fact, this frequency is barely discernable in only a
small stretch of the late Pliocene ice volume record and is
absent over most (Figures 1 and 2). One must conclude that
summer insolation at high northern latitudes does not exert a
dominant (linear) influence on climate over most of the
northern hemisphere Ice Ages.
4. The 41 kyr Problem
[10] While many investigators have attempted to model
the 100 kyr world, few have focused their attention on the
41 kyr world. A notable exception is Andre Berger and
colleagues who used a two dimensional ice sheet-climate
model to try to simulate the growth and decay of ice sheets
over the last 3 million years [e.g., Berger et al., 1999].
While the obliquity period is present in the model output,
precessional variance in ice sheet mass is also strongly
present. In other words, although they successfully model
the lack of the 100 kyr eccentricity cycle, they were not able
to model an ice sheet that varies only at the obliquity
frequency. This appears to be because the model is ultimately very sensitive to high-latitude summer insolation.
[11] Secondly, a discussion of the 41 kyr problem can be
found in Richard Muller and Gordon MacDonald’s book ‘‘Ice
Ages and Astronomical Causes’’ [Muller and MacDonald,
2000]. Following Kukla [1968], they propose that northern
latitude winter insolation (e.g., January 65N) may drive late
Pliocene/early Pleistocene climate cycles, even though the
total insolation received in January is a factor of 20 less than
summer insolation at the same latitude. However, they go on
to say this proposition is speculative and that the geologic
record is posing a problem that needs to be solved.
5. Insolation Gradients
[12] Given that summer insolation has too much precession
and (one could argue) winter insolation appears too weak to
drive anything, what is left? We propose that the gradient in
insolation between high and low latitudes may, through its
influence on the poleward flux of moisture which fuels ice
sheet growth, play the dominant role in controlling climate
from 3 to 1 million years ago. Summer half-year insolation,
for instance, is calculated as the mean of insolation received
between the vernal (0) and autumnal (180) equinoxes as
defined by the longitude of the sun in degrees. The gradient
(or difference) in summer half-year insolation between 25
and 70N (Figure 3) is almost completely dominated by
obliquity (spectra shown in Figure 4). It is this temperature
gradient that drives the poleward heat, moisture, and momentum fluxes in the atmosphere; the correlation between d18O
120
11 - 4
RAYMO AND NISANCIOGLU: THE 41 KYR WORLD
Figure 3. Various insolation and insolation gradient curves compared with obliquity. Curves derived
using the Laskar [1990] orbital solution and Analyseries software of Paillard et al. [1996].
and the insolation gradient (Figure 5) suggests that increased
gradients promote ice sheet growth (although given the
uncertainties in the timescale we cannot definitively rule
out other possibilities). Note that the oxygen isotope record in
Figure 5 has been shifted to an older age by 8 kyr, a
reasonable lag to assume for the climate response to obliquity
and consistent with late Pleistocene observations [e.g.,
Imbrie et al., 1992]. Of course the true lag of response after
forcing would be almost impossible to determine directly in
sediments of this age.
[13] The idea that insolation gradients could exert an
important control on climate on Milankovitch timescales
is not new; Young and Bradley [1984] proposed that hemispheric insolation gradients may have contributed to the
growth and decay of continental ice sheets through their
modulating influence on the poleward transport of moisture.
They, and previously Berger [1976], suggest that times of
rapid ice growth and decay correspond to especially pronounced deviations in latitudinal insolation gradients. Johnson [1991] similarly invokes a decrease in the insolation
gradient, rather than direct summer insolation at high
latitudes, as the immediate cause of the deglaciation at
Termination 2, offering this mechanism as the explanation
for paleoclimate data which suggest that deglaciation
occurred prior to the increase in summer insolation. This
perplexing mismatch in timing between the deglaciation at
Termination 2 and the timing predicted by Milankovitch
theory has also been discussed by Winograd et al. [1992]
and more recently by Gallup et al. [2002].
[14] It may be that we are underestimating the influence
of meriodional fluxes of sensible and latent heat, driven by
hemispheric temperature gradients, on continental ice sheet
size. The mass balance of an ice sheet is set by the relative
rates of accumulation and ablation. The rate of ablation is
controlled by local incoming solar radiation and local
atmospheric temperature. The rate of accumulation is controlled by the amount of moisture available for precipitation
as well as the local temperature. As temperature and
precipitation at high latitudes are strongly influenced by
the magnitude of the atmospheric meridional heat and
moisture fluxes, we would thus expect these fluxes to exert
a strong influence on ice sheet mass balance. Today, annual
mean poleward transports of heat by the atmosphere peak at
about 5.0 PW in the mid-latitudes of both hemispheres
121
RAYMO AND NISANCIOGLU: THE 41 KYR WORLD
Figure 4. Power spectra of the gradient in insolation
between 25 and 70N for the summer half year between 0
(vernal equinox) and 180 (autumnal equinox). Bandwidth
(bw) and confidence interval (ci) shown on figure.
[Trenberth and Caron, 2001], compensating the loss of heat
to space from the polar regions. At high latitudes (>60N)
the Earth emits on an annual basis approximately twice as
much energy (as long wave radiation) as it receives from
absorbed solar radiation [Hartmann, 1994]; hence the meridional heat flux is comparable in magnitude to that received
from high-latitude insolation. Lastly, recent studies suggest
that a much greater portion of poleward energy transport
occurs in the atmosphere, rather than the ocean which
11 - 5
would be less directly affected by insolation gradients
[Trenberth and Caron, 2001].
[15] All of the above observations suggest the possibility
that variations in meridional heat and moisture fluxes (driven
by orbital obliquity variations) could be large enough to
override the effects of local insolation variations and imprint
the dominant 41 kyr signal on the ice volume record. This
seems especially plausible given the powerful ice-albedo
feedback that would enhance the effects of insolation gradients on poleward energy transports. As the polar atmospheric temperature cooled at the onset of a glacial period,
snow and ice would expand into regions previously covered
by surfaces such as forests that have relatively low albedo.
This increased snow/ice cover would raise the surface albedo
dramatically, reflecting incoming radiation, and causing a
further decrease in local temperature [e.g., Bonan et al.,
1992; Kutzbach and Gallimore, 1998]. Such an albedo
change could have two effects: (1) act as a strong positive
feedback on the meridional temperature gradients, further
enhancing the poleward transport of the moisture that feeds
ice sheet growth; and (2) act as a negative feedback by
causing local cooling which decrease moisture availability
through the temperature-precipitation feedback. Perhaps it is
the interplay between these two feedbacks that determines
the maximum ice sheet size (and perhaps the difference
between the early and late Pleistocene climate behavior).
[16] Support for the idea that insolation gradients influence moisture flux to the ice sheets is found in the Antarctic
deuterium excess record of the past 150 Ka which shows a
strong correlation with the mean annual insolation gradient
from 20S to 60S [Vimeux et al., 1999]. As deuterium
excess is a measure of the evaporative conditions of the
oceanic source region for the moisture, this data indicates
that there is a strong link between the insolation gradient and
atmospheric moisture supply to the ice sheets. Note, that the
deuterium excess values are high during glacial inception, a
Figure 5. Site 607 benthic d18O record plotted versus the summer half-year gradient between 25 and
70N. Isotope data is plotted to paleomagnetic timescale given in Table 1 and then shifted older by an
assumed response lag of 8 kyr after the forcing [after Imbrie et al., 1992] to better illustrate the correlation
between the two parameters. The gradient in insolation is in phase with and shows a similar amplitude
modulation as obliquity.
122
11 - 6
RAYMO AND NISANCIOGLU: THE 41 KYR WORLD
period characterized by low obliquity (cold high-latitude
summers), and high meridional insolation gradient (e.g., a
strong meridional atmospheric moisture flux).
[17] In most climate modeling studies of the long-term
evolution of glacial-interglacial cycles (typically 2-D zonal
energy balance models), the meridional atmospheric moisture flux has been kept fixed. Typically, models perturb the
modern observed precipitation field according to changes in
temperature. As a result, changes in atmospheric moisture
flux have no impact on high-latitude precipitation and the
accumulation of ice sheets. A notable exception is the study
by Gildor and Tziperman [2000]. However, the hypothesis
that insolation gradients as they control variations in poleward moisture fluxes and precipitation at high latitudes has
never been directly examined within the context of an ice
sheet model. In light of what we know about modern energy
fluxes and their relative influence on high-latitude climate,
this possibility should be examined.
6. Future Directions
[18] Above, we propose a ‘‘gradient hypothesis’’: that the
strong obliquity signal imprinted on the Ice Age record is
caused by the control meridional temperature gradients
exert on the poleward transport of moisture. As obliquity
decreases, cooling at high latitudes occurs and the gradient
in solar heating between high and low latitude increases.
Both effects, cooling polar regions and the enhanced delivery of moisture, promote ice sheet growth. The ice volume/
temperature history of the last few million years is now well
known and, as discussed above, poses a challenge both to
climate modelers and paleoclimatologists. Ultimately, ocean
circulation and ice volume are being controlled by atmospheric dynamics that must be sensitive to Milankovitch
variations in incoming solar radiation. These are the same
physical processes that will determine the response of
Earth’s climate to rising greenhouse gases. Building a
model which can reproduce the first-order features of the
Earth’s Ice Age history over the Plio-Pleistocene would be
an important step forward in the understanding of the
dynamic processes that drive global climate change.
[19] Acknowledgments. The authors would like to thank Bill Ruddiman, Eli Tziperman, Peter Stone, Larry Peterson and an anonymous
reviewer for comments, questions, and encouragement that significantly
improved this manuscript as well as John Imbrie for his suggestions and his
continued inspiration in thinking about Milankovitch. MER also thanks the
NSF Marine Geology and Geophysics and Atmospheric Sciences Paleoclimate Programs for the support of this research.
References
Berger, A., Long term variations of daily and
monthly insolation during the last Ice Age,
Eos Trans AGU, 57(4), 254, 1976.
Berger, A., X. S. Li, and M. F. Loutre, Modeling
northern hemisphere ice volume over the last 3
Ma, Quat. Sci. Rev., 18(1), 1 – 11, 1999.
Berggren, W., F. Hilgen, C. Langereis, D. Kent,
J. Obradovich, I. Raffi, M. E. Raymo, and
N. Shackleton, Late Neogene chronology: New
perspectives in high resolution stratigraphy,
Geol. Soc. Am. Bull., 107, 1272 – 1287, 1995.
Bonan, G. B., D. Pollard, and S. Thompson, Effects of boreal forest vegetation on global climate, Nature, 359, 716 – 718, 1992.
Cande, S. C., and D. V. Kent, A new geomagnetic polarity time scale for the Late Cretaceous and Cenozoic, J. Geophys. Res., 97(B10),
13,917 – 13,951, 1992.
Cande, S. C., and D. V. Kent, Revised calibration
of the geomagnetic polarity timescale for the
late Cretaceous and Cenozoic, J. Geophys.
Res., 100(B4), 6093 – 6095, 1995.
Clement, B. M., and D. V. Kent, Geomagnetic
polarity transition records from five hydraulic
piston core sites in the North Atlantic, Init. Rep.
Deep Sea Drill. Proj., 96, 831 – 852, 1986.
Hartmann, D. L., Global Physical Climatology,
409 pp., Academic, San Diego, Calif., 1994.
Gallup, C. D., H. Cheng, F. W. Taylor, and R. L.
Edwards, Direct determination of the timing of
sea level change during Termination II,
Science, 295, 310 – 313, 2002.
Gildor, H., and E. Tziperman, Sea ice as the
glacial cycles climate switch: Role of seasonal
and orbital forcing, Paleoceanography, 15,
605 – 615, 2000.
Imbrie, J., et al., On the structure and origin of
major glaciation cycles, 1, Linear responses to
Milankovitch forcing, Paleoceanography, 7,
701 – 738, 1992.
Imbrie, J., A. Berger, and N. J. Shackleton, Role of
orbital forcing: a two-million-year perspective,
in Global Changes in the Perspective of the
Past, edited by J. A. Eddy and H. Oeschger,
pp. 263 – 277, John Wiley, New York, 1993a.
Imbrie, J., et al., On the structure and origin of
major glaciation cycles, 2, The 100,000-year
cycle, Paleoceanography, 8, 699 – 736, 1993b.
Johnson, R. G., Major Northern Hemisphere deglaciation caused by a moisture deficit 140 ka,
Geology, 19, 686 – 689, 1991.
Kallen, E., C. Crafoord, and M. Ghil, Free oscillations in a climate model with ice-sheet dynamics, J. Atmos. Sci., 36(12), 2292 – 2303,
1979.
Kukla, J., The Pleistocene epoch and the evolution of man-A reply, Curr. Anthropology, 9,
27 – 47, 1968.
Kutzbach, J. E., and R. G. Gallimore, Sensitivity
of a coupled atmosphere/mixed layer ocean
model to changes in orbital forcing at 9000
years B.P., J. Geophys. Res., 93, 803 – 821,
1998.
Laskar, J., The chaotic motion of the solar system: A numerical estimate of the chaotic
zones, Icarus, 88, 266 – 291, 1990.
Mix, A. C., J. Le, and N. J. Shackleton, Benthic
foraminiferal stable isotope stratigraphy of Site
846: 0 – 1.8 Ma, Proc. Ocean Drill. Prog. Sci.
Results, 138, 839 – 854, 1995.
Muller, R. A., and G. J. MacDonald, Ice Ages
and Astronomical Causes, 318 pp., SpringerVerlag, New York, 2000.
Paillard, D., L. Labeyrie, and P. Yiou, Macintosh
program performs time-series analysis, Eos
Trans. AGU, 77, 379, 1996.
Peltier, W. R., and S. Marshall, Coupled energybalance/ice-sheet model simulations of the glacial cycles: a possible connection between terminations and terrigenous dust, J. Geophys.
Res., 100, 14,269 – 14,289, 1995.
Raymo, M. E., W. F. Ruddiman, J. Backman,
B. M. Clement, and D. G. Martinson, Late
Pliocene variation in Northern Hemisphere
123
ice sheets and North Atlantic deep circulation,
Paleoceanography, 4, 413 – 446, 1989.
Ruddiman, W. F., M. E. Raymo, D. G. Martinson,
B. M. Clement, and J. Backman, Mid-Pleistocene evolution of Northern Hemisphere climate, Paleoceanography, 4, 353 – 412, 1989.
Shackleton, N. J., A. Berger, and W. R. Peltier,
An alternative astronomical calibration of the
lower Pleistocene time scale based on ODP
site 677, Trans. R. Soc. Edinburgh Earth
Sci., 81, 251 – 261, 1990.
Tiedeman, R., M. Sarnthein, and N. J. Shackleton, Astronomic timescale for the Pliocene
Atlantic d18O and dust flux records of Ocean
Drilling Program site 659, Paleoceanography,
9, 619 – 638, 1994.
Trenberth, K. E., and J. Caron, Estimates of meridional atmosphere and ocean heat transports,
J. Clim., 14, 3433 – 3443, 2001.
Vimeux, F., V. Masson, J. Jouzel, M. Stievenard,
and J. R. Petit, Glacial-interglacial changes in
ocean surface conditions in the Southern
Hemisphere, Nature, 398, 410 – 413, 1999.
Winograd, I. J., T. Coplen, J. Landwehr,
A. Riggs, K. Ludwig, B. Szabo, P. Kolesar,
and K. Revesz, Continuous 500,000-year climate record from vein calcite in Devil’s Hole,
Nevada, Science, 258, 255 – 260, 1992.
Young, M. A., and R. S. Bradley, Insolation gradients and the paleoclimatic record, in Milankovitch and Climate, Part 2, edited by A. L.
Berger et al., pp. 707 – 713, D. Reidel, Norwell, Mass., 1984.
K. H. Nisancioglu, Department of Earth,
Atmospheric, and Planetary Sciences, Massachusetts Institute of Technology, Room 54-1715,
Cambridge, MA 02139, USA.
M. E. Raymo, Department of Earth Sciences,
Boston University, 685 Commonwealth Ave.,
Boston, MA 02215, USA. ([email protected])
Appendix B
Reorganization of Miocene Deep Water
Circulation in Response to the Shoaling of
the Central American Seaway
124
PALEOCEANOGRAPHY, VOL. 18, NO. 1, 1006, doi:10.1029/2002PA000767, 2003
Reorganization of Miocene deep water circulation in response
to the shoaling of the Central American Seaway
Kerim H. Nisancioglu
Program in Atmospheres, Oceans, and Climate, Department of Earth, Atmospheric, and Planetary Sciences, Massachusetts
Institute of Technology, Cambridge, Massachusetts, USA
Maureen E. Raymo
Department of Earth Sciences, Boston University, Boston, Massachusetts, USA
Peter H. Stone
Program in Atmospheres, Oceans, and Climate, Department of Earth, Atmospheric, and Planetary Sciences, Massachusetts
Institute of Technology, Cambridge, Massachusetts, USA
Received 5 February 2002; revised 10 August 2002; accepted 26 August 2002; published 11 February 2003.
[1] The response of ocean circulation to the shoaling of the Central American Seaway (CAS) is investigated
using the Massachusetts Institute of Technology (MIT) Ocean General Circulation Model (OGCM). In contrast to
earlier model studies, it is found that significant amounts of deep water are formed in the North Atlantic prior to
the closure of the CAS. However, the circulation pattern is fundamentally different from the modern ocean. In the
upper layers of the CAS, there is a relatively strong geostrophic flow from the Pacific to the Atlantic, controlled
by the pressure difference across the seaway. However, when the CAS is deeper than the level of North Atlantic
Deep Water (NADW) outflow, a significant amount of NADW passes through the CAS to the Pacific Ocean. In
the Pacific, the deep water traverses the basin from east to west in a relatively narrow zonal jet, and becomes a
southward flowing western boundary current, before it joins with the Antarctic Circumpolar Current (ACC) to
the south. This implies that deep sea sediment records from the Miocene Pacific Ocean could have been
influenced by relatively young NADW and provides a new framework for the interpretation of geochemical
tracer data. Diversification of benthic foraminifer fauna suggests that the CAS shoaled to a depth of about 1000 m
toward the end of the middle Miocene. This would have prevented the passage of NADW to the Pacific and
INDEX TERMS: 1620 Global Change: Climate
established the modern deep water circulation pattern at that time.
dynamics (3309); 3344 Meteorology and Atmospheric Dynamics: Paleoclimatology; 4255 Oceanography: General: Numerical modeling;
4267 Oceanography: General: Paleoceanography; KEYWORDS: paleoceanography, miocene, climate, model, deep water, world ocean
Citation: Nisancioglu, K. H., M. E. Raymo, and P. H. Stone, Reorganization of Miocene deep water circulation in response to the
shoaling of the Central American Seaway, Paleoceanography, 18(1), 1006, doi:10.1029/2002PA000767, 2003.
1. Introduction
[2] Some of the most important changes to past ocean
circulation and climate have been connected with tectonic
events involving the closure or opening of oceanic gateways. The most recent event of this nature was the closure
of the Central American Seaway (CAS) between North and
South America. The details of the tectonics in the region are
complicated and not well constrained. However, the shoaling of the seaway is thought to have been gradual, beginning 16 Ma at the early to middle Miocene boundary
[Keller and Barron, 1983; Duque-Caro, 1990; Droxler et
al., 1998], with final closure at about 3 Ma in the middle
Pliocene [Keigwin, 1982; Marshall et al., 1982; Coates et
al., 1992].
[3] Before the closure of the CAS, water flowed between
the Pacific and Atlantic Oceans at a latitude of about 10N.
Copyright 2003 by the American Geophysical Union.
0883-8305/03/2002PA000767$12.00
Experiments with the Hamburg Ocean General Circulation
Model (OGCM) suggest that North Atlantic Deep Water
(NADW) production was severely reduced, or nonexistent
when the CAS was open, due to a flow of relatively fresh
Pacific water into the North Atlantic [Maier-Reimer et al.,
1990; Mikolajewicz et al., 1993; Mikolajewicz and Crowley,
1997]. While some deep sea sediment core data indicate that
the production of NADW increased in the early Pliocene as
the CAS closed [Tiedemann and Franz, 1997; Haug and
Tiedemann, 1998; Billups et al., 1999], other data suggests
that NADW production was significant in the Miocene
when the CAS was open [Keller and Barron, 1983; Miller
and Fairbanks, 1985; Woodruff and Savin, 1989; Delaney,
1990; Wright et al., 1992].
[4] In this study, the structure of the meridional ocean
circulation and its sensitivity to flow through the CAS at
different depths is reexamined with the Massachusetts
Institute of Technology (MIT) OGCM. It is important to
note that the opening of the CAS is applied to the control
run of the modern ocean, where modern values are used for
6-1
125
6-2
NISANCIOGLU ET AL.: REORGANIZATION OF MIOCENE DEEP WATER CIRCULATION
Table 1. Common Parameters Used in Experiments With the MIT Ocean GCM
Ah
Az
Kh
Kz
DTmom
DTtracer
Smax
KI
Dz
Parameters
Value
Horizontal viscosity
Vertical viscosity
Horizontal diffusivity
Vertical diffusivity
Momentum time step
Tracer time step
GM Maximum slope
GM isopycnal diffusivity
Thickness of model layers (m)
5 105 m2/s
103 m2/s
0
5 105 m2/s
40 min
1 day
0.01
103 m2/s
50, 70, 100, 140, 190, 240, 290, 340, 390, 440, 540, 590, 640, and 690
the boundary conditions. Therefore, the results should be
viewed as a sensitivity study of the meridional overturning
circulation to an open CAS, with possible connections to
ocean circulation before the closure, rather than as a
simulation of climate before 3 Ma.
2. Description of the Model and Boundary
Conditions
2.1. The MIT OGCM
[5] The MIT OGCM is based on the incompressible,
Boussinesq form of the Navier – Stoke’s equations. Full
details of the equations and solution techniques are
described by Marshall et al. [1997a, 1997b]. In this study
the model is used in the hydrostatic limit, and the horizontal
resolution is constant at about 2.8 2.8. The maximum
depth is 5200 m and there are 15 vertical levels, with the
thickness of the layers gradually increasing from 50 m at the
ocean surface to 690 m at the bottom (Table 1). As coarse
resolution OGCMs are not capable of resolving mesoscale
eddies, and ocean circulation is affected by mixing processes due to these eddies, it is necessary to apply a subgridscale parameterization for the advection of tracers (such as
temperature and salinity). In this version of the model, the
parameterization of Gent and McWilliams [1990], together
with diffusion of tracers along isopycnals [Redi, 1982], is
implemented as described by Griffies [1998].
2.2. Surface Boundary Conditions
[6] In the experiments, the model ocean is initialized with
temperature and salinity fields from the studies of Levitus
and Boyer [1994] and Levitus et al. [1994] at all depths for
the month of March, when maximum convection takes
place in the North Atlantic [Marshall and Schott, 1999].
Subsequently, the surface is forced with observed surface
heat and freshwater fluxes, as well as sea surface temperature, i.e., mixed boundary conditions given by
@S
¼ SR FW
@z
ð1Þ
@T Dz
Qobs
¼ ðTobs T Þ þ
@z tT
r 0 CP
ð2Þ
Kz
Kz
where SR is a reference salinity taken to be equal to 35 psu,
FW is the net freshwater flux in m/s, due to evaporation,
precipitation, and runoff (from rivers and ice discharge), r0
is a reference density, CP is the specific heat capacity of
seawater, and Qobs is the net heat flux into the oceans in W/
m2. The sea surface temperature relaxation time step tT is
set to 75 days.
[7] Traditionally, when using mixed boundary conditions,
the surface freshwater flux (FW) and the heat flux (Qobs) are
diagnosed from a spin-up run using restoring boundary
conditions [e.g., Maier-Reimer et al., 1990]. In this study,
the diagnosed fluxes are not used because they do not
correspond to the observed atmospheric fluxes of heat and
freshwater, and numerous model studies have found that
GCMs forced with traditional mixed boundary conditions
are unrealistically unstable to perturbations in the surface
fresh water field [Marotzke and Willebrand, 1991; Mikolajewicz and Maier-Reimer, 1994; Tziperman et al., 1994].
[8] The annual mean surface freshwater flux is based on
precipitation minus evaporation data of Schmitt et al. [1989]
for the Atlantic and Baumgartner and Reichel [1975] for the
other oceans, combined with river runoff data of Perry et al.
[1996], as well as Greenland ice-calving data of Reeh [1994]
(Figure 1). It should be noted that the Arctic and its influence
on the freshwater budget is not represented in the present
version of the model. Monthly data for surface heat flux is
compiled from an updated version of Trenberth and Solomon [1994] [see Jiang et al., 1999]. Monthly wind stress
fields are obtained from the data of the European Center for
Medium-Range Weather Forecasts (ECMWF) for the years
1980– 1986, as computed by Trenberth et al. [1989].
2.3. Experiment Implementation
[9] The time steps for the momentum and tracer equations
are 40 min and 24 hours, respectively. Both time steps are
constant with integration time, and the same values are used
for all experiments. Note that the tracer time step is larger
than the momentum time step by a factor of 36. This speeds
up the relatively slow abyssal processes, and significantly
accelerates the convergence of the model to equilibrium
[Bryan, 1984]. The disadvantage is that it distorts the
seasonal cycle [Danabasoglu et al., 1996]. However, the
annual mean conditions, which are discussed in this paper,
are preserved [Kamenkovich et al., 2002].
[10] The following three model experiments were undertaken to test the sensitivity of ocean circulation to the CAS:
1. CNTR: Control experiment with modern configuration
of the continents and bathymetry,
2. CAS2700: Perturbed experiment with the bathymetry
of the control experiment modified by introducing a three
126
NISANCIOGLU ET AL.: REORGANIZATION OF MIOCENE DEEP WATER CIRCULATION
6- 3
Figure 1. Annual mean surface freshwater flux (FW) in m/yr equal to evaporation minus precipitation
and runoff (from rivers and ice calving).
grid boxes wide and 2700 m deep channel, separating North
and South America at latitudes of about 11 –19.5N,
3. CAS1000: Same as CAS2700, but with a 1000 m deep
channel.
The structure of the CAS chosen for experiment CAS2700 is
similar to that used by Maier-Reimer et al. [1990]. The
intermediate sill depth in experiment CAS1000 is chosen,
because there is evidence from benthic foraminifera suggesting that the sill could have shoaled to about 1000 m by the
end of the middle Miocene (12.5 Ma) [Duque-Caro,
1990]. The same climatological data sets are used for the
experiments with an open CAS as for the control experiment.
3. Model Results
3.1. Meridional Overturning Transport
[11] The Eulerian mean meridional overturning stream
function, for the Global and Atlantic Oceans of the control
experiment (CNTR), shows that the overturning circulation
is dominated by the thermohaline cell, originating primarily
in the North Atlantic (Figure 2a). This clockwise overturning consists of a warm, relatively saline northward flow
in the upper branch, and a cold, southward return flow in
the lower branch, recognized as NADW. The maximum
strength of the cell is positioned at about 55N and 1200 m
depth in the North Atlantic with an overturning strength of
30 Sv. This is associated with 18 Sv of NADW transport
southward across the equator, which agrees well with the
modern value of 18 ± 3 Sv, estimated by Ganachaud
[1999]. From the same figure it can be seen that there is
very little Antarctic Bottom Water (AABW) produced in
the Atlantic basin of the model. This is thought to be due
partly to the lack of a representation of sea ice and the
associated brine rejection in the present version of the
model.
[12] A comparison of the control (CNTR) with the perturbed experiments (CAS2700 and CAS1000), reveals that
the global meridional overturning circulation remains relatively unchanged when the CAS is open (Figures 2b and 2c).
The maximum overturning in the Atlantic is reduced to 26 Sv
in CAS2700 and CAS1000. Also, the deep southward transport of NADW across the equator in CAS1000 is reduced to
14 Sv. However, in experiment CAS2700 this transport
decreases to 6 Sv, which is about one third of the value
found in the control experiment. In other words, although the
global meridional overturning is relatively strong, the influence of NADW in the South Atlantic appears to be greatly
reduced when the CAS is deeper than about 1000 m.
3.2. Surface and Deep Currents
[13] The surface and deep circulation for experiment
CNTR (Figures 3a and 4a) correspond reasonably well with
the observed currents of the modern ocean. NADW originating in the Labrador and Greenland Seas, flows south as a
western boundary current in the Atlantic, before it joins the
Antarctic Circumpolar Current (ACC) and eventually
upwells in the Pacific and Indian Oceans. A deep western
boundary current is also observed in the southern Pacific,
where a combination of AABW and NADW flows north
from the ACC region.
[14] In experiment CAS2700, the surface circulation
(Figures 3b and 3c) does not change much from that of
the control experiment, except for a weakening of the
western boundary current in the South Atlantic. However,
a significant amount of water is shown flowing from the
Pacific to the Atlantic. This flow increases in strength
below the surface to a maximum at a depth of about
500 m (Figure 5).
[15] At depths below about 1000 m, the flow reverses,
and a strong flow from the Atlantic through the CAS to the
Pacific is observed. The result is a reduction in the amount
of deep water crossing the equator from the North Atlantic
to the South Atlantic compared to experiment CNTR, as
was observed in the figures of the meridional overturning
stream function (Figure 2b). In the same model run, it can
be seen that the deep western boundary current in the
southern Pacific reverses direction, and flows south toward
the ACC region (Figures 4b and 4c). The reversal of the
deep western boundary current is observed down to depths
of about 3500 m, below which the circulation remains
relatively unchanged (not shown).
[16] The deep flow from the Atlantic to the Pacific is
caused by the disappearance of the E-W pressure gradient
when the western boundary of the Atlantic basin is removed
and replaced by the seaway. This allows the deep western
boundary current in the Atlantic (which today consists
127
6-4
NISANCIOGLU ET AL.: REORGANIZATION OF MIOCENE DEEP WATER CIRCULATION
Figure 2. Eulerian mean Meridional Overturning Stream (MOS) function in units of Sv (Sv = 106 m3/s)
for the Global and Atlantic oceans. The MOS is calculated from a 10-year mean of meridional velocity
after 2000 years of model integration. Negative values (solid contours) imply clockwise overturning, and
positive values (stippled contours) imply counterclockwise overturning. The contour interval is 4 Sv. (a)
CNTR, (b) CAS2700, and (c) CAS1000.
mostly of NADW) to pass through the CAS to the Pacific,
and reduces the southward flow to the South Atlantic. The
total eastward and westward transports through the CAS
with a sill depth of 2700 m are about 17 and 10 Sv,
respectively (Figure 5).
[17] In experiment CAS1000, deep westward transport of
NADW through the CAS is prevented by the shallow sill.
This sill creates a boundary between the Atlantic and Pacific
at depths below 1000 m, which supports the flow of the
deep western boundary current to the South Atlantic. There-
fore, the transport in the CAS is mainly eastward, with a
strength of about 16 Sv (Figure 5).
3.3. Poleward Heat Transport (PHT)
[18] According to Ganachaud and Wunsch [2000], the
estimated maximum modern PHT by the global ocean is
1.8 ± 0.3 and 0.8 ± 0.6 PW for the Northern and Southern
Hemispheres, respectively. These are relatively close to the
values found for experiment CNTR, where the maximum
PHTs are 1.7 and 1.4 PW (Figure 6a). In this experiment,
128
NISANCIOGLU ET AL.: REORGANIZATION OF MIOCENE DEEP WATER CIRCULATION
Figure 3. Ocean currents in m/s after 2000 years of model integration at a depth of 25 m for
experiments (a) CNTR, (b) CAS2700, and (c) CAS2700 minus CNTR.
129
6- 5
6-6
NISANCIOGLU ET AL.: REORGANIZATION OF MIOCENE DEEP WATER CIRCULATION
Figure 4. Ocean currents in m/s after 2000 years of model integration at a depth of 2500 m for
experiments (a) CNTR, (b) CAS2700, and (c) CAS2700 minus CNTR.
130
NISANCIOGLU ET AL.: REORGANIZATION OF MIOCENE DEEP WATER CIRCULATION
Figure 5. Transport through the CAS in Sv as a function
of depth for experiments CAS2700 (solid blue line) and
CAS1000 (dashed green line), with sill depths of 2700 and
1000 m, respectively. Positive and negative values imply
eastward and westward transport, respectively.
transport by the Atlantic is northward at all latitudes, and
dominates the PHT in the Northern Hemisphere, with a
smaller northward contribution by the Indo-Pacific. In the
Southern Hemisphere, the Indo-Pacific dominates with a
strong southward transport.
[19] In experiments CAS2700 and CAS1000, the global
PHT in the Northern Hemisphere is reduced by 14% and the
global PHT in the Southern Hemisphere is enhanced by
13%, compared to the control experiment. The decrease in
the Northern Hemisphere is due to a smaller Atlantic heat
transport, whereas the contribution by the Indo-Pacific
remains relatively unchanged.
[20] The observed increase in the heat transport in the
Southern Hemisphere is linked to the reduced export of
NADW south of the equator. In experiment CNTR, 18 Sv of
NADW is exported southward across the equator. This has
to be balanced by an import of surface waters from the
South Atlantic. However, when the NADW export is
reduced, as in CAS2700 (6 Sv) and CAS1000 (14 Sv),
the import of South Atlantic surface water is reduced as
well. The result is a warming of South Atlantic surface
waters, and an increase in southward heat transport.
4. Discussion
[21] The model experiments suggest that the shoaling of
the CAS had a significant impact on the circulation of the
6- 7
world’s oceans. The most notable result is the strong flow
through the CAS. This flow is found to vary in strength and
direction with depth, and has a significant impact on the
global overturning circulation; in particular the deep circulation of the Pacific Ocean.
[22] The experiments show that when the CAS is open,
deep water forms in the high latitudes of the North Atlantic
(NADW) and passes through the CAS to the Pacific. Once
in the Pacific, the deep water flows westward in a relatively
narrow zonal jet, and approaches the western boundary of
the basin. At the western boundary, the water becomes a
southward flowing boundary current, joining with the ACC
to the south (see Figure 4b). In the Atlantic, part of the
NADW produced flows past the CAS as a western boundary current even when it is open, however, the strength of
the current is greatly reduced compared to the control
experiment.
[23] These model results are in agreement with the
dynamical framework of Stommel and Arons [1960a,
1960b] for the abyssal circulation. According to this
theory, deep water produced at its source in the modern
North Atlantic flows south to the Southern Ocean, where
it combines with deep water produced in the Weddell Sea,
and increases the transport of the ACC (Figure 7a). This
in turn, feeds the northward flowing deep western boundary currents in the Indian and Pacific Oceans, and
eventually upwells in the interior of the basins. The
opening of the CAS (Figure 7b) is equivalent to introducing a source of deep water at the eastern boundary of the
North Pacific, which connects to the western boundary of
the basin through a zonal jet. This additional source of
deep water in the Pacific results in a reversal of the deep
western boundary current, which flows south and joins the
ACC.
[24] It is further conceivable that part of the NADW
could have continued from the Pacific through the Indonesian Passage and into the Indian Ocean, if the passage
was sufficiently deep. However, an investigation of this
idea requires further research as the depth of the Indonesian passage and its development through time is
uncertain.
4.1. Comparisons With Earlier Model Studies
[25] The results from the experiments presented here
differ from earlier model studies with the Hamburg OGCM
[Maier-Reimer et al., 1990]. According to the experiments
with the Hamburg model, a CAS with a sill depth of about
2700 m, results in a flow of 10 Sv from the Pacific to the
North Atlantic. This dilutes the surface salinity in the North
Atlantic by >1.0 psu, with the largest changes seen in the
subpolar North Atlantic (2.0 – 3.0 psu fresher). This
decrease in salinity prevents the formation of NADW.
Thus, no deep flow from the Atlantic to the Pacific is
observed in these experiments, due to the negligible
production of deep water in the North Atlantic, and the
concomitant reduction in the strength of the deep western
boundary current.
[26] In the experiments presented here, the North Atlantic
is freshened by about 0.5 psu at latitudes above 30N, due
to the eastward flow in the upper layers of the CAS. This
131
6-8
NISANCIOGLU ET AL.: REORGANIZATION OF MIOCENE DEEP WATER CIRCULATION
Figure 6. Total PHT in units of PW, where 1015 W = 1 PW for the Global (solid red line), Atlantic
(dashed blue line), and Indo-Pacific (dash-dotted green line) basins. Positive and negative values imply
northward and southward transport, respectively.
change is significantly smaller than that found by MaierReimer et al. [1990]. One reason for this is that the low
surface salinity in the subpolar North Atlantic of the
Hamburg model is not caused solely by the influx of Pacific
water, which has a mean salinity about 1.5 psu lower than
the Atlantic [Broecker, 1989]. The salinity is also reduced
by the import of Arctic water, as the northeastward flow in
the subpolar North Atlantic switches to southwestward. This
effect is not observed in the experiments with the MIT
OGCM, as the model does not include a representation of
the Arctic Ocean. At the same time, the northward flowing
Gulf Stream and North Atlantic Drift remain relatively
strong when the CAS is open, preventing the southward
flow of fresher water from the northernmost part of the
basin.
[27] Earlier model studies, including studies with the
Hamburg OGCM, suggest that more than one mode of
meridional overturning circulation may exist in experiments
with modern bathymetry and identical boundary conditions
[e.g., Broecker et al., 1985; Manabe and Stouffer, 1988;
Marotzke and Willebrand, 1991; Maier-Reimer et al., 1993].
To test whether the MIT OGCM can exhibit additional
steady states, experiment CAS2700 is initialized with fields
of homogeneous salinity and stratified, horizontally uniform
temperature, instead of the full temperature and salinity
fields of Levitus and Boyer [1994] and Levitus et al. [1994].
Except for these fields, all the boundary conditions remain
the same. When integrated the model converges toward the
same equilibrium state as the original experiment. Thus, the
model gives the same solution with different initial conditions, and there are no indications of secondary modes of
circulation.
[28] The test described above is not sufficient to rule out
the possible existence of multiple steady states. However,
during the course of this study, as well as numerous other
experiments with the MIT OGCM, there is no evidence of
multiple steady states of the ocean circulation. It is conceivable that the lack of a full Arctic basin in the present
model configuration reduces the magnitude of the freshening in the North Atlantic as well as the change in deep water
formation in response to an open CAS. However, it is
unlikely that the circulation would switch to a steady state
without NADW formation.
[29] The increased stability of the meridional overturning
circulation in the MIT OGCM is believed to be mainly
due to improvements in the surface boundary conditions
and the parameterization of mixing due to subgrid-scale
eddies. In the experiments with the Hamburg model,
traditional mixed boundary conditions are used, where
the freshwater flux is diagnosed from a control run with
restoring of the surface fields to observed temperatures and
salinities. The resulting freshwater flux does not agree well
132
NISANCIOGLU ET AL.: REORGANIZATION OF MIOCENE DEEP WATER CIRCULATION
Figure 7. A simplified sketch of the reorganization of the
abyssal circulation implied by the model experiments. The
interior flow has not been drawn and the only sources of
deep water are in the North Atlantic and Southern Ocean.
(a) Control experiment with ‘‘modern’’ bathymetry. (b)
Perturbed experiment with a CAS extending to depths
below the core of NADW.
with observations, and leads to a meridional overturning
circulation which is unrealistically sensitive to perturbations in surface salinity [e.g., Marotzke and Willebrand,
1991; Mikolajewicz and Maier-Reimer, 1994; Tziperman et
al., 1994].
4.2. Implications for Neogene Ocean Circulation
and Climate
[30] The experiments described in this study employ
modern continental configurations and atmospheric forcing. These conditions do not realistically represent the
climate of the Miocene and Pliocene, and the results do
not represent a simulation of ocean circulation during
these periods. Instead, the results test the sensitivity of
ocean circulation to a CAS with different depths. The
model makes several physically based predictions which
might lead to a better understanding of the role of the
CAS in ocean circulation and climate. The following
predictions can be compared with geochemical and sedimentological tracers from ocean sediment cores: (1)
NADW was formed before the final closure of the CAS,
(2) a sill depth greater than about 1000 m allows for the
passage of a westward jet of NADW into the Pacific
6- 9
Ocean, thus greatly reducing the amount of NADW transported to the South Atlantic, (3) in the western Pacific
relatively young NADW flows south as a deep western
boundary current, and (4) a shallow sill prevents the flow
of NADW to the Pacific and enhances the flow of NADW
to the South Atlantic.
4.2.1. Miocene Deep Water Circulation
[31] Data from deep sea sediment cores strongly suggests that NADW was being formed during parts of the
middle Miocene (17– 11 Ma) [Keller and Barron, 1983;
Miller and Fairbanks, 1985; Woodruff and Savin, 1989],
and that the flux of NADW increased in the late Miocene
(11– 5 Ma) [Keller and Barron, 1983; Woodruff and
Savin, 1989; Delaney, 1990; King et al., 1997], approaching modern values at the time of the final closure of the
CAS in the early Pliocene (5 – 3 Ma) [Tiedemann and
Franz, 1997; Haug and Tiedemann, 1998; Billups et al.,
1999]. While these observations are consistent with this
study, the model results also demonstrate that as the CAS
shoaled it could have caused dramatic changes to the
intermediate and deep circulation in the tropical Pacific,
and along the western boundaries of the South Pacific and
South Atlantic. It should be possible to test these findings
with tracer data from deep sediment cores in these
regions.
[32] A common procedure used to infer the relative
influences of NADW and AABW in the Atlantic basin
has been to compare paleo tracer data (such as d13C and
Cd/Ca) from core sites in the Atlantic and Pacific. Convergence of tracer data between the two basins has typically been interpreted as evidence of NADW being
replaced in the Atlantic basin by water of Southern Ocean
(Pacific) origin [e.g., Delaney, 1990; Wright et al., 1992].
However, the model experiments presented here (Figure
4b) suggest that cores in the western South Pacific and
western Atlantic could have been influenced by a common
water mass during the middle Miocene, with age and
nutrient characteristics similar to modern NADW. In other
words, convergence of geochemical tracers could indicate
NADW in both basins, not water of Pacific or Southern
Ocean origin.
[33] Delaney [1990], in a study of relative cadmium
content (Cd/Ca) of Miocene benthic foraminifera from the
deep South Atlantic (Site 289) and deep western equatorial
Pacific (Site 525), showed that the difference in Cd/Ca ratio
between the Atlantic and Pacific was negligible in the early
Miocene, small in the middle Miocene, and relatively large
in the late Miocene. This data can be interpreted to indicate
that relatively young NADW from the CAS could have
been present in the western equatorial Pacific during the
early and middle Miocene. Similarly, a study comparing
benthic foraminiferal d13C data from cores in the North
Atlantic (Deep Sea Drilling Project (DSDP) 563) and
western equatorial Pacific (DSDP 289) [Wright et al.,
1992], shows a negligible difference between North Atlantic and South Pacific d13C values during the early middle
Miocene (15 – 13 Ma), again consistent with the model
predictions.
[34] The sill depth is of great importance in controlling
the flow in the CAS. A study of the foraminiferal biostra-
133
6 - 10
NISANCIOGLU ET AL.: REORGANIZATION OF MIOCENE DEEP WATER CIRCULATION
tigraphy of the Atrato Basin in northern South America
during the Neogene (24 – 2 Ma) [Duque-Caro, 1990],
indicates that there was a major uplift of the sill to a depth
of about 1000 m during the middle Miocene, at 13– 12
Ma, disrupting the exchange of deep water between the
Atlantic and Pacific Oceans. This would be analogous to a
transition from experiments CAS2700 to CAS1000, and
would predict that relatively young NADW would no
longer be entering the Pacific via the CAS.
[35] The proposed time for the shoaling of the CAS is
close in time to observed increases in NADW formation
[Keller and Barron, 1983; Woodruff and Savin, 1989;
Delaney, 1990; King et al., 1997], as well as the late
Miocene ‘‘carbonate crash.’’ Studies from Leg 138 of the
Ocean Drilling Program (ODP) find that the east equatorial
Pacific was characterized by relatively high carbonate concentrations and accumulation rates before about 11 Ma.
However, the accumulation rates declined between 11 and
9.8 Ma, and at about 9.5 Ma, an almost complete loss of
carbonate was observed [Lyle et al., 1995; Farrell et al.,
1995]. According to Lyle et al. [1995], the carbonate crash
could not have been caused by an abrupt increase in
productivity, or by loss of organic carbon from continental
shelves. Instead, the authors conclude that the crash was
probably caused by a relatively small reduction in deep
water exchange between the Atlantic and Pacific Oceans
through the CAS [Lyle et al., 1995; Farrell et al., 1995].
This interpretation is fully consistent with this study, and
suggests that the CAS shoaled to less than 1000 m depth by
the late Miocene. At this time, the deep water exchange was
cut off and in the Pacific carbonate saturated NADW was
replaced with older, carbonate undersaturated deep water
from the Southern Ocean.
[36] Starting about 10.5 Ma, close in time to the east
equatorial carbonate crash, ODP Leg 154 sites at Ceara Rise
in the west equatorial Atlantic, experienced a permanent
increase in carbonate preservation [King et al., 1997].
Again, this agrees with the model results that predict an
increased influence of less corrosive NADW in the South
Atlantic as the CAS shoals to 1000 m depth.
4.2.2. Pliocene Climate Change
[37] While the exact sequence of tectonic events leading
to the shoaling of the CAS are uncertain [e.g., Droxler et al.,
1998; Mann, 1999], the timing of the final closure is better
constrained. An early study by Keigwin [1982], compares
d18O and d13C values of benthic and planktonic foraminifera
from DSDP sites in the eastern Pacific and western Caribbean, covering a period 8 – 2 Ma. From the d18O data it
is found that the salinity of Caribbean surface waters started
increasing 4.6 Ma. Further, the d13C difference between
the two basins increased at 6 Ma and again at 3 Ma,
approaching modern values. From these data, Keigwin
[1982] estimated that the final closure of the CAS took
place about 3 Ma. Supporting these results are data presented by Marshall et al. [1982], which show that the
exchange of marine organisms across the CAS was eliminated 3 Ma, while the exchange of terrestrial biota
between the two continents was enhanced.
[38] The proximity of the final closure of the CAS to the
onset of enhanced northern hemisphere glaciation 2.9– 2.6
Ma [Shackleton et al., 1984; Raymo, 1994; Maslin et al.,
1996; Spielhagen et al., 1997] has prompted several authors
to connect the two events [Kaneps, 1979; Keigwin, 1982;
Stanley, 1995; Haug and Tiedemann, 1998; Driscoll and
Haug, 1998]. According to the study by Haug and Tiedemann [1998], the closure of the CAS strengthened the Gulf
Stream and the transport of warm saline water to high
latitudes of the North Atlantic. This increased the production of NADW, leading to greater evaporative cooling of
surface waters and increased atmospheric moisture content.
Combined with favorable orbital obliquity, the enhanced
moisture content facilitated a buildup of ice sheets in the
Northern Hemisphere. The study by Driscoll and Haug
[1998] proposes a similar mechanism, involving enhanced
freshwater delivery to the Arctic via Siberian rivers and the
formation of sea ice.
[39] The differences between experiments CAS1000 and
CNTR imply that the production of NADW increased at the
time of the final closure of the CAS, in agreement with the
studies of Haug and Tiedemann [1998] and Driscoll and
Haug [1998]. However, the increased strength of the meridional overturning circulation also implies that additional
heat is transported to high latitudes of the North Atlantic
(Figure 6). According to the model experiments, the total
global PHT is enhanced by about 10% when the CAS is
closed, due mostly to a 30% increase in the PHT by the
North Atlantic. Indeed, Berger and Wefer [1996] have
suggested that such an increase in PHT may have delayed
the onset of Northern Hemisphere glaciation by several
million years, and may possibly have been the cause of the
early Pliocene warm period (5 – 3) [e.g., Dowsett et al.,
1996; Crowley, 1996]. Both interpretations above are consistent with the model results: It is not clear whether the
effect of the heat or the moisture fluxes dominates in
controlling the growth of Northern Hemisphere ice sheets.
To further investigate the effect of the final closure of the
CAS on the onset of Northern Hemisphere glaciation, it will
be necessary to include an atmospheric component to the
ocean model. With such a coupled ocean – atmosphere
model it will be possible to model changes to the heat
and moisture transport by the atmosphere induced by the
closure of the CAS.
5. Summary and Conclusions
[40] Experiments with the MIT OGCM were performed
to investigate the response of ocean circulation to the
shoaling and eventual closure of the CAS. Three model
experiment were conducted: A control experiment with
modern bathymetry, and two experiments with a CAS with
sill depths of 2700 and 1000 m. The model experiments
make several physically based predictions, providing a new
framework with which to interpret Miocene geochemical
tracer data:
1. Deep water is formed in the North Atlantic when the
CAS is open, in agreement with Miocene and Pliocene
geochemical tracer data. However, the formation rate is
reduced by about 10% compared to the control experiment.
2. The reduced rate of NADW formation is due to a flow
of relatively fresh water from the Pacific to the Atlantic in the
134
NISANCIOGLU ET AL.: REORGANIZATION OF MIOCENE DEEP WATER CIRCULATION
upper 1000 m of the CAS. The flow increases in strength
below the surface to a maximum at a depth of about 500 m.
3. A sill depth greater than about 1000 m allows for the
passage of a westward jet of NADW into the Pacific Ocean,
thus greatly reducing the amount of NADW transported to
the South Atlantic.
4. In the western Pacific, the NADW flows southward as
a deep western boundary current, eventually joining the
ACC. The presence of relatively young NADW in the
western Pacific is consistent with records of d13C and Cd/Ca,
which show similar values in the western Pacific and western
Atlantic during the middle Miocene.
5. In response to the reduced export of NADW to the
South Atlantic, the amount of imported South Atlantic
surface water is reduced. The result is a warming of surface
waters in the South Atlantic, and an increase in southward
heat transport.
6 - 11
6. As the CAS shoals, the flow of NADW to the Pacific
is prevented, and the flow of NADW to the South Atlantic
is enhanced, creating the modern deep water circulation
pattern. This change in the path of relatively young
carbonate saturated NADW, is consistent with sediment
core data. The data show a loss of carbonate in the east
equatorial Pacific and an increase in carbonate preservation
in the west equatorial Atlantic at the middle to late Miocene
boundary, when the CAS is believed to have shoaled to
intermediate depths.
[41] Acknowledgments. We are grateful to Veronique Bugnion, Baylor Fox-Kemper, Jeff Scott, John Marshall, Jochem Marotzke, and Ed
Boyle for their encouragement and helpful comments. We would also like
to thank Tom Crowley and Uwe Mikolajewicz for their valuable comments
on an earlier version of this paper. Funding for M.E.R. was provided by
MGG NSF grant OCE-0049011.
References
Baumgartner, A., and E. Reichel, The World
Water Balance, 179 pp., Elsevier Sci., New
York, 1975.
Berger, H. W., and G. Wefer, Expedition into the
past: Paleoceanographic studies in the South
Atlantic, in The South Atlantic: Present and
Past Circulation, edited by G. Wefer et al.,
pp. 363 – 410, Springer-Verlag, New York,
1996.
Bifllups, K., A. C. Ravelo, J. C. Zachos, and
R. D. Norris, Link between oceanic heat transport, thermohaline circulation, and the Intertropical Convergence Zone in the early Pliocene
Atlantic, Geology, 27(4), 319 – 322, 1999.
Broecker, W. S., The salinity contrast between
the Atlantic and Pacific Oceans during glacial time, Paleoceanography, 4, 207 – 212,
1989.
Broecker, W. S., D. M. Peteet, and D. Rind, Does
the ocean – atmosphere system have more than
one stable mode of operation, Nature, 315,
21 – 26, 1985.
Bryan, K., Accelerating the convergence to equilibrium of ocean climate models, J. Phys.
Oceanogr., 14, 666 – 673, 1984.
Coates, A. G., J .B. C. Jackson, J. S. Collins, T. M.
Cronin, H. J. Dowsett, L. M. Bybell, P. Jung,
and J. A. Obando, Closure of the Isthmus of
Panama: The near-shore marine record of Costa
Rica and western Panama, Geol. Soc. Am. Bull.,
104, 814 – 828, 1992.
Crowley, T. J., Pliocene climates: The nature of
the problem, Mar. Micropaleontol., 27, 3 – 12,
1996.
Danabasoglu, G., J. C. McWilliams, and W. G.
Large, Approach to equilibrium in accelerated
global oceanic models, J. Clim., 9, 1092 –
1110, 1996.
Delaney, L. M., Miocene benthic foraminiferal
Cd/Ca records: South Atlantic and western
equatorial Pacific, Paleoceanography, 5, 743 –
760, 1990.
Dowsett, H., J. Barron, and R. Poore, Middle
Pliocene sea surface temperatures: A global
reconstruction, Mar. Micropaleontol., 27,
13 – 25, 1996.
Driscoll, N. W., and G. H. Haug, A short circuit
in thermohaline circulation: A cause for Northern Hemisphere glaciation?, Science, 282,
436 – 438, 1998.
Droxler, A. W., K. C. Burke, A. D. Cunningham,
A. C. Hine, E. Rosencrantz, D. S. Duncan, P.
Hallock, and E. Robinson, Caribbean constraints on circulation between Atlantic and
Pacific oceans over the past 40 million years,
in Tectonic Boundary Conditions for Climate
Reconstructions, edited by T. J. Crowley and
K. C. Burke, pp. 169 – 191, Oxford Univ.
Press, New York, 1998.
Duque-Caro, H., Neogene stratigraphy, paleoceanography and paleobiogeography in northwest
South America and the evolution of the Panama Seaway, Palaeogeogr. Palaeoclimatol.
Palaeoecol., 77, 203 – 234, 1990.
Farrell, J. W., I. Raffi, T. R. Janacek, D. W. Murray, M. Levitan, K. A. Dadey, K.-C. Emeis, M.
Lyle, J.-A. Flores, and S. Hovan, Late Neogene
sedimentation patterns in the eastern equatorial
Pacific Ocean, in Proc. Ocean Drill. Program
Sci. Results, vol. 138, edited by N. G. Pisias et
al., pp. 717 – 756, Ocean Drill. Program, College Station, Tex., 1995.
Ganachaud, A. S., Large scale oceanic circulation and fluxes of freshwater, heat, nutrients
and oxygen, doctoral dissertation thesis, Mass.
Inst. of Technol., Cambridge, Mass., 1999.
Ganachaud, A. S., and C. Wunsch, Improved
estimates of global ocean circulation, heat
transport and mixing from hydrographic data,
Nature, 408, 453 – 456, 2000.
Gent, P. R., and J. C. McWilliams, Isopycnal
mixing in ocean circulation models, J. Phys.
Oceanogr., 20, 150 – 155, 1990.
Griffies, S. M., The Gent – McWilliams skew
flux, J. Phys. Oceanogr., 28, 831 – 841, 1998.
Haug, G. H., and R. Tiedemann, Effect of the
formation of the Isthmus of Panama on Atlantic Ocean thermohaline circulation, Nature,
393, 673 – 676, 1998.
Jiang, S., P. H. Stone, and P. Malanotte-Rizzoli, An assessment of the Geophysical Fluid
Dynamics Laboratory ocean model with
coarse resolution: Annual-mean climatology,
J. Geophys. Res., 104(C11), 25,623 – 25,645,
1999.
Kamenkovich, I. V., A. Sokolov, and P. H. Stone,
An efficient climate model with a 30 ocean and
statistical-dynamical atmosphere, Clim. Dyn.,
19, 585 – 598, 2002.
Kaneps, A. G., Gulf Stream: Velocity fluctuations during the late Cenozoic, Science, 204,
297 – 301, 1979.
Keigwin, L., Isotopic paleoceanography of the
Caribbean and East Pacific: Role of Panama
135
uplift in Late Neogene time, Science, 217,
350 – 352, 1982.
Keller, G., and J. A. Barron, Paleoceanographic
implications of the Miocene deep-sea hiatuses,
Geol. Soc. Am. Bull., 94, 590 – 613, 1983.
King, T. A., W. G. Ellis Jr., D. W. Murray,
N. J. Shackleton, and S. Harris, Miocene
evolution of carbonate sedimentation at the
Ceara rise: A multivariate data/proxy approach, in Proc. Ocean Drill. Program Sci.
Results, edited by N. J. Shackleton et al., pp.
349 – 365, Ocean Drill. Program, College Station, Tex., 1997.
Levitus, S., and T. P. Boyer, World Ocean Atlas
1994 Volume 4: Temperature, 117 pp., U.S.
Dept. of Commer., Washington, D. C., 1994.
Levitus, S., R. Burgett, and T. P. Boyer, World
Ocean Atlas 1994 Volume 3: Salinity, 99 pp.,
U.S. Dept. of Commer., Washington, D. C.,
1994.
Lyle, M., K. A. Dadey, and J. W. Farrell, The
Late Miocene (11 – 8 Ma) eastern Pacific carbonate crash: Evidence for reorganization of
deep-water circulation by the closure of the
Panama gateway, in Proc. Ocean Drill. Program Sci. Results, edited by N. G. Pisias et al.,
pp. 821 – 838, Ocean Drill. Program, College
Station, Tex., 1995.
Maier-Reimer, E., U. Mikolajewicz, and T. Crowley, Ocean general circulation model sensitivity
experiment with an open Central American
isthmus, Paleoceanography, 5, 349 – 366,
1990.
Maier-Reimer, E., U. Mikolajewicz, and K. Hasselmann, Mean circulation of the Hamburg
LSG OGCM and its sensitivity to the thermohaline surface forcing, J. Phys. Oceanogr., 23,
731 – 757, 1993.
Manabe, S., and R. J. Stouffer, Two stable equilibria of a coupled ocean – atmosphere model,
J. Clim., 1, 841 – 866, 1988.
Mann, P., Caribbean sedimentary basins: Classification and tectonic setting from Jurassic to
present, in Caribbean Basins, edited by P.
Mann, pp. 3 – 31, Elsevier Sci., New York,
1999.
Marotzke, J., and J. Willebrand, Multiple equilibria of the global thermohaline circulation,
J. Phys. Oceanogr., 21, 1372 – 1385, 1991.
Marshall, J., and F. Schott, Open-ocean convection: Observations, theory, and models, Rev.
Geophys., 37(1), 1 – 64, 1999.
6 - 12
NISANCIOGLU ET AL.: REORGANIZATION OF MIOCENE DEEP WATER CIRCULATION
Marshall, J., A. Adcroft, C. Hill, L. Perelman,
and C. Heisey, A finite-volume, incompressible Navier Stokes model for studies of the
ocean on parallel computers, J. Geophys.
Res., 102(C3), 5753 – 5766, 1997a.
Marshall, J., C. Hill, L. Perelman, and A. Adcroft, Hydrostatic, quasi-hydrostatic, and nonhydrostatic ocean modeling, J. Geophys. Res.,
102(C3), 5733 – 5752, 1997b.
Marshall, L. G., S. D. Webb, J. J. Sepkoski, and
D. M. Raup, Mammalian evolution and the
great American interchange, Science, 215,
1351 – 1357, 1982.
Maslin, M. A., G. H. Haug, M. Sarntheim, and
R. Tiedemann, The progressive intensification
of Northern Hemisphere glaciation as seen
from the North Pacific, Geol. Rundsch., 85,
452 – 465, 1996.
Mikolajewicz, U., and T. J. Crowley, Response of
a coupled ocean/energy balance model to restricted flow through the central American isthmus, Paleoceanography, 12, 429 – 441, 1997.
Mikolajewicz, U., and E. Maier-Reimer, Mixed
boundary-conditions in ocean general-circulation models and their influence on the stability
of the models conveyor belt, J. Geophys. Res.,
99(C11), 22,633 – 22,644, 1994.
Mikolajewicz, U., E. Maier-Reimer, T. J. Crowley, and K. Y. Kim, Effect of Drake and Panamanian Gateways on the circulation of an ocean
model, Paleoceanography, 8, 409 – 426, 1993.
Miller, K. G., and R. G. Fairbanks, Oligocene to
Miocene carbon isotope cycles and abyssal circulation changes, in The Carbon Cycle and
Atmospheric CO2: Natural Variations Archean
to Present, edited by E. T. Sundquist and W. S.
Broecker, pp. 469 – 486, AGU, Washington,
D. C., 1985.
Perry, G. D., P. B. Duffy, and N. L. Miller, An
extended data set of river discharges for vali-
dation of general circulation models, J. Geophys. Res., 101(D16), 21,339 – 21,349, 1996.
Raymo, M. E., The initiation of Northern Hemisphere glaciation, Annu. Rev. Earth Planet.
Sci., 22, 353 – 383, 1994.
Redi, M. H., Oceanic isopycnal mixing by coordinate rotation, J. Phys. Oceanogr., 12, 1154 –
1158, 1982.
Reeh, N., Calving from Greenland glaciers: Observations, balance estimates of calving rate,
calving laws, in Workshop on the Calving Rate
of West Greenland Glaciers in Response to
Climate Change, p. 171, Dan. Polar Cent., Copenhagen, 1994.
Schmitt, R. W., P. S. Bogden, and C. E. Dorman,
Evaporation minus precipitation and density
fluxes for the North Atlantic, J. Phys. Oceanogr., 19, 1208 – 1221, 1989.
Shackleton, N. J., et al., Oxygen isotope calibration of the onset of ice-rafting and history of
glaciation in the North Atlantic region, Nature,
307, 620 – 623, 1984.
Spielhagen, R. F., et al., Arctic ocean evidence
for late Quaternary initiation of northern Eurasian ice sheets, Geology, 25(9), 783 – 786,
1997.
Stanley, S. M., New horizons for paleontology
with two examples: The rise and fall of the
Cretaceous supertethys and the cause of the
modern ice age, J. Paleontol., 69(6), 999 –
1007, 1995.
Stommel, H., and A. B. Arons, On the abyssal
circulation of the world ocean, 1, Stationary
planetary flow patterns on a sphere, Deep
Sea Res., 6, 140 – 154, 1960a.
Stommel, H., and A. B. Arons, On the abyssal
circulation of the world ocean, 2, An idealized
model of the circulation pattern and amplitude
in oceanic basins, Deep Sea Res., 6, 217 – 233,
1960b.
136
Tiedemann, R., and S. O. Franz, Deep-water
circulation, chemistry, and terrigenous sediment supply in the equatorial Atlantic during
the Pliocene, 3.3 – 2.6 Ma and 5 – 4.5 Ma, in
Proc. Ocean Drill. Program Sci. Results, edited by N. J. Shackleton et al., pp. 299 – 318,
Ocean Drill. Program, College Station, Tex.,
1997.
Trenberth, K. E., and A. Solomon, The global
heat-balance: Heat transports in the atmosphere and ocean, Clim. Dyn., 10(3), 107 –
134, 1994.
Trenberth, K. E., J. G. Olson, and W. G. Large, A
global ocean wind stress climatology based on
ECMWF analysis, NCAR Tech. Note, NCAR/
TN-338+STR, Natl. Cent. for Atmos. Res.,
Boulder, Colo., 1989.
Tziperman, E., J. R. Toggweiler, Y. Feliks, and K.
Bryan, Instability of the thermohaline circulation with respect to mixed boundary-conditions:
Is it really a problem for realistic models,
J. Phys. Oceanogr., 24, 217 – 232, 1994.
Woodruff, F., and S. Savin, Miocene deep-water
oceanography, Paleoceanography, 4, 87 – 140,
1989.
Wright, J. D., K. G. Miller, and R. G. Fairbanks,
Early and middle Miocene stable isotopes: Implications for deepwater circulation and climate, Paleoceanography, 7, 357 – 389, 1992.
K. H. Nisancioglu and P. H. Stone, Program in
Atmosphere, Oceans, and Climate, Department
of Earth, Atmospheric, and Planetary Sciences,
Massachusetts Institute of Technology, Cambridge, MA 02141, USA. ([email protected];
[email protected])
M. E. Raymo, Department of Earth Sciences,
Boston University, 685 Commonwealth Avenue,
Boston, MA 02215, USA. ([email protected])
Bibliography
Andrews, J. T., The Wisconsin Laurentide ice sheet: Dispersal centers, problems og rates of retreat,
and climatic implications, Arctic Alpine Research, 5, 185–199, 1973.
Barrett, P., C. Adams, W. McInstosh, C. Swisher, and G. Wilson, Geochronological evidence supporting Antarctic deglaciation 3 million years ago, Nature, 359(6398), 816 – 818, 1992.
Berger, A., Long-term variation of the daily and monthly insolation during the Last Ice Age.,
Transactions of the American Geophysical Union, 57, 254, 1976.
Berger, A., Long-term variations of daily insolation and quaternary climatic changes, Journal of
the Atmospheric Sciences, 35(12), 2362 – 2367, 1978a.
Berger, A., Long-term variations of caloric insolation resulting from earths orbital elements, Quaternary Research, 9(2), 139 – 167, 1978b.
Berger, A., and M. Loutre, Long-term climatic variations: data and modelling, NATO ASI series.
Series I, Global environmental change, vol. 22, chap. Precession, eccentricity, obliquity, insolation and paleocilmates, Springer-Verlag, Berlin, New York, proceedings of the NATO Advanced
Study Institute on Long-Term Climatic Variations–Data and Modelling, Siena, Italy, September
27-October 11, 1992, 1994.
Berger, A., M. Loutre, and C. Tricot, Insolation and earths orbital periods, Journal of Geophysical
Research - Atmospheres, 98(D6), 10,341 – 10,362, 1993.
137
Berger, A., X. S. Li, and M. F. Loutre, Modelling northern hemisphere ice volume over the last 3
Ma, Quaternary Science Reviews, 18(1), 1–11, 1999.
Berger, W. H., and G. Wefer, Expedition into the Past: Paleoceanographic Studies in the South
Atlantic, pp. 363–410, Springer-Verlag, Berlin Heidelberg, 1996.
Berggren, W. A., F. J. Hilgen, C. G. Langereis, D. V. Kent, J. D. Obradovich, I. Raffi, M. E.
Raymo, and N. J. Shackleton, Late Neogene Chronology - New Perspectives in High-Resolution
Stratigraphy, Geological Society of America Bulletin, 107(11), 1272–1287, 1995.
Birchfield, G., J. Weertman, and A. Lunde, A paleoclimate model of northern hemispheric ice
sheets, Quaternary Research, 15(126-142), 1981.
Birchfield, G., J. Weertman, and A. Lunde, A model study of the role of high-latitude topography
in the climatic response to orbital insolation anomalies, Journal of the Atmospheric Sciences,
39(1), 71 – 87, 1982.
Braithwaite, R., and Y. Zhang, Sensitivity of mass balance of five Swiss glaciers to temperature
changes assessed by tuning a degree-day model, Journal og Glaciology, 46(152), 7 – 14, 2000.
Braithwaite, R. J., Positive degree-day factors for ablation on the Greenland ice sheet studied by
energy-balance modelling, Journal of Glaciology, 41(137), 153–160, 1995.
Braithwaite, R. J., and O. B. Olesen, Calculation of glacier ablation from temperature, West Greenland., pp. 219–223, Dordrecht, Kluwer Academic Publishers, 1989.
Bromwich, D., and F. Robasky, Recent precipitation trends over the polar ice sheets, Meteorology
and Atmospheric Physics, 51(3-4), 259 – 274, 1993.
Brown, C., M. Meier, and A. Post, Calving speed of Alaska tidewater glaciers, with application to
Columbia Glacier, Tech. Rep. 1258-C, U.s. Geological Survey Professional Paper, 1982.
Budyko, M. I., The effect of solar radiation variations on the climate of the Earth, Tellus, 5, 611–
619, 1969.
138
Cande, S. C., and D. V. Kent, A New Geomagnetic Polarity Time Scale For the Late Cretaceous
and Cenozoic, Journal of Geophysical Research-Solid Earth, 97(B10), 13,917–13,951, 1992.
Cande, S. C., and D. V. Kent, Revised Calibration of the Geomagnetic Polarity Timescale For the
Late Cretaceous and Cenozoic, Journal of Geophysical Research-Solid Earth, 100(B4), 6093–
6095, 1995.
Cane, M., and P. Molnar, Closing of the Indonesian seaway as a precursor to east African aridircation around 3-4 million years ago, Nature, 411(6834), 157 – 162, 2001.
Cess, R., Climate change - appraisal of atmospheric feedback mechanisms emplying zonal climatology, Journal of the Atmospheric Sciences, 33(10), 1831 – 1843, 1976.
Charles, C., D. Rind, J. JouzelL, R. Koster, and R. Fairbanks, Glacial-interglacial changes in moisture sources for Greenland - influences on the ice core record of climate, Science, 263(5146),
508 – 511, 1994.
Charney, J., The dynamics of long waves in a baroclinic westerly current, Journal of Meteorology,
1947.
Clapperton, C., and D. Sugden, Late Cenozoic glacial history of the Ross embayment, Antarctica,
Quatenary Science Reviews, 9(2-3), 253 – 272, 1990.
Clark, P., and D. Pollard, Origin of the middle Pleistocene transition by ice sheet erosion of regolith,
Paleoceanography, 13(1), 1 – 9, 1998.
Clark, P. U., R. B. Alley, and D. Pollard, Climatology - Northern hemisphere ice-sheet influences
on global climate change, Science, 286(5442), 1104–1111, 1999.
Clement, B. M., and D. V. Kent, Geomagnetic Polarity Transition Records From 5 Hydraulic Piston
Core Sites in the North-Atlantic, Initial Reports of the Deep Sea Drilling Project, 94, 831–852,
1987.
139
Coakley, J., Study of climate sensitivity using a simple energy-balance model, Journal of the Atmospheric Sciences, 36(2), 260 – 269, 1979.
Cortijo, E., S. Lehman, L. Keigwin, M. Chapman, D. Paillard, and L. Labeyrie, Changes in meridional temperature and salinity gradients in the North Atlantic Ocean (30 degrees-72 degrees N)
during the last interglacial period, Paleoceanography, 14(1), 23–33, 1999.
Croll, J., Climate and Time, Appleton, 1875.
Crowley, T. J., Pliocene climates: The nature of the problem, Marine Micropaleontology, 27, 3–12,
1996.
Crowley, T. J., and G. R. North, Paleoclimatology, Oxford Monographs on Geology and Geophysics, vol. 18, 349 pp., Oxford University Press, New York, 1991.
Cubasch, U., and G. Meehl, Projections of Future Climate Change, chap. 9, p. 881, Cambridge
University Press, Cambridge, UK, 2001.
Cuffey, K., and G. Clow, Temperature, accumulation, and ice sheet elevation in central Greenland
through the last deglacial transition, Journal of Geophysical Research-Oceans, 102(C12), 26,383
– 26,396, 1997.
Cuffey, K., G. Clow, R. Alley, M. Stuiver, E. Waddington, and R. Saltus, Large Arctic temperaturechange at the Wisconsin-Holocene glacial transition, Science, 270(5235), 455 – 458, 1995.
Dansgaard, W., The O18 abundance in fresh water, Geochimica et Cosmochimica Acta, 6, 241–260,
1954.
Dansgaard, W., Stable isotopes in precipitation, Tellus, 16, 436–468, 1964.
Deblonde, G., W. R. Peltier, and W. T. Hyde, Simulations of Continental Ice-Sheet Growth Over
the Last Glacial Interglacial Cycle - Experiments With a One Level Seasonal Energy-Balance
Model Including Seasonal Ice Albedo Feedback, Global and Planetary Change, 98(1), 37–55,
1992.
140
Denton, G., M. Prentice, D. Kellogg, and T. Kellogg, Late Tertirary history of the Antarctic icesheet - evidence from the dry valleys, Geology, 12(5), 263 – 267, 1984.
Dowsett, J. H, R. S. Thompson, J. A. Barron, T. M. Cronin, S. E. Ishman, R. Z. Poore, D. A. Willard,
and J. T. R. Holtz, Joint investigation of the Middle Pliocene climate I: PRISM palaeoenvironmental reconstructions, Palaeogeography Palaeoclimatology Palaeoecology, 9, 1994.
Dowsett, H., J. Barron, and R. Poore, Middle Pliocene sea surface temperatures: A global reconstruction, Marine Micropaleontology, 27(1-4), 13–25, 1996.
Driscoll, N. W., and G. H. Haug, A short circuit in thermohaline circulation: A cause for northern
hemisphere glaciation?, Science, 282(5388), 436–438, 1998.
Ellis, J., and T. Vonder Haar, Zonal average earth radiation budget measurements from satellites for
climate studies, Tech. rep., Colorado State University, Fort Collins, 1976.
Emiliani, C., Pleistocene temperatures, Journal of Geology, 63(6), 538–578, 1955.
Emiliani, C., The Pleistocene epoch and the evolution of man, Current Anthropology, 9(1), 27–47,
1968.
Gallee, H., J. P. Vanypersele, T. Fichefet, C. Tricot, and A. Berger, Simulation of the Last Glacial
Cycle By a Coupled, Sectorially Averaged Climate-Ice Sheet Model .1. the Climate Model,
Journal of Geophysical Research-Atmospheres, 96(D7), 13,139–13,161, 1991.
Gallee, H., J. P. Vanypersele, T. Fichefet, I. Marsiat, C. Tricot, and A. Berger, Simulation of the
Last Glacial Cycle By a Coupled, Sectorially Averaged Climate-Ice Sheet Model .2. Response to
Insolation and Co2 Variations, Journal of Geophysical Research-Atmospheres, 97(D14), 15,713–
15,740, 1992.
Gallimore, G. R, and J. E. Kutzbach, Snow cover and sea ice sensitivity to generic changes in Earh
orbital parameters, Journal of Geophysical Research, 100(D1), 1103–1120, 1995.
141
Gallup, C., H. Cheng, F. Taylor, and R. Edwards, Direct determination of the timing of sea level
change during termination II, Science, 295(5553), 310 – 313, 2002.
Ghil, M., Cryothermodynamics - the Chaotic Dynamics of Paleoclimate, Physica D Nonlinear
Phenomena, 77(1-3), 130–159, 1994.
Ghil, M., M. Allen, M. Dettinger, K. Ide, D. Kondrashov, M. Mann, A. Robertson, A. Saunders,
Y. Tian, F. Varadi, and P. Yiou, Advanced spectral methods for climatic time series, Rviews of
Geophysics, 40(1), 2002.
Gildor, H., and E. Tziperman, Sea ice as the glacial cycles’ climate switch: Role of seasonal and
orbital forcing, Paleoceanography, 15(6), 605–615, 2000.
Gildor, H., and E. Tziperman, A sea ice climate switch mechanism for the 100-kyr glacial cycles,
Journal of Geophysical Research-Oceans, 106(C5), 9117–9133, 2001.
Green, J. S. A., Transfer properties of the large-scale eddies and the general circulation of the
atmosphere, Quarterly Journal of the Royal Meteorological Society1, 96(408), 157–185, 1970.
Haney, R., Surface thermal boundary condition of ocean circulation model, Journal of Physical
Oceanography, 1, 145–167, 1971.
Hasselmann, K., Construction and verification of stochastic climate models, pp. 481–497, D. Reidel Publishing Company, 1981.
Haug, G. H., and R. Tiedemann, Effect of the formation of the Isthmus of Panama on Atlantic
Ocean thermohaline circulation, Nature, 393(6686), 673–676, 1998.
Hays, J. D., J. Imbrie, and N. J. Shackleton, Variations in the Earth’s orbit: Pacemakers of the ice
ages, Science, 194(4270), 1121–1132, 1976.
Held, I., and M. Suarez, Simple albedo feedback models of ice-caps, Tellus, 26(6), 613 – 629, 1974.
142
Held, I. M., The vertical scale of an unstable baroclinic wave and its importance for eddy heat flux
parameterizations, Journal of the Atmospheric Sciences, 35, 572–576, 1978.
Henderson, G., and N. Slowey, Evidence from U-Th dating against Northern Hemisphere forcing
of the penultimate deglaciation, Nature, 404(6773), 61 – 66, 2000.
Herbert, T. D., J. D. Schuffert, D. Andreasen, L. Heusser, M. Lyle, A. Mix, A. C. Ravelo, L. D.
Stott, and J. C. Herguera, Collapse of the California Current during glacial maxima linked to
climate change on land, Science, 293(5527), 71–76, 2001.
Hughes, T., North America and Adjacent Oceans During the Last Deglaciation, The Geology of
North America, vol. K-3, chap. Ice dynamics and deglaction models when ice sheets collapsed,
pp. 183–220, Geological Society of America, Boulder, Colorado, 1987.
Hughes, T., G. Denton, and M. Grosswald, Was there a late-Wurm Arctic ice sheet?, Nature, 266,
596–602, 1977.
Huybers, P., and C. Wunsch, Rectification and precession signals in the climate system, Geophysical Research Letters, 30(19), 2003.
Huybrechts, A 3-D model for the Antarctic ice sheet:: a sensitivity study on the glacial-interglacial
contrast, Climate Dynamics, 5, 79–92, 1990.
Huybrechts, P., and J. de Wolde, The dynamic response of the Greenland and Antarctic ice sheets
to multiple-century climatic warming, Journal of Climate, 12(8), 2169 – 2188, 1999.
Huybrechts, P., A. Letreguilly, and N. Reeh, The Greenland Ice-Sheet and Greenhouse Warming,
Global and Planetary Change, 89(4), 399–412, 1991.
Huybrechts, P., D. Steinhage, F. Wilhelms, and J. Bamber, Balance velocities and measured properties of the Antarctic ice sheet from a new compilation of gridded data for modelling, Annals of
Glaciology, 30, 52 – 60, 2000.
143
Hyde, W., and W. Peltier, Sensitivity experiments with a model of the ice-age cycle - the response
to Milankovich forcing, Journal of the Atmospheric Sciences, 44(10), 1351 – 1374, 1987.
Imbrie, J., E. A. Boyle, S. C. Clemens, A. Duffy, W. R. Howard, G. Kukla, J. Kutzbach, D. G.
Martinson, A. McIntyre, A. C. Mix, B. Molfino, J. J. Morley, L. C. Peterson, N. G. Pisias,
W. L. Prell, M. E. Raymo, N. J. Shackleton, and J. R. Toggweiler, On the Structure and Origin
of Major Glaciation Cycles .1. Linear Responses to Milankovitch Forcing, Paleoceanography,
7(6), 701–738, 1992.
Imbrie, J., A. Berger, E. A. Boyle, S. C. Clemens, A. Duffy, W. R. Howard, G. Kukla, J. Kutzbach,
D. G. Martinson, A. McIntyre, A. C. Mix, B. Molfino, J. J. Morley, L. C. Peterson, N. G. Pisias,
W. L. Prell, M. E. Raymo, N. J. Shackleton, and J. R. Toggweiler, On the Structure and Origin of
Major Glaciation Cycles .2. the 100,000-Year Cycle, Paleoceanography, 8(6), 699–735, 1993.
Jansen, E., and J. Sjoholm, Reconstruction of glaciation over the past 6 myr from ice-borne deposits
in the Norwegian Sea, Nature, 349(6310), 600 – 603, 1991.
Jansen, E., U. Bleil, R. Henrich, L. Kringstad, and B. Slettemark, Paleoenvironmental changes
in the Norwegian Sea and the northeast Atlantic during the last 2.8 m.y.: Deep Sea Drilling
Project/Ocean Drilling Program sites 610, 642, 643, and 644, Paleoceanography, 3, 563–581,
1988.
Johnson, R. G., Major Northern Hemisphere deglaciation caused by a moisture deficit 140 ka,
Geology, 19(7), 686 – 689, 1991.
Joussaume, S., and P. Braconnot, Sensitivity of paleoclimate simulation results to season definitions, Journal of Geophysical Research-Atmospheres, 102(D2), 1943 – 1956, 1997.
Kageyama, M., F. D’Andrea, G. Ramstein, P. Valdes, and R. Vautard, Weather regimes in past
climate atmospheric general circulation model simulations, Climate Dynamics, 15(10), 773 –
793, 1999.
144
Kallen, E., C. Crafoord, and M. Ghil, Free Oscillations in a Climate Model With Ice-Sheet Dynamics, Journal of the Atmospheric Sciences, 36(12), 2292–2303, 1979.
Kalnay, E., M. Kanamitsu, R. Kistler, W. Collins, D. Deaven, L. Gandin, M. Iredell, S. Saha,
G. White, J. Woollen, Y. Zhu, M. Chelliah, W. Ebisuzaki, W. Higgins, J. Janowiak, K. Mo,
C. Ropelewski, J. Wang, A. Leetmaa, R. Reynolds, R. Jenne, and D. Joseph, The NCEP/NCAR
40-year reanalysis project, Bulletin of the American Meteorological Society, 77(3), 437 – 471,
1996.
Kaneps, A. G., Gulf Stream: Velocity fluctuations during the late Cenozoic., Science, 204, 297–
301, 1979.
Kaspner, W., R. Alley, C. Shuman, S. Anandakrishnan, and P. Grootes, Dominant influence of
atmospheric circulation on snow accumulation in Greenland over the past 18,000 years, Nature,
373(6509), 52 – 54, 1995.
Keigwin, L., Isotopic paleoceanography of the Caribbean and East Pacific: role of Panama uplift
in Late Neogene time, Science, 217, 350–352, 1982.
Kennett, J., Cenozoic evolution of Antarctic glaciation, cirum-Antarctic ocean, and their impacet on
global paleoceanography, Journal of Geophysical Research - Oceans and Atmospheres, 82(27),
3843 – 3860, 1977.
Kennett, J. P., and D. A. Hodell, Evidence for Relative Climatic Stability of Antarctica during
the Early Pliocene: A Marine Perspective, Geografiska Annaler. Series A. Physical Geography,
75(4), 205–220, 1993.
Khodri, M., Y. Leclainche, G. Ramstein, P. Braconnot, O. Marti, and E. Cortijo, Simulating the
amplification of orbital forcing by ocean feedbacks in the last glaciation, Nature, 410(6828),
570–574, 2001.
Kim, S., T. Crowley, and A. Stossel, Local orbital forcing of Antarctic climate change during the
last interglacial, Sciece, 280(5364), 728 – 730, 1998.
145
Kominz, M., and N. Pisias, Pleistocene climate - deterministic or stochastic, Science, 204(4389),
171 – 173, 1979.
Köppen, W. P., and A. Wegener, Die klimate der geologischen vorzeit, Gebrüder Borntraeger, Berlin, 1924.
Krantz, D., A chronology of Pliocene sea-level flucturations - the United-Sates middle Atlantic
coastal-plain record, Quaternary Science Reviews, 10(2-3), 163 – 174, 1991.
Kukla, J., The Pleistocene epoch and the evolution of man - a reply, Current Anthropology, 9,
27–47, 1968.
Kutzbach, J., R. Bryson, and W. Shen, An evaluation of the thermal Rossby number in the Pleistocene, Meteorological Monographs, 8(3), 134–138, 1968.
Laskar, J., F. Joutel, and F. Boudin, Orbital, precessional, and insolation quantities for the earth
from -20 myr to +10 myr, Astronomy and Astrophysics, 270(1-2), 522 – 533, 1993.
Lee, W., and G. North, Small ice cap instability in the presence of fluctuations, Climate Dynamics,
11(4), 242 – 246, 1995.
Leovy, C., Exchange of water vapor between atmosphere and surface of Mars, Icarus, 18(1), 120 –
125, 1973.
Letreut, H., and M. Ghil, Orbital forcing, climatic interactions, and glaciation cycles, Journal of
Geophysical Research - Oceans and Atmospheres, 88(NC9), 5167 – 5190, 1983.
Li, Z., and H. Leighton, Global climatologies of solar-radiation budgets at the surface and in the
atmosphere from 5 years of ERBE data, Journal of Geophysical Research - Atmospheres, 98(D3),
4919 – 4930, 1993.
Lian, M., and R. Cess, Energy-balance climate models - reappraisal of ice-albedo feedback, Journal
of the Atmospheric Sciences, 34(7), 1058 – 1062, 1977.
146
Lin, C., Eddy heat fluxes and stability of planetary-waves .1., Journal of the Atmospheric Sciences,
37(11), 2353 – 2372, 1980a.
Lin, C., Eddy heat fluxes and stability of planetary-waves .2., Journal of the Atmospheric Sciences,
37(11), 2373 – 2380, 1980b.
Lorenz, E., Forced and free variations of weather and climate, Journal of the Atmospheric Sciences,
36(8), 1367 – 1376, 1979.
Maier-Reimer, E., U. Mikolajewicz, and T. Crowley, Ocean general circulation model sensivity
experiment with an open central american isthmus, Paleoceanography, 5(3), 349–366, 1990.
Milankovitch, M., Kalorische Jahreszeiten und deren Anwendung im palaoklimalen Problem,
Bericht der konigl. serbischen Akademie, CIX, 1923.
Milankovitch, M., Handbuch der klimatologie (editors Koppen, W. , Geiger, R.), Gebruder
Borntraeger, Berlin, 1936.
Milankovitch, M., Canon of insolation and the ice-age problem. K. Serb. Akad. Geogr., Spececial
Publication, (132), 484, translated by Israel Program for Scientific Translations, Jerusalem,
1969, US. Depte of Commerce, 1941.
Miller, K. G., and R. G. Fairbanks., Oligocene to Miocene carbon isotope cycles and abyssal
circulation changes, pp. 469–486, Geophysical Monograph Series, Washington D.C., 1985.
Myhre, G., E. Highwood, K. Shine, and F. Stordal, New estimates of radiative forcing due to well
mixed greenhouse gases, Geophysical Research Letters, 25(14), 2715 – 2718, 1998.
Nakamura, M., P. H. Stone, and J. Marotzke, Destabilization of the Thermohaline Circulation By
Atmospheric Eddy Transports, Journal of Climate, 7(12), 1870–1882, 1994.
Nisancioglu, K., M. Raymo, and P. Stone, Reorganization of Miocene deep water circulation in
response to the shoaling of the Central American Seaway, Paleoceanography, 18(1), art. no. –
1006, 2003.
147
North, G., The small ice cap instability in diffusive climate models, Journal of the Atmospheric
Sciences, 41(23), 3390 – 3395, 1984.
North, G., and J. Coakley, Differences between seasonal and mean annual energy-balance model
calculations of climate and climate sensitivity, Journal of the Atmospheric Sciences, 36(7), 1189
– 1204, 1979.
North, G., and T. Crowley, Application of a seasonal climate model to Cenozoic glaciation, Journal
of the Geological Society, 142, 475 – 482, 1985.
North, G., R. Cahalan, and J. Coakley, Energy-blance climate models, Reviews of Geophysics,
19(1), 91 – 121, 1981.
North, G. R., Theory of energy-balance climate models, Journal of the atmospheric sciences,
32(11), 2033–2041, 1975.
North, G. R., J. G. Mengel, and D. A. Short, Simple Energy-Balance Model Resolving the Seasons and the Continents - Application to the Astronomical Theory of the Ice Ages, Journal of
Geophysical Research-Oceans and Atmospheres, 88(C11), 6576–6586, 1983.
Nye, J. F., The flow of glaciers and ice-sheets as a problem in plasticity, Proceedings of the Royal
Society of London, Series A: Mathematical and Physical Sciences, 207(1091), 554–572, 1951.
Oerlemans, J., Model experiments of the 100,000-yr glacial cycle, Nature, 287(5781), 430 – 432,
1980.
Ohmura, A., Physical basis for the temperature-based melt-index method, Journal of Applied Meteorology, 40(4), 753 – 761, 2001.
Ohmura, A., M. Wild, and L. Bengtsson, A possible change in mass balance of Greenland and
Antarctic ice sheets in the coming century, Journal of Climate, 9(9), 2124 – 2135, 1996.
Oort, A., NOAA Professional Paper No. 14, Tech. rep., U.S. Department of Commerce, Rockville,
Md., 1983.
148
Orlanski, I., The influence of bottom topography on the stability of jets in a baroclinic fluid, Journal
of Atmospheric Sciences, 26, 1216–1232, 1969.
Orowan, E., and M. F. Perutz, The flow of ice and other solids, Journal of Glaciology, 1(5), 231–
240, 1949.
Owen, T., R. Cess, and V. Ramanathan, Enhanced CO2 greenhouse to compensate for reduced solar
luminosity on early earth, Nature, 277(5698), 640 – 642, 1979.
Paillard, D., Glacial cycles: Toward a new paradigm, Reviews of Geophysics, 39(3), 325–346,
2001.
Patterson, W. S. B., The physics of glaciers, Pergamon, Oxford, UK, 1994.
Pedlosky, J., The stability of currents in the atmosphere and the ocean: Part I, Journal of the
Atmospheric Sciences, 21, 201–219, 1964a.
Pedlosky, J., The stability of currents in the atmosphere and the ocean: Part II, Journal of the
Atmospheric Sciences, 21, 342–353, 1964b.
Pedlosky, J., Secondary baroclinic instability and meridional scale of motion in ocean, Journal of
Physical Oceanography, 5(4), 603 – 607, 1975.
Peixoto, J. P., and A. H. Oort, Physics of Climate, American Institute of Physics, New York, 1992.
Peltier, W., Ice-age paleotopography, Science, 265(5169), 195 – 201, 1994.
Percival, D., and A. Walden, Spectral analysis for physical applications - Multitaper and conventional univariate techniques, Cambridge University Press, 1993.
Philander, S., and A. Fedorov, Role of tropics in changing the response to Milankovich forcing
some three million years ago, Paleoceanography, 18(2), 2003.
Pisias, N. G., and J. Moore, T. C., The evolution of Pleistocene climate; a time series approach,
Earth and Planetary Science Letters, 52(2), 450–458, 1981.
149
Pollard, D., Investigation of astronomical theory of ice ages using a simple climate ice sheet model,
Nature, 272(5650), 233 – 235, 1978.
Pollard, D., A simple parameterization for ice-sheet ablation rate, Tellus, 32(4), 384 – 388, 1980.
Pollard, D., A simple ice-sheet model yields realistic 100 kyr glacial cycles, Nature, 296(5855),
334 – 338, 1982.
Pollard, D., A. Ingersoll, and J. Lockwood, Response of a zonal climate ice-sheet model to the
orbital perturbations during the Quaternary ice ages, Tellus, 32(4), 301 – 319, 1980.
Raymo, M., and K. Nisancioglu, The 41 kyr world: Milankovitch’s other unsolved mystery, Paleoceanography, 18(1), art. no. – 1011, 2003.
Raymo, M. E., The Initiation of Northern-Hemisphere Glaciation, Annual Review of Earth and
Planetary Sciences, 22, 353–383, 1994.
Raymo, M. E., W. F. Ruddiman, J. Backman, B. M. Clement, and D. G. Martinson, Late Pliocene
variation in northern hemisphere ice sheets and North Atlantic deepwater circulation, Paleoceanography, 4, 413–446, 1989.
Reeh, N., A plasticity theory approach to the steady-state shape of a 3-dimensional ice sheet,
Journal of Glaciology, 28(100), 431 – 455, 1982.
Reeh, N., Paramterization of melt rate and surface temperature on the Greenland ice sheet, Polarforschung, 59(3), 113–128, 1989.
Rivin, I., and E. Tziperman, Linear versus self-sustained interdecadal thermohaline variability in a
coupled box model, Journal of Physical Oceanography, 27(7), 1216–1232, 1997.
Roe, G., and R. Lindzen, The mutual interaction between continental-scale ice sheets and atmospheric stationary waves, Journal of Climate, 14(7), 1450 – 1465, 2001a.
150
Roe, G., and R. Lindzen, A one-dimensional model for the interaction between continental-scale
ice sheets and atmospheric stationary waves, Climate Dynamics, 17(5-6), 479 – 487, 2001b.
Ruddiman, W. F., and A. McIntyre, Oceanic Mechanisms For Amplification of the 23,000-Year
Ice- Volume Cycle, Science, 212(4495), 617–627, 1981.
Ruddiman, W. F., M. Raymo, and A. McIntyre, Matuyama 41,000-Year Cycles - North-Atlantic
Ocean and Northern-Hemisphere Ice Sheets, Earth and Planetary Science Letters, 80(1-2), 117–
129, 1986.
Ruddiman, W. F., M. E. Raymo, D. G. Martinson, B. M. Clement, and J. Backman, Pleistocene
evolution; Northern Hemisphere ice sheets and North Atlantic Ocean, Paleoceanography, 4(4),
353–412, 1989.
Saltzman, B., and A. Sutera, A model of the internal feedback system involved in late Quaternary
Climatic variations, Journal of the Atmospheric Sciences, 41, 736–745, 1984.
Saltzman, B., A. R. Hansen, and K. A. Maasch, The Late Quaternary Glaciations As the Response
of a 3- Component Feedback-System to Earth-Orbital Forcing, Journal of the Atmospheric Sciences, 41(23), 3380–3389, 1984.
Sarnthein, M., U. Pflaumann, and M. Weinelt, Past extent of sea ice in the northern North Atlantic
inferred from foraminiferal paleotemperature estimates, Paleoceanography, 18(2), 2003.
Schneider, S., and S. Thompson, Ice ages and orbital variations - some simple theory and modeling,
Quaternary Research, 12(2), 188 – 203, 1979.
Scott, J., Simple parameterization of meridional sensible and latent heat eddy fluxes due to seasonal
forcing, 1995.
Sellers, W. D., A global climatic model based on the energy balance of the earth-atmosphere system, Journal of Applied Meteorology, 8, 392–400, 1969.
151
Shackleton, N., A. Berger, and W. Peltier, An alternative astronomical calibration of the lower
pleistocene timescale based on ODP site 677, Transactions of the Royal Society of Edinburgh Earth Sciences, 81, 251 – 261, 1990.
Shackleton, N. J., and N. D. Opdyke, Oxygen Isotope and Palaeomagnetic Stratigraphy of Equatorial Pacific Core V28-238: Oxygen Isotope Temperatures and Ice Volumes on a 100,000 and
1,000,000 Year Scale (CLIMAP Program), Quaternary Research (New York), 3(1), 39–55, 1973.
Shackleton, N. J., M. A. Hall, and A. Boersma, Oxygen and Carbon Isotope Data From Leg-74
Foraminifers, Initial Reports of the Deep Sea Drilling Project, 74, 599–612, 1984.
Short, D., J. Mengel, T. Crowley, W. Hyde, and G. North, Filtering of Milankovitch cycles by earths
geography, Quaternary Research, 35(2), 157 – 173, 1991.
Solomon, A., and P. Stone, Equilibration in an eddy resolving model with simplified physics,
Journal of the Atmospheric Sciences, 58(6), 561 – 574, 2001.
Stanley, S. M., New horizons for paleontology with two examples: The rise and fall of the Cretaceous supertethys and the cause of the modern ice age., Journal of Paleontology, 69(6), 999–
1007, 1995.
Stone, P., S. Ghan, D. Spiegel, and S. Rambaldi, Short-term fluctuations in the eddy heat-flux and
baroclinic stability of the atmosphere, Journal of the Atmospheric Sciences, 39(8), 1734 – 1746,
1982.
Stone, P. H., A simplified radiative-dynamical model for the static stability of rotating atmospheres,
Journal of the Atmospheric Sciences, 29(3), 405–418, 1972.
Stone, P. H., Constraints on dynamical transports of energy on a spherical planet, Dynamics of
Atmospheres and Oceans, 2, 123–139, 1978.
Stone, P. H., and D. A. Miller, Empirical relations between seasonal changes in meridional
152
temeprature gradients and meridional fluxes of heat, Journal of the atmospheric sciences, 37,
1708–1721, 1980.
Stone, P. H., and M.-S. Yao, Development of a two-dimensional zonally averaged statisticaldynamical model. Part III: The parameterization of the eddy fluxes of heat and moisture., Journal
of Climate, 3, 726–740, 1990.
Suarez, M., and I. Held, Sensitivity of an energy-balance climate model to variations in the orbital
parameters, Journal of Geophysical Research-Oceans and atmospheres, 84(NC8), 4825 – 4836,
1979.
Tarasov, L., and W. R. Peltier, Terminating the 100 kyr ice age cycle, Journal of Geophysical
Research-Atmospheres, 102(D18), 21,665–21,693, 1997.
Thomson, D., Spectrum estimation and harmonic-analysis, Proceedings of the IEEE, 70(9), 1055
– 1096, 1982.
Trenberth, K. E., and J. M. Caron, Estimates of meridional atmosphere and ocean heat transports,
Journal of Climate, 14(16), 3433–3443, 2001.
Trenberth, K. E., and D. P. Stepaniak, Co-variability of components of poleward atmospheric energy transports on seasonal and interannual timescales., 2003.
Urey, H. C., The thermodynamic propoerties of isotopic substances, J. Chemical Society, 1947,
562–581, 1947.
Van der Veen, C., Tidewater calving, Journal of Glaciology, 42(141), 375 – 385, 1996.
Van der Veen, C., Calving glaciers, Progress in Physical Geography, 26(1), 96 – 122, 2002.
van der Veen, C., D. Bromwich, B. Csatho, and C. Kim, Trend surface analysis of Greenland
accumulation, Journal of Geophysical Research-Atmospheres, 106(D24), 33,909 – 33,918, 2001.
153
Vernekar, A. D., Long-period global variations of incoming solar radiation, Meteorological Monographs, 12(34), 1, 1972.
Wang, W., and P. H. Stone, Effect of ice-albedo feedback on global sensitivity in a one-dimensional
radiative-convective climate model, Journal of the atmospheric sciences, 37, 545–552, 1980.
Wardlaw, B., and T. Quinn, The record of Pliocene sea-level change at Enewetak atoll, Quaternary
Science Reviews, 10(2-3), 247 – 258, 1991.
Webb, P., and D. Harwood, Late Cenozoic glacial history of the Ross embayment, Antarctica,
Quaternary Science Reviews, 10(2-3), 215 – 223, 1991.
Weertman, J., Rate of growth or shrinkage of nonequilibrium ice sheets, Journal of Glaciology,
5(38), 145–158, 1964.
Weertman, J., Milankovitch solar radiation variations and ice age ice sheet sizes, Nature,
261(5555), 17 – 20, 1976.
Werner, M., M. Heimann, and G. Hoffmann, Isotopic composition and origin of polar precipitation
in present and glacial climate simulations, Tellus Series B-Chemical and Physical Meteorology,
53(1), 53–71, 2001.
Wetherald, R., and S. Manabe, Effects of changing solar constant on climate of a general circulation
model, Journal of the Atmospheric Sciences, 32(11), 2044 – 2059, 1975.
Winograd, I. J., T. B. Coplen, J. M. Landwehr, A. C. Riggs, K. R. Ludwig, B. J. Szabo, P. T.
Kolesar, and K. M. Revesz, Continuous 500,000-year climate record from vein calcite in Devils
Hole, Nevada, Science, 258(5080), 255 – 260, 1992.
Young, M. A., and R. S. Bradley, Insolation gradients and the paleoclimate record, pp. 707–713,
D. Reidel Publishing Company, 1984.
154
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement