ICE AGES (MILANKOVITCH THEORY) 995
ICE AGES (MILANKOVITCH THEORY)
M F Loutre, Université catholique de Louvain, Louvain-
la-Neuve, Belgium
Copyright 2003 Elsevier Science Ltd. All Rights Reserved.
Introduction
An astronomical theory of paleoclimate aims to
provide an understanding of the relationship between
insolation and climate on the global scale. One such
theory was developed by Milankovitch who first
computed the radiation received by the Earth from
the Sun (insolation). This quantity undergoes diurnal,
seasonal, as well as millennial scale (long-term)
variations. Milankovitch argued that insolation
changes induce surface temperature changes and that
long-term insolation changes are sufficient to produce
ice ages by changing the geographic and seasonal
distribution of sunlight received by the Earth. He
estimated the relationship between summer radiation
and the altitude of the snowline, which allowed him
to compute the latitude of the ice sheet margin for the
last 650 000 years. Milankovitch’s curves were found
to be remarkably consistent with the reconstruction of
the glacial period made by Penck and Brückner.
However, Milankovitch theory is not the only astro-
nomical theory of paleoclimate. For example, in the
nineteenth century, Croll developed an astronomical
theory of paleoclimate, which was later rejected
because it did not fit with the evidence of that time.
Today astronomical theories are still a focus of
research. Sophisticated climate models are used to
test how insolation, and in particular its distribution in
Orbital parameters
Insolation change
Simulated
climate change
Figure 1 Schematic outline of an astronomical theory of paleoclimate.
latitude through the year, is forcing climate. Four main
steps can be identified in an astronomical theory
(Figure 1):

the theoretical computation of the long-term vari-
ations of the Earth’s orbital parameters and related
insolations;

the design of climate models to study climate
changes related to insolation changes;

the collection of geological data and their interpre-
tation in terms of climate; and

the comparison of these proxy data to the simulated
climatic variables.
Therefore astronomical theory is at a crossing point of
geology, astronomy, physics, chemistry, biology, and
geophysics. Moreover, it is related to processes of the
solid Earth, the atmosphere, the hydrosphere, and the
ocean.
The Ice Ages
Our planet Earth was born some 4.5 billion years ago.
Many events have marked its evolution; in particular,
some periods have experienced markedly colder
climate than others. Up to 0.9 billion years ago the
Earth was apparently mostly ice-free, despite low solar
luminosity, although there is some evidence for a first
glaciation approximately 2.5 billion years ago. At
least three major phases of glaciation occurred be-
tween 900 and 600 million years ago. Glacial deposits
from the late Proterozoic have been found at low
... Geological data
Recorded
Comparison climate change
996 ICE AGES (MILANKOVITCH THEORY)
latitudes, suggesting that at that time ice sheets
covered the Earth from pole to pole. This is the so-
called ‘snowball Earth’ hypothesis. The return to
warmer conditions would then have resulted from the
accumulation in the atmosphere of CO2 from volcanic
activity. The large cap carbonate found in Namibia,
for example, could be the witness of this time.
However, this hypothesis is still debated. From 600
to 100 million years ago mild climates prevailed,
punctuated by several major phases of ice growth.
These long geological cold periods, such as the late
Precambrian Ice Age, the late Ordovician-Silurian Ice
Age, and the Permo-Carboniferous Ice Age, are also
called Ice Ages. A gradual cooling and drying of the
globe started some 50 million years ago. The Antarctic
ice sheet started to grow some 40 million years ago,
whereas the Greenland and midlatitude ice sheets
formed later (4–2.4 million years ago). The Quater-
nary Ice Age, the cold period in which we are living, is
characterized by a succession of colder and warmer
periods, the glacial–interglacial cycles. During colder
phases (or ice ages) the ice sheets spread out in the high
latitudes. The purpose of astronomical theory is to
explain these broad climatic features, which charac-
terize not only the Quaternary, but also older periods
including the Early Mesozoic, more than 150 million
years ago.
A Brief History of the Astronomical
Theory of Paleoclimates
The inspiration for the pioneering work on the
astronomical theory of paleoclimate was probably
Louis Agassiz’s lecture about his theory of a Great Ice
Age at a meeting of the Swiss Society of Natural
Sciences, held in Neuchâtel in 1837. Only a few years
later, Joseph Adhémar proposed an explanation of the
existence of ice ages based on the precession of the
equinoxes. At the same time the French astronomer,
Urbain Leverrier, calculated the changes in the Earth’s
orbit over the last 100 000 years. James Croll would
eventually take Adhémar’s idea and put it into an
astronomical theory of climate. Croll’s major hypoth-
esis was that the critical season for producing an ice
age is winter. He determined that the precession of the
equinoxes must play a decisive role in the amount of
sunlight received during winter. Indeed, winter would
be warmer if it occurred when the Earth were closer to
the Sun and it would be colder if the Earth were farther
from the Sun. Moreover, the shape of the Earth’s orbit
could reinforce this effect. He concluded that periods
of greater eccentricity could lead to exceptionally
warm or cold winters. He argued that ice ages
therefore occurred alternately in one hemisphere or
the other during Glacial Epochs, when eccentricity is
large. According to Croll, the last Glacial Epoch began
some 250 000 years ago and ended about 80 000 years
ago. Later, he also took into account the role of the tilt
of the Earth’s axis of rotation. He hypothesized that an
ice age would be more likely to occur when the tilt was
small because the polar regions would then receive less
heat. However, he acknowledged that orbital changes
could only be a triggering mechanism. He identified
the albedo–temperature feedback and the change in
ocean currents as two mechanisms responsible for
amplifying the direct climatic effect of the astronom-
ical forcing. Meanwhile, geologists collected evidence
around the world that several ice ages had occurred in
the past, separated by nonglacial epochs, as predicted
by Croll’s theory. However, with time the majority
of geologists in Europe and America became opposed
to Croll’s theory as more and more new evidence
suggested that the last Glacial Period ended later
than 15 000 years ago, instead of 80 000 years
ago as required by Croll’s astronomical theory. By
the end of the nineteenth century, the theory fell into
disfavor.
The attention of the scientific world was drawn
back to the ice age problem with the publication in
1924 of Milankovitch’s theory. This was the first full
astronomical theory of the Pleistocene ice ages,
including the computation of the orbitally induced
changes in the insolation and climate. According to
Milankovitch’s theory, the summer in northern high
latitudes had to be cold to prevent the winter snow
from melting. In turn this would allow a positive value
in the annual budget of ice, and a positive feedback
cooling would be initiated over the Earth through a
further extension of the snow cover and subsequent
increase of surface albedo. This hypothesis requires
that summer insolation is a minimum in the high
latitude Northern Hemisphere. One of Milankovitch’s
first major contributions consisted of radiation curves
showing past insolation changes at high northern
latitudes (Figure 2). He claimed that each minimum of
these curves should cause an ice age. Comparing
Milankovitch’s curves with the Penck and Brückner
curve for the succession of European ice ages, Köppen
and Wegener concluded that the theory matched the
facts amazingly well. Although the timing of the ice
ages and the radiation minima did not agree in detail,
the general pattern of the two curves was quite similar.
But by 1955, the astronomical theory was rejected by
geologists. Indeed, using new techniques for dating
Pleistocene fossils (radiocarbon dating) they showed
that there were more glacial advances during the past
80 000 years (or at least the time interval believed to be
80 000 years) than could be explained by the Milan-
kovitch theory.
 60°

Equivalent latitude
65°

70°

75°

600 550 500 450 400 350 300 250 200 150 100 50 0

Riss I

Riss II
Günz II

Mindel II

Wurm I
Günz I

Mindel I

Wurm II

Wurm III
(A)

60°
Equivalent latitude

65°

70°

75°

(B) 600 550 500 450 400 350 300 250 200 150 100 50 0

500

ICE AGES (MILANKOVITCH THEORY) 997

450

400

600 550 500 450 400 350 300 250 200 150 100 50 0
(C) Time (ky BP)

Figure 2 The Milankovitch amplitude of the secular variations of the summer radiation at 651 N, (A) after Stockwell and Pelgrim’s, and (B) after Le Verrier and Miskovitch’s astronomical solution.
The ordinate axis in (A) and (B) gives the latitude that receives the same radiation as 651 N for the last 600 ky. Part (C) gives the mean irradiance (W m  2) according to Berger’s computation.
998 ICE AGES (MILANKOVITCH THEORY)
The theory was still largely disputed until the early
1970s. Nevertheless, progress was being made toward
a better understanding of the ice ages, in particular the
Pleistocene. New dating techniques were developed
and accurate climatic interpretation was given to
geological records, such as variation of the oxygen
isotope records of forams in deep sea cores, or
sequences of soils and loess. New evidence was put
forward that major climate changes have accompa-
nied variations in obliquity and precession over the
last 500 000 years. This finding stimulated the revival
of the astronomical theory. Vernekar, Bretagnon,
Berger, and others refined the calculations of orbital
history; geologists obtained new records of past
climate; the improved dating techniques clarified the
detail of the Quaternary time scale; global past climate
changes were reconstructed with better accuracy; and
finally, climate models were designed to test this
theory. Within a few years it became increasingly clear
that major changes in global climate were caused by
changes in the astronomically driven insolation forc-
ing. Moreover, the importance of mechanisms and
processes such as the carbon cycle, vegetation change,
ocean circulation, and dynamics of the cryosphere was
also acknowledged.
Orbital Parameters
The German astronomer Johannes Kepler (1571–
1630) formulated the three laws of planetary motion,
which are satisfied with a high accuracy not only by the
system of planets and Sun, but also by the system of
each set of satellites moving about their primary
planet. They are:

The orbit of each planet is an ellipse with the Sun at
one focus.

The line (the radius vector) joining the Sun to each
planet sweeps out equal areas of its ellipse in equal
times.

The cubes of the semimajor axis of the planetary
orbits are proportional to the squares of the planets’
periods of revolution.
While Kepler gave a description of the orbital motion,
Sir Isaac Newton (1642–1727) defined the law of
gravitation, which is the basis for explaining
the planetary motion. In particular, during its
travel counterclockwise on its orbit around the Sun
(Figure 3), the Earth is sometimes nearer to and
sometimes farther away from the Sun. The distance
from the Earth to the Sun (r) can be computed through
the equation of the ellipse, given here as eqn [1].
r ¼ ½að1  e2 Þð1 þ e cos vÞ1 ½1
In eqn [1], a, the semimajor axis of the orbit, gives its
size. The value of a is constant through geological time
to a very good accuracy. The eccentricity, e, is a
measure of the departure of the ellipse from a circle,
and the true anomaly, v, of the Earth is a measure of the
position of the Earth in its orbit. The true anomaly is
measured counterclockwise from perihelion (where
the Earth is nearest to the Sun). Usually the angles that
describe the position of the Earth in its orbit are not
measured from the perihelion, but rather from the
spring equinox (SE). Therefore, we have v ¼ l  o,
where l is the longitude of the Earth in its orbit
measured from the spring equinox of the year, or
moving spring equinox, and o is the longitude of the
perihelion relative to this same moving spring equi-
nox. Alternatively, the position of the perihelion
(o
 ) is often measured from the First Point of Aries
(vernal point). This point on the Earth’s orbit gives the
direction of the Sun as seen from the Earth at the spring
equinox. Consequently, we have v ¼ l  o   p. In
addition, the Earth’s axis of rotation is tilted with
respect to the orbital plane. The obliquity, e, is the
angle between the Earth’s axis of rotation and the
perpendicular to the orbital plane.
The astronomical parameters, e, e, and o, experi-
ence long-term variations. These variations can be
obtained by solving two sets of equations, one set for
the mutual gravitational forces in the planetary system
and the other for the Sun–Earth–Moon system.


AE


WS
A
P
S
v
SS

SE
Figure 3 Elements of the Earth’s orbit around the Sun (S). Some
reference points are drawn on the orbit, i.e., the perihelion (P) and
the aphelion (A), the spring equinox (SE), the summer solstice
(SS), the autumn equinox (AE), and the winter solstice (WS). The
vernal point is given by g. l is the longitude of the Earth in its orbit
measured from the spring equinox of the year, or moving spring
equinox; o is the longitude of the perihelion relative to this same
moving spring equinox; and v is the true anomaly measured
counterclockwise from the perihelion.

Different methods, from fully analytical to fully
numerical, were developed following the first calcula-
tions by Leverrier in the nineteenth century. Moreover,
the accuracy of these solutions has been greatly
improved. However, the orbital motion of the planets
in the solar system is chaotic, i.e., the solution is
strongly dependent on the initial conditions, which
limits the possibility of obtaining an accurate solution
for the astronomical parameters of the Earth over
more than 35–50 million years.
Figure 4 shows the long-term variations of the three
orbital parameters (e, e, e sin o ) from 400 000 years
Before Present (400 ky BP) to 100 000 years After
Present (100 ky AP), a time slice over which the
solution is very accurate. The eccentricity of the orbit
varies between near circularity (e ¼ 0) and small
ellipticity (e ¼ 0:07). These variations are quasi-peri-
odic with a mean period of about 100 ky. However, a
longer period of about 400 ky can also be discerned. In
practice, the analytical solution for the eccentricity,
expressed in trigonometrical series, puts forwards
many terms having different periods. The major
periods in the development are 404, 95, 124, 99,
and 131 ky (in order of decreasing amplitude). The
obliquity (tilt of the Earth’s axis) varies between 221
and 251 with a very clear quasi-period of 41 ky. The
0.06
0.05 Eccentricity (e)
0.04
0.03
0.02
0.01
0.00
_
Climatic precession (e sin )
25
Obliquity ()
24
23
22
100 50 0 _ 50 _ 100
(kyr AP) Time (kyr BP)
Figure 4 Long-term variations of eccentricity, climatic precession and obliquity from 400 000 years ago to 100 000 years into the future
(from Berger).
ICE AGES (MILANKOVITCH THEORY) 999
largest amplitude terms correspond to periods of 41
(by far the largest term), 54, and 39 ky. The variations
of the climatic precession (e sin o ) reflect two oppos-
ing motions, i.e., the counterclockwise motion of the
perihelion along the ecliptic with a period of  100 ky
and the clockwise motion of the vernal point along the
ecliptic with a period of  25 700 years. The two
effects taken together result in the climatic precession,
which corresponds to the slow shift of the seasons
about the Earth’s orbit relative to the perihelion. This
motion has a mean quasi-period of 21 ky, which is
derived from periods at 19 and 23 ky.
Insolation
The amount of solar radiation received at a mean
Earth–Sun distance, rm , by a surface perpendicular to
the incident radiation, is about 1370 W m 2 (this is the
so-called total solar irradiance, S0 ). However, rm varies
over time according to the eccentricity. Therefore,
instead of S0 , it is often preferred to use the related
quantity, S, defined at the constant distance a from
Sun. As the solar energy decreases according to the
square of the distance from the Sun, the amount of
solar radiation received by the Earth on a unit surface
perpendicular to the Sun’s rays at a distance r from the
0.06
0.04
0.02
0.00
_ 0.02
_ 0.04
_ 0.06
_ 150 _ 200 _ 250 _ 300 _ 350 _ 400
1000 ICE AGES (MILANKOVITCH THEORY)
Sun is given by W ¼ Sða=rÞ2. Moreover, the incident
radiation is usually not perpendicular to the Earth’s
surface, but it is tilted according to the elevation of the
Sun above the horizon. The elevation angle at a given
point on the Earth is measured by the zenith distance,
z, which is the angular distance from the zenith (the
point vertically overhead) along the vertical circle
through the point. The instantaneous insolation (irra-
diance) received by a unit horizontal surface at a
particular time characterized by a zenith distance, z, is
given by eqn [2].
W ¼ Sða=rÞ2 cos z
Spherical trigonometry applied to the astronomical
coordinates on the celestial sphere for the Earth’s
orbital motion provides cos z (Figure 5), expressed as
eqn [3].
cos z ¼ sin f sin d þ cos f cos d cos H
Here f is the latitude on the Earth, d is the declination
(the angular distance from the Equator along the
meridian), and H is the hour angle (measured clock-
wise on the Equator from the meridian of the observer
to the meridian of the Sun). The declination depends
on the obliquity (e). It can be computed for any given
H
or
Eq
izo
ua
n
tor
Eclipti
c
Figure 5 Position of a celestial body (the Sun for example) on the celestial sphere. The different variables are explained in the text.
time in the year l using eqn [4].
sin d ¼ sin l sin e ½4
This shows that the energy (irradiance) available at
any given latitude f on the Earth (on the assumption of
a perfectly transparent atmosphere) is a single-valued
function of the total solar irradiance, S, the semimajor
axis, a, of the Earth orbit about the Sun, its eccentricity
e, its obliquity e, and the longitude of the perihelion
measured from the moving spring equinox, o. There-
fore it appears that the irradiance varies only accord-
ing to three astronomical parameters, i.e., the
½2 eccentricity (e), the climatic precession (e sin o  ) and
the obliquity (e). Moreover, climatic precession and
eccentricity on one side, and obliquity on the other
side, appear in two distinct factors in the formulation
for the irradiance. Each of these factors has a physical
meaning. The distance factor (r ¼ r=a) depends on the
½3 precession and eccentricity, and the inclination factor
(cos z) is solely a function of the obliquity.
The daily irradiation is the irradiance integrated
over a whole day, either from sunrise to sunset or over
24 h, in case of no sunset. The 24 h mean irradiance
(Wd ), i.e., the average daily irradiation over 24 h,
expressed in W m 2, is often preferred. The value of
Wd depends on the latitude.
PN π _
2
π _
O 2
S H
 
SE 
PS


For the latitudes where there is a daily sunrise and
sunset, i.e., jfj  p=2  jdj, Wd is expressed by eqn
[5],
Wd ¼ S=pr2 ðH0 sin f sin d
þ cos f cos d sin H0 Þ
where H0 , the absolute value of the hour angle at
sunrise and sunset, is given by eqn [6].
cos H0 ¼  tan f tan d

For the other latitudes, i.e., jfj > p=2  jdj:
Either it is the long polar night (H0 ¼ 0), i.e.,
fd  0, in which case Wd is given by [7]
Wd ¼ 0
Or it is the long polar day (H0 ¼ p), i.e., fd > 0, in
which case Wd is given by [8].
Wd ¼ Sr2 sin f sin d
The daily irradiation varies through the year as well
as according to the latitude. Moreover, it also exhibits
long-period variations caused by the secular variations
in the elements of the Earth’s orbit and rotation.
Some features of the long-term variations in eccen-
tricity, obliquity, and climatic precession can be
discerned in the insolation variations. Low values of
the eccentricity are mirrored in the small amplitude of
the insolation change, such as for the recent past and
near future; conversely, large values of e induce large
amplitudes of the insolation change, for example, at
about 100 ky BP (Figure 2C). Solar radiation is mostly
90° N
60° N
30° N
Latitude
0°
30° S
60° S
90° S
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Time (month)
(A) _20 _10 0 10 20 30 40
Figure 6 Variation of the seasonal contrast of the mean irradiance (W m ) following (A) an increase of the obliquity from 22.51 to 251
(e ¼ 0:05 and winter at perihelion) and (B) a change in the climatic precession (from summer solstice at perihelion to summer solstice at
aphelion; e ¼ 0:05 and e ¼ 25 ).
ICE AGES (MILANKOVITCH THEORY) 1001
affected by variations in precession, although the
obliquity plays a relatively more important role for
high latitudes, mainly in the winter hemisphere. The
variations in the obliquity are perceptible in the same
way in both hemispheres (Figure 6A), i.e., an increase
½5 in the obliquity induces an increase in the insolation
during the local summer (March to September in the
Northern Hemisphere and September to March in the
Southern Hemisphere) and an insolation decrease
during local winter. Consequently the seasonal con-
½6
trast in daily irradiation is reinforced. A change in the
climatic precession (Figure 6B) such that the June
summer solstice is moving from the perihelion to the
aphelion (i.e., close to the present-day situation)
induces a decrease of insolation over the whole Earth
½7 (Northern and Southern Hemispheres) simultaneous-
ly over one half of the year (Northern Hemisphere
summer season and Southern Hemisphere winter
season, i.e., from March to September). Climatic
½8 precession plays an opposite role in both hemispheres.
At present, perihelion occurs in early January. This
situation favors mild winters and cool summers in the
Northern Hemisphere, and cold winters and warm
summers in the Southern Hemisphere.
Comparison between changes in the orbital param-
eters and/or in the solar radiation received by the Earth
with geological reconstruction of past climate changes
is also used to provide a clock for dating the records. In
this case it is assumed that the quasi-periods observed
in the data are a direct response to the quasi-periodic
variations of the forcing. Consequently the astronom-
ical chronology is directly applied to the geological
data, possibly with a time lag.
90° N
60° N
30° N
Latitude
0°
30° S
60° S
90° S
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Time (month)
50 60 (B) _100 _ 80 _ 60 _ 40 _20 0 20 40 60 80 100
2
1002 ICE AGES (MILANKOVITCH THEORY)
Paleoclimate Modeling
Climate models are simplifications of reality, designed
to describe the complexity of the interactions within
the climate system. These numerical climate models
can be used to test the astronomical theory of
paleoclimate, i.e., to study whether astronomically
induced changes in insolation are able to drive
the climate system into glacial–interglacial cycles
similar to these recorded in geological data. The
modeling effort has led to a better understanding of the
physical mechanisms involved in the climate system
response to the astronomically forced changes in the
pattern of incoming solar radiation. Such mechanisms
are related, in particular, to the ice sheets, the
lithosphere, the hydrological cycle, the cloud proper-
ties, the albedo temperature feedback, the land–sea ice
temperature gradient, the CO2 cycle, and the ocean
circulation. The different parts of the climate system,
i.e., the atmosphere, the hydrosphere, the cryosphere,
the biosphere, and the lithosphere, are becoming
convincingly modeled separately, and work is going
on towards the design of comprehensive coupled
models including several parts, if not all of them. A
hierarchy of models, climate models of different
complexities that differ in their degree of spatial and
temporal resolution, are used for paleoclimate
purposes.
General Circulation Models (GCMs) are primarily
used for simulating geographic features of paleocli-
mates. Their major limitation is their high computing
cost. For this reason they are used for simulations
covering a few thousand years at maximum. They
provide a ‘snapshot’ view of the climate in equilibrium
with the boundary conditions.
At the last interglacial, some 125 ky BP, modeling
experiments led to warmer conditions, especially in
the high latitudes, reduced sea-ice extent, enhanced
northern tropical monsoon and northward displace-
ment of the tundra and taiga biomes, in good agree-
ment with geological reconstruction. However, the
strong cooling induced by changes in the orbital
parameters at 115 ky BP are not sufficient to initiate
glaciation, at least if vegetation changes are not
properly taken into account. This clearly puts forward
the importance of the vegetation–albedo–temperature
feedback. Several GCMs have been used to simulate
the climate of the Last Glacial Maximum, some 20 ky
ago. Again, important processes at work at that time
were identified. They are related to the CO2 concen-
tration, sea ice, ocean temperature, and land albedo.
As part of the Paleoclimate Modeling Intercomparison
Project (PMIP), several GCMs performed the same
simulation of the mid-Holocene climate (6 ky BP).
Some robust features have been identified, for instance
a northward shift of the main regions of monsoon
precipitation over Africa and India.
The conceptual models are simple models designed
to assess whether a climate process can explain
past climate changes. For example, the simple thresh-
old (or multistate) climate model due to Paillard
simulates ice volume increase as a function of a
smoothed truncation of the insolation. The
model distinguishes three distinct states (interglacial,
mild glacial, and full glacial) and the transition
between them occurs when insolation and ice volume
cross prescribed thresholds in insolation and ice
volume. This model reproduced reasonably well the
succession of glacial–interglacial cycles over the late
Pleistocene.
Models of intermediate complexity are the only
climate models to be able to simulate the time-
dependent behavior of the fully coupled climate
system over a time interval long enough to test the
astronomical theory of paleoclimate. Earth system
Models of Intermediate Complexity (EMICs) include
most of the processes described in comprehensive
models, in particular the slow-response climate vari-
ables such as ice volume, bedrock depression, deep-
ocean temperature, and atmospheric concentration of
greenhouse gases. They also simulate the interactions
between the different parts of the climate system.
Moreover, they are simple enough to allow for long-
term climate simulations (several glacial–interglacial
cycles).
The LLN 2D NH climate model (two-dimensional
climate model developed in Louvain-la-Neuve) is one
of these EMICs. It was designed in order to understand
the response of the climate system to astronomical
forcing. It links the atmosphere, the upper mixed layer
of the ocean, the sea ice, the continents, the ice sheets,
and their underlying lithosphere. It is forced by
computed insolation and reconstructed atmospheric
CO2 concentration. It considers only the Northern
Hemisphere (the Southern Hemisphere is not consid-
ered) and it has no explicit representation of the
thermohaline circulation. It has been able to simulate
many of the different situations that characterize the
last 3 million years: the entrance into glaciation
around 2.75 My BP, the dominance of the obliquity
cycle during the late Pliocene–early Pleistocene, the
emergence of the 100 ky cycle around 900 ky BP, and
the glacial–interglacial cycles of the last 600 ky. The
climatic changes over the Northern Hemisphere, in
particular the continental ice volume, simulated by the
LLN 2D NH climate model during the last 400 ky
shows a broad good agreement with reconstruction
(Figure 7).
However, a major discrepancy in this model is the
too frequent melting of the ice sheets during the
 ICE AGES (MILANKOVITCH THEORY) 1003

Ice volume (106 km3)

10

20

30
(C)
_2 40

50

Sea level
δ18O

(B)
2 550

Insolation (W m 2 )
CO2 (ppmv)

250 500

_
200 450

0 50 100 150 200 250 300 350 400

(A) Age (ky BP)

Figure 7 Comparison of records and modeled data over the last 400 000 years. (A) Variation in the mean irradiance in July at 601 N (red
full line) and the CO2 concentration (dashed green line). (B) Proxy records for the variation of continental ice volume, i.e., stacked,
smoothed oxygen isotope record as function of age in the SPECMAP time scale (full dark blue line), d18 O record from the oceanic core
MD900963 (long-dashed blue line), and reconstructed sea level from benthic forams in the oceanic core V19-30 (short-dashed blue line).
(C) Northern Hemisphere continental volume as simulated by the LLN 2D NH climate model.

interglacial. The largest difference between the simu- See also

lated and the reconstructed Northern Hemisphere
continental ice volume appears between 180 and
150 ky BP. Moreover an unusual feature shows up
between 400 and 350 ky BP. This time interval is
characterized by a very long interglacial, which
does not seem to be recorded in data. This behavior
is possibly caused by the interplay between insolation
forcing and CO2 concentration forcing. This
model also confirms that the orbital forcing acts
as a pacemaker for the glacial–interglacial cycles
and that the climate response to orbital forcing is
amplified by CO2. Moreover, important processes in
climate change were identified, such as albedo–
temperature feedback, water vapor–temperature feed-
back, the snow aging process, and the isostatic
rebound.
New observation techniques, accurate dating meth-
ods, improved transfer functions, and comprehensive
climate models will lead to increasingly accurate
knowledge of the past evolution of the atmosphere
and the oceans, the waxing and waning of the ice
sheets, and the growth and retreat of the forests and
deserts.
Carbon Dioxide. Climate Variability: Glacial, Intergla-
cial Variations. Energy Balance Model, Surface. Gen-
eral Circulation: Models. Glaciers. Numerical
Models: Methods. Paleoclimatology: Ice Cores; Varves.
Further Reading
Berger A, Imbrie J, Hays J, Kukla G and Saltzman B (eds)
(1984) Milankovitch and Climate. NATO ASI Series C,
vol. 126. Dordrecht: Reidel Publishing Company.
Bradley R (1999) Paleoclimatology. Reconstructing
Climates of the Quaternary. New York and London:
Academic Press.
Imbrie J and Imbrie KP (1979) Ice Ages. Solving the Mystery.
Cambridge, MA and London, England: Harvard Univer-
sity Press.
Milankovitch MM (1941) Kanon der Erdbestrahlung
und seine Anwendung auf des Eizeitenproblem.
R. Serbian Acad. Spec. Publ. 132, Sect. Math. Nat.
Sci., 33. Beograd: Köninglich Serbische Akademie.
Reprinted in English: Canon of Insolation and the
Ice-Age Problem. Zavod za udzbenikb i nastavna
sredstva, Beograd (1998).
Roy AE (1978) Orbital Motion. Bristol, Philadelphia and
New York: Adam Hilger.
1004 INSTABILITY / Inertial Instability
INSTABILITY
Contents
Inertial Instability
Symmetric Stability
Wave–CISK
Inertial Instability
J A Knox, University of Georgia, Athens, GA, USA
Copyright 2003 Elsevier Science Ltd. All Rights Reserved.
Introduction
Inertial instability is a fundamental, but infrequently
documented, hydrodynamic instability characterized
by strongly divergent anticyclonic flow. It is the
geophysical analog to the centrifugal (or Taylor–
Couette) instability in fluid dynamics examined by
Rayleigh and Taylor nearly a century ago. Bergen
School meteorologists seeking an explanation for
cyclogenesis pioneered inertial instability research in
the 1930s and 1940s. However, the triumph of
baroclinic instability theory and balanced dynamics
in the 1950s and 1960s sharply curtailed interest in
inertial instability. A revival of research during the past
two decades, particularly in middle-atmosphere and
mesoscale dynamics, has led to a growing appreciation
of the role of inertial instability in geophysical flows.
Today, inertial instability arises in a wide range of
subjects: the dynamics of mesoscale convection and
monsoons, wave generation and breaking in the
stratosphere and mesosphere, and the maintenance
of jets in planetary atmospheres and equatorial
oceanography. Below, we examine what inertial
instability looks like physically, how it is represented
mathematically, and how it is manifested geophysi-
cally. Those new to the subject may wish to begin with
the latter.
Physical Description
Centrifugal instability occurs in the Taylor–Couette
problem when the angular momentum of a fluid
contained between two rotating cylinders (Figure 1A)
decreases radially outward, violating Rayleigh’s sta-
bility criterion. Parcels then rearrange themselves to
achieve a stable radial profile of angular momentum.
Inertial instability is the geophysical equivalent of
centrifugal instability and occurs when angular
momentum decreases as one moves outward from
the axis of rotation of the flow. This can be visualized
on the global scale (Figure 1B) by imagining the
roles of the cylinders being played by latitude lines,
with the Equator serving as the inner cylinder. The
latitude at which angular momentum is nondecreasing
outward from the rotation axis forms the outer
cylinder.
This analogy between centrifugal and inertial
instabilities is correct only if the rotation in the
Taylor–Couette apparatus is equated to the total
vertical rotation of the flow in the geophysical
case. In meteorological terminology, the rotation in
Figure 1B is the sum of the planetary and relative
vertical vorticities. From this viewpoint, inertial
instability should be expected where the relative
vorticity is opposite in sign to the planetary vorticity,
and at least equal in magnitude to it. The threshold for
inertial instability – the latitude corresponding to the
outer cylinder in Figure 1B – is thus the location of zero
absolute vorticity. With some restrictions, this is
proved mathematically in the next section.
Inertial instability, like centrifugal instability, leads
to roll-like vortical motions. The flattened, ‘pancake’-
like divergent circulations of inertial instability
(Figure 2) attempt to reduce the anticyclonic local
rotation via momentum transports so that the total
rotation is the same sign throughout the domain.
An inertially unstable flow can exist on a variety of
scales. As illustrated in Figures 1 and 2, the flow can be
as large as a zonally symmetric latitudinal ring of air
around the Earth. It can also occur on scales as small as
a mesoscale vortex. When viewed from the perspective
of a circular vortex, this instability is triggered when a
very strong outward pressure gradient force and the
centrifugal force combine to overwhelm the Coriolis
force and lead to the breakdown of balance.
 INSTABILITY / Inertial Instability 1005

In layman’s terms, the salient point of inertial

R2 instability research is that a planet’s rotation sets a
local rotational ‘speed limit’ beyond which violations
R1 are corrected via inertial instability. This ‘speed limit’
Ω1 Ω2 = 0 can be indirectly inferred from weather charts in a
number of ways: for example, the strong tendency for
nonnegative absolute vorticity in the Northern Hemi-
sphere, and also the complete absence of intense
g ‘bull’s-eye’ high-pressure centers analogous to ‘bomb’
cyclones.
Inertial instability depends on planetary rotation
and the horizontal shear and/or curvature of the
horizontal wind. In this sense, it shares an affinity with
barotropic instability. However, the rapid, strongly
divergent motions associated with inertial instability
invite analogies with another fundamental, strongly
ageostrophic instability: static instability. This ana-
logy is strengthened by the close similarity in the
r derivation of these two instabilities’ criteria, discussed
below.
(A)
Mathematical Criteria for Instability
Inertial instability represents a large departure from
Ω
geostrophic balance. As a result, inertial instability
theory cannot rely on quasi-geostrophic or nonlinear
balances to facilitate a generalized theory, making it a
far less tractable problem than barotropic or barocli-
nic instability. In fact, no fully three-dimensional
theory for inertial instability exists currently. Instead,
the classic derivation relies on the parcel-mode
approach, an extremely simplified flow geometry,
Ωa = f +  < 0
and ignores nonlinear, frictional, and diabatic effects.
Even so, the result is remarkably useful when applied
to observed and simulated flows.
We begin by assuming that we have a perturbed
z parcel embedded in a purely zonal geostrophic flow, as
illustrated in Figure 3. The horizontal equations of
motion describing the parcel are:

DH u
Ωa = f +  = 0 ¼ fvag ½1
Dt

(B)
Figure 1 (A) Centrifugal instability in the Taylor–Couette
experiment and (B) inertial instability near the equator. (Adapted
with permission from Hua BL, Moore DW and Le Gentil S (1997)
Inertial nonlinear equilibration of equatorial flows. Journal of Fluid
Mechanics 331: 345–371.)
DH vag

¼ f ug  u ½2
Dt
In these equations, the horizontal Lagrangian
derivative DH/Dt 5 q/qt 1 uq/qx 1 vq/qy and f is the
Coriolis parameter (assumed to be constant for
simplicity). In eqn [2], the meridional pressure
gradient term has been rewritten in terms of the
geostrophic wind.
These two equations can be combined into one
equation for the meridional ageostrophic wind in the
following manner. Taking the Lagrangian derivative
1006 INSTABILITY / Inertial Instability

C W

CONV DIV
Altitude

W C

DIV CONV

C W

f =0 f+ζ=0
Latitude

Figure 2 Schematic view of inertially unstable circulations in (y, z). DIV and CONV refer to regions of divergence and convergence due
to the horizontally divergent inertial circulations, shown in heavy bold arrows. Conservation of mass leads to the vertical motions shown in
the lighter arrows, which adiabatically create the warm and cold temperature anomalies labeled W and C respectively. (Adapted with
permission from Dunkerton TJ (1981) On the inertial stability of the equatorial middle atmosphere. Journal of the Atmospheric Sciences
38: 2354–2364, and Hayashi H, Shiotani M and Gille JC (1998) Vertically stacked temperature disturbances near the equatorial
stratopause as seen in cryogenic limb array etalon spectrometer data. Journal of Geophysical Research 103: 19469–19483.

of eqn [2] yields that the parcel will not locally change the geostrophic
wind but it can advect geostrophic relative vorticity
D2H vag DH ug DH u
 
meridionally; thus the analysis is non-quasi-geos-
¼ f  ½3
Dt2 Dt Dt trophic but omits the two-way interaction of the
geostrophic adjustment problem.
The second term in the parenthesis in eqn [3] can be Insertion of eqns [1] and [4] into eqn [3] and
replaced with the right-hand side of eqn [1], coupling rearrangement leads directly to
the two equations of motion. The first term in the
parenthesis can also be related to the meridional D2H vag 

þ f f þ zg vag ¼ 0 ½5
ageostrophic wind via the following approximation: Dt 2

DH ug qug This second-order differential equation bears a close

ffi v ¼ vag zg ½4 resemblance to the stability equation for static
Dt qy
instability, with the meridional ageostrophic wind
The relation in eqn [4] derives from the steady, replacing the vertical displacement and f(f 1 zg)
zonally uniform mass field in Figure 3. It also assumes replacing the buoyancy frequency. As in the static
 INSTABILITY / Inertial Instability 1007

L Note that the inertial oscillation period in eqn [6]

p − ∆p reduces to the usual textbook form if zg 5 0; the more
ug general form in eqn [6] has been applied successfully to
p
constant-pressure radiosonde trajectory periodicities.
Weak inertial stability is analogous to weak static
p + ∆p stability, in which forcing leads to a larger response
H
y than in strongly stable conditions.
The criterion in eqn [8] can be interpreted in a
x
variety of ways. In the Northern Hemisphere, it is
Figure 3 Schematic illustrating the physical situation posed in equivalent to negative geostrophic absolute vorti-
the mathematical derivation of the inertial instability criterion (eqn city. For statically stable conditions, eqn [8] implies
[8]), in which a parcel (small circle) in a background flow with
negative potential vorticity in a geostrophic flow in the
horizontal shear is subject to a perturbation (heavy arrow).
Northern Hemisphere. If eqn [8] is calculated on an
isentropic surface then it is identical to the criterion for
stability problem, we assume a wave solution and symmetric instability (see Instability: Symmetric
obtain the following cases: Stability). Interestingly, the necessary criterion for

barotropic instability is the meridional derivative of
f f þ zg > 0 eqn [8].
2p The e-folding times of inertial instability are, by eqn
stable inertial oscillation with period 
1=2 [8], dependent on the latitude and the magnitude of the
f f þ zg anomalous absolute vorticity. Observations and mod-
½6 eling studies suggest e-folding times as short as a few

hours in the mesoscale midlatitide troposphere and
f f þ zg ¼ 0 inertial neutrality ½7 around one day in the equatorial middle atmosphere.

These time scales are much longer than for static
f f þ zg o0 instability but are usually shorter than for barotropic
1 instability.
inertial instability with e-folding time 
1=2 Extensions beyond eqn [8] are possible in some
f f þ zg
cases. For a circular vortex, the criterion in eqn [8] is
½8 modified only slightly, with the sum of the shear and
0
1.20•1 -8

0
0
1.50•10 -8

6.00•10-9

3.00•10 -9
9.00•10 -9
0

0.1
3.00•10

•10-8-8
0
-9

2.70•1
2.40

1.0
Pressure (hPa)

2.10•10
-8
1.80•10-8
1.50•10 -8

10.0
6.00•10

1.50•10-8
1.20•10-8
•10 -9

9.00•10

1.80•10
-9
3.00
9.00•10

-9

-8
-9

100.0
−80 −60 −40 −20 0 20 40 60 80
Latitude
Range from −1.3828446e−09 to 2.5954147e-07 s−2 Contour = 1.50000e−09

Figure 4 The inertial instability criterion (eqn [8]), calculated from Limb Infrared Monitoring of the Stratosphere (LIMS) geopotential
heights for the period 12–17 December 1978. Negative values, corresponding to inertial instability according to eqn [8], are shaded.
(Reproduced with permission from Knox JA (1997) Generalized nonlinear balance criteria and inertial stability. Journal of the Atmospheric
Sciences 54: 967–985.)
1008 INSTABILITY / Inertial Instability

1 1
1
1
0.1 −1 −3
3

1 3 5 1
1 1

−1
−3 1
−1 −1

1 1
1.0
−1
−1
Pressure (hPa)

−1
1 1
1

−1
−1 1
1 −1
1 −1
10.0

−1 1
1

1 1

−1 −1
100.0
−20 −10 0 10 20 30 40
Latitude

Figure 5 Vertically layered ‘pancake structures’ in the lower mesosphere in LIMS eddy temperatures (contour interval 1 K, negative
regions shaded) at the international dateline on 10 December 1978, at the beginning of a week-long episode of inertial instability. Note that
the temperature anomalies are strongest just south of 101 N, where the inertial instability is centered in Figure 4, and that anomalies of the
opposite sign exist near 301 N, as predicted by the schematic in Figure 2.

curvature vorticities replacing the geostrophic relative
vorticity. Extension of the analysis to the beta plane
does not alter the instability criterion, although the
condition for stability becomes necessary, no longer
sufficient. On the sphere, the metric terms due to the
Earth’s curvature can alter the criterion, but the
change is significant only for high wind speeds and/
or high latitudes. The vertical equation of motion may
be incorporated into the analysis via the thermal wind
law; the resulting instability criterion is eqn [8], with
the absolute vorticity replaced with the Ertel potential
vorticity. Friction in the form of Rayleigh drag
stabilizes the flow linearly, i.e., the more drag there
is, the less unstable/more stable the flow is. Extension
of the theory to zonally asymmetric flows has been
achieved but does not yield an alternative criterion; in
general eqn [8] is used locally with good results, except
in very strong anticyclones.
Unfortunately, the assumptions in the analysis
above preclude any insight into the vertical scale of
the instability. The growth rates for inertial instability
are greatest at smallest scales; it is hypothesized that
eddy diffusion damps these scales out and leads to a
preferred intermediate vertical scale. However, the
instability itself may be the source of the turbulent
eddies that in turn select the vertical scale of the
instability. Thus the eddy diffusion hypothesis for
vertical scale selection, which depends on externally
prescribed approximations such as Rayleigh drag, is
incomplete.
Observed Phenomena Related to
Inertial Instability
Inertial instability is possible wherever the antic-
yclonic relative vorticity rivals the Coriolis parameter
in magnitude. Therefore, it is not confined to any one
region of the planet, although it is likeliest where the
Coriolis parameter is smallest, i.e., the tropics.
Middle Atmosphere
Large-scale inertial instability is observed in the
equatorial lower mesosphere, often lasting about a
 INSTABILITY / Inertial Instability 1009

300

200

100

50

20

10

010000 days of inertial instability at 250 hPa

Figure 6 Climatology of inertial instability using National Centers for Environmental Prediction daily geopotential height analyses
(horizontal resolution 5 381 km at 601 N) at 250 mb for December 1966–December 1994 (29 Decembers). Contours indicate the number
of analyses in which eqn [8] was satisfied over the 899 days in the study. (Figure courtesy Russ Schumacher, Colorado State University,
and David Schultz, National Severe Storms Laboratory.)

−2
−2
930 0
T flank of the stratopause semiannual oscillation east-
4 −5
−2
942
4 shown that Rossby wave breaking in the tropics acts
Figure 7 Eta model initial analysis of absolute geostrophic
vorticity (shaded, only negative regions shown; scale is
 105 s  1), geopotential height (heavy contours; in decameters),
and horizontal divergence (light contours; scale is  105 s  1) at
300 mb over Texas and Oklahoma valid at 0000 UTC on 4 May
1999. The centroid of tornado reports at this time is indicated with a
large T. Shaded regions, corresponding to inertial instability via eqn
[8], are colocated with severe weather and with high values of
divergence. The inertially unstable region over Texas was
associated with 29 severe weather reports. (Figure courtesy David
Schultz, National Severe Storms Laboratory.)
week and occurring within a week or two of the boreal
winter solstice. The instability develops in regions of
negative absolute vorticity (Figure 4) on the poleward
erlies; there is also evidence for it around the summer
solstice. Its hallmarks are layered, nearly stationary
‘pancake structures’ in the eddy temperature field
(Figure 5) that have a latitudinal width of 10–20
degrees of latitude and a vertical wavelength of
roughly 10 km.
Numerical simulations and observations have
as a trigger for the onset of equatorial inertial
instability and influences its zonally asymmetric
nature in winter. The divergent character of the
instability, and some recent observational work,
suggest a role for inertial instability in the latitudinal
transport of tracers in the tropical middle atmosphere.
Recent research strongly indicates that Rossby-
wave-triggered inertial instability excites the two-day
wave of the equatorial stratopause region. The
instability may also play a role in some stratospheric
sudden warmings and other high-latitude phenomena.
1010 INSTABILITY / Inertial Instability
Figure 8 National Weather Service Doppler radar image from Sullivan, Wisconsin (circle at center of image) at 1133 UTC on 14 July
1995. The banded echoes extending east-southeast across central Wisconsin correspond to elevated convection at dawn during a deadly
heat wave.
Upper Troposphere
Inertial instability has been sought for in jet stream
analyses since the 1940s. A long-term climatology of
inertial instability based on NCEP geopotential
heights (Figure 6) reveals why this has been a difficult
search: the criterion for the instability is achieved in
the data-rich midlatitudes only once every few years at
most. (The occurrence rate is of course higher for
higher-resolution data sets, but is still rare.) However,
Figure 6 suggests that inertial instability is a fairly
common phenomenon in the subtropical upper tropo-
sphere, particularly on the equatorward flank of the
East Asian jet. Recent work has shown layered
disturbances in low-PV regions near the tropopause
just south of Japan.
Inertial instability can enhance the outflow from
mesoscale convective systems such as thunderstorms,
‘tropical plumes’, and hurricanes (the Rossby radius of
deformation, an estimator of the outflow width, is
closely related to eqn [8] and is infinite in the case of
inertial instability). However, the evidence for its role
in both hurricane outflow and severe thunderstorms is
mixed. In Figure 7, the upper-tropospheric flow pattern
for the devastating 3 May 1999 Oklahoma City,
Oklahoma, tornado outbreak is depicted. Inertially
unstable regions are colocated with strong divergence
aloft and severe weather at the surface in this case, but
not in others. The relationship between inertial
instability and convection is still not well understood.
Inertial instability may also have a connection to
hazardous weather from an aviation perspective. The
gravity wave radiation expected from strongly antic-
yclonic regions, such as inertially unstable circula-
tions, may lead to some otherwise unexplained
instances of clear-air turbulence.
Lower-to-Middle Troposphere
Inertial instability below the jet stream level is
confined to the tropics and a few rare instances in the
 INSTABILITY / Inertial Instability 1011

Figure 9 Potential vorticity (.025 PV unit contour interval, only negative values contoured) and equivalent potential temperature (2 K
contour interval) at 4 km over the upper Midwest United States, as determined by a 24-hour forecast from the University of Wisconsin non-
hydrostatic model (UW-NMS; horizontal resolution 6.67 km, vertical resolution 200–1000 m) valid at 1200 UTC 14 July 1995. Note the
region of negative PV extending east-southeastward across central Wisconsin. The equivalent potential temperature field indicates the
near-horizontal character of the flow over the region.

midlatitudes. The cross-equatorial Asian monsoon
circulation appears to accelerate toward the coast of
India under the influence of inertial instability. The
divergence – convergence couplets of inertial instabil-
ity appear to determine the location of near-equatorial
convection and the mean latitude of the Intertropical
Convergence Zone.
Intense extratropical anticyclones may possibly
exhibit inertial instability. This is the best explanation
for an unforecast pre-dawn elevated convection event

Figure 10 Cross-section of 24-hour forecast vertical motion (contoured every 0.025 m s  1 5 2.5 cm s  1) from UW-NMS valid 1200
UTC 14 July 1995. The cross-section slices NNW-SSE through the region of maximum inertial instability in Figure 9 (vertical mark below
horizontal axis). The checkerboard pattern of vertical motion in the vicinity of the inertially unstable region should be compared with the
schematic in Figure 2.
1012 INSTABILITY / Inertial Instability
near the center of the mid-July 1995 Chicago heat
wave high-pressure system. Elongated bands of light
precipitation (Figure 8) coincided closely in time and
space with a narrow region of negative potential
vorticity (Figure 9). Very high-resolution model
simulations of this event, shown in Figure 10, indicate
a checkerboard pattern of vertical motion (and thus
divergence and convergence) strongly suggestive of
Figure 2. It appears that the rising motions induced by
the instability were substantial enough to cause
condensation and elevated convection where strong
subsidence would normally be expected.
Other Geophysical Fluids
Like the atmosphere, the oceans may also contain
inertially unstable flows wherever the anticyclonic
current is fast and the effect of the planetary rotation is
weak. The layered structure of subthermocline equa-
torial ocean currents, reminiscent of the middle
atmosphere ‘pancake structures’ and confined to
within a degree or two of the Equator, has been linked
with inertial instability. Furthermore, anticyclonic
ocean eddies, even at high latitudes, can satisfy the
criterion in eqn [8]. For example, the overwhelming
tendency for ‘spiral eddies’ on the scales of a few
kilometers to rotate cyclonically has been attributed to
the limiting effects of inertial instability on small-scale
anticyclones.
Farther afield, the atmospheres of Mars and the
large gaseous planets are likely venues for inertial
instability owing to their strong jet structures at low
latitudes. Modeling studies suggest that the very
strong shears on the equatorial flanks of the Martian
jets should be inertially unstable, even in long-term
means.
Summary
After several decades of off-and-on attention, inertial
instability now seems firmly ensconced in the lexicon
of geophysical fluid dynamics. The instability is at its
largest and most observable in connection with
breaking Rossby waves in the tropical middle atmo-
sphere, but it can happen anywhere anticyclonic shear
and/or curvature becomes unusually intense. Its purely
horizontal origins and flattened pancake circulations
make it orthogonal to static instability and much more
difficult to observe. However, as a strongly ageos-
trophic instability it otherwise shares much in
common with static instability. It is perhaps not too
much of a stretch to call inertial instability by the
nickname ‘horizontal convection’, while keeping in
mind that rotation, not density, is at the heart of
inertial instability.
Some outstanding research issues involving
inertial instability include observation and theoretical
explanation of its onset and three-dimensional
structure; elucidation of its relationship to wave
dynamics, other instabilities, and balanced dynamics;
and further investigation of its role in mixing on a wide
range of scales, from convection to planetary-scale
flows.
Glossary
Taylor–Couette instability Also known as centri-
fugal instability, this is a fundamental fluid
dynamical instability that occurs between
differentially rotating cylinders. The bibliographic
entry for Donnelly should satisfy the non-expert’s
curiosity.
Rayleigh drag An expedient approximation to dis-
sipation used in theory and modeling, it assumes
that dissipation is proportional to the wind speed
times a constant, i.e., it is a linear drag on the wind
speed.
Ertel potential vorticity A generalized version of
potential vorticity that is valid for baroclinic flows
and is conserved for adiabatic and frictionless
conditions.
‘PV unit’ A way of reducing the complexity of the
units of Ertel potential vorticity; 1 PV unit 5 10  6
m2 s  1 K kg  1. Values below zero indicate inertial
instability according to extensions to eqn [8]; values
above approximately 1.5 units usually indicate
stratospheric air.
Rossby radius of deformation A fundamental length
scale of the atmosphere that is, among other things,
the distance over which the height field adjusts
during the geostrophic adjustment process. It is
also a measure of the radially confined nature of a
vortex and is smallest for cyclones, largest for
anticyclones.
See also
Angular Momentum of the Atmosphere. Anticy-
clones. Dynamic Meteorology: Balanced Flows;
Waves. Instability: Symmetric Stability. Laboratory
Geophysical Fluid Dynamics. Mesoscale Meteoro-
logy: Mesoscale Convective Systems. Middle Atmo-
sphere: Planetary Waves; Semiannual Oscillation; Trans-
port Circulation. Monsoon: Dynamical Theory. Ocean
Circulation: General Processes. Planetary Atmo-
spheres: Jupiter and the Outer Planets; Mars. Rossby
Waves. Tropical Meteorology: Inter Tropical Conver-
gence Zones (ITCZ). Turbulent Diffusion. Vorticity.
Wave Mean-Flow Interaction.

Further Reading
Andrews DG, Holton JR and Leovy CB (1987)
Middle Atmosphere Dynamics. New York: Academic
Press.
Angell JK (1962) The influence of inertial instability upon
transosonde trajectories and some forecast implications.
Monthly Weather Review 90: 245–251.
Bjerknes J (1951) Extratropical cyclones. In: Malone TF
(ed.) Compendium of Meteorology, pp. 577–598. Bos-
ton, MA: American Meteorological Society.
Blanchard DO, Cotton WR and Brown JM (1998)
Mesoscale circulation growth under conditions of
weak inertial instability. Monthly Weather Review 126:
118–140.
Boyd JP and Christidis ZD (1982) Low wavenumber
instability on the equatorial beta-plane. Geophysical
Research Letters 9: 769–772.
Clark PD and Haynes PH (1996) Inertial instability on an
asymmetric low-latitude flow. Quarterly Journal of the
Royal Meteorological Society 122: 151–182.
Donnelly RJ (1991) Taylor-Couette flow: the early days.
Physics Today 44(11): 32–39.
Dunkerton TJ (1981) On the inertial stability of the
equatorial middle atmosphere. Journal of the Atmo-
spheric Sciences 38: 2354–2364.
Dunkerton TJ (1993) Inertial instability of nonparallel flow
on an equatorial b-plane. Journal of the Atmospheric
Sciences 50: 2744–2758.
Emanuel KA (1979) Inertial instability and mesoscale
convective systems. Part I: Linear theory of inertial
instability in rotating viscous fluids. Journal of the
Atmospheric Sciences 36: 2425–2449.
Hayashi H, Shiotani M and Gille JC (1998) Vertically
stacked temperature disturbances near the equatorial
stratopause as seen in cryogenic limb array etalon
spectrometer data. Journal of Geophysical Research
103(D16): 19469–19483.
Hayashi H, Shiotani M and Gille JC (2002) Horizontal wind
disturbances induced by inertial instability in the
equatorial middle atmosphere as seen in rocketsonde
observations. Journal of Geophysical Research 10.1029/
2001JD000922 31 July 2002.
Hitchman MH, Leovy CB, Gille JC and Bailey PL (1987)
Quasi-stationary zonally asymmetric circulations in the
equatorial lower mesosphere. Journal of the Atmospheric
Sciences 44: 2219–2236.
Hoskins BJ (1974) The role of potential vorticity
in symmetric stability and instability. Quarterly
Journal of the Royal Meteorological Society 100:
480–482.
Hua BL, Moore DW and Le Gentil S (1997) Inertial
nonlinear equilibration of equatorial flows. Journal of
Fluid Mechanics 331: 345–371.
Hunt BG (1981) The maintenance of the zonal mean state
of the upper atmosphere as represented in a three-
dimensional general circulation model extending up
to 100 km. Journal of the Atmospheric Sciences 38:
2172–2186.
INSTABILITY / Inertial Instability 1013
Knox JA (1997) Generalized nonlinear balance criteria and
inertial stability. Journal of the Atmospheric Sciences 54:
967–985.
Knox JA (1997) Possible mechanisms of clear-air turbulence
in strongly anticyclonic flows. Monthly Weather Review
125: 1251–1259.
Mecikalski JR and Tripoli GJ (1998) Inertial available
kinetic energy and the dynamics of tropical plume
formation. Monthly Weather Review 126: 2200–2216.
Munk W (2001) Spirals on the sea. Scientia Marina 65(Supp.
2): 193–198.
Orsolini YJ, Limpasuvan V and Leovy CB (1997) The
tropical stratopause in the UKMO assimilated analyses:
Evidence for a 2-day wave and inertial circulations.
Quarterly Journal of the Royal Meteorological Society
123: 1707–1724.
O’Sullivan DJ and Hitchman MH (1992) Inertial instability
and Rossby wave breaking in a numerical model. Journal
of the Atmospheric Sciences 49: 991–1002.
Rayleigh Lord (1916) On the dynamics of revolving fluids.
Proceedings of the Royal Society of London Series A 93:
148–154.
Rodwell MJ and Hoskins BJ (1995) A model of the Asian
summer monsoon. Part II: Cross-equatorial flow and
PV behavior. Journal of the Atmospheric Sciences 52:
1341–1356.
Rosier SM and Lawrence BN (1999) The January 1992
stratospheric sudden warming: A role for tropical inertial
instability? Quarterly Journal of the Royal Meteorologi-
cal Society 125: 2575–2596.
Sato K and Dunkerton TJ (2002) Layered structure
associated with low potential vorticity near the tropo-
pause seen in high-resolution radiosondes over Japan.
Journal of the Atmospheric Sciences 59: in press.
Sawyer JS (1949) The significance of dynamic instability in
atmospheric motions. Quarterly Journal of the Royal
Meteorological Society 75: 364–374.
Smith AK and Riese M (1999) Cryogenic Infrared Spectro-
meters and Telescopes for the Atmosphere (CRISTA)
observations of tracer transport by inertially unstable
circulations. Journal of Geophysical Research 104(D16):
19171–19182.
Stevens DE (1983) On symmetric stability and instability of
zonal mean flows near the equator. Journal of the
Atmospheric Sciences 40: 882–893.
Stevens DE and Ciesielski PE (1986) Inertial instability of
horizontally sheared flow away from the equator. Journal
of the Atmospheric Sciences 43: 2845–2856.
Taylor GI (1923) Stability of a viscous liquid contained
between two rotating cylinders. Philosophical Transac-
tions of the Royal Society of London Series A 223:
289–343.
Tomas RA and Webster PJ (1997) The role of inertial
instability in determining the location and strength of
near-equatorial convection. Quarterly Journal of the
Royal Meteorological Society 123: 1445–1482.
Wilson RJ (1997) A general circulation model simulation of
the Martian polar warming. Geophysical Research
Letters 24: 123–126.
1014 INSTABILITY / Symmetric Stability
Symmetric Stability
H B Bluestein, University of Oklahoma, Norman,
OK, USA
Copyright 2003 Elsevier Science Ltd. All Rights Reserved.
Introduction
Symmetric stability is a state of the atmosphere in
which an inviscid, dry-air parcel displaced from its
equilibrium position with respect to some axis along
or about which the flow has no variations, i.e., along or
about an axis of symmetry, experiences a restoring
force which makes it oscillate about its original
position. For axially symmetric displacements in a
ring about an axially symmetric vortex the wave
motions are called inertial or centrifugal waves.
Centrifugal oscillations are like buoyancy waves
with the horizontal centrifugal (inertial) force playing
the role of buoyancy (gravity).
Similar oscillations can also occur in a statically
stable, rotating atmosphere when the thermal
wind shear vector is unidirectional and does not vary
along the direction it is oriented. Parcels in the form
of a tube are displaced in a vertical plane normal to
the thermal-wind vector. In this case the axis of
symmetry is the axis along which the thermal wind is
directed.
If potential temperature increases with height and
the geostrophic absolute vorticity is anticylonic then
the atmosphere is inertially unstable; if the potential
temperature decreases with height and the geostrophic
absolute vorticity is cyclonic then the atmosphere is
gravitationally (or statically) unstable. If the geo-
strophic absolute vorticity is cyclonic and potential
temperature increases with height then the atmos-
phere is both inertially stable and gravitationally
stable; however, if infinitesimal displacements in the
plane normal to the vertical shear are accompanied by
forces that move the air parcel farther away from its
equilibrium position then the atmosphere is symme-
trically unstable. Since the atmosphere is baroclinic,
owing to the thermal wind, this instability is a special
case of baroclinic instability for a flow in which there is
no temperature gradient component along the axis of
symmetry.
When tubes of moist, unsaturated air are lifted in a
symmetrically stable atmosphere to a level at which
condensation occurs (and water and ice loading are
not significant or are neglected) and thence to a level at
which the atmosphere is symmetrically unstable with
respect to saturated processes (i.e., when vertical
trajectories follow surfaces of constant equivalent or
wet-bulb potential temperature instead of surfaces of
potential temperature), then the atmosphere is in a
state of conditional symmetric instability (CSI). CSI is
analogous to conditional instability for air parcels
lifted vertically. Since CSI involves forces that are both
horizontally and vertically directed, the process by
which the instability is released is also referred to as
slantwise convection. When a layer of moist air that is
initially symmetrically stable is lifted to saturation,
and the vertical displacement of air itself creates the
conditions for slantwise convection, then the process is
referred to as potential symmetric instability (PSI),
which is analogous to potential instability for
upright convection. At saturation, CSI and PSI are
equivalent.
The importance of CSI is that it is thought to be
responsible for the formation of some mesoscale bands
of precipitation that are oriented along the thermal
wind. Since the thermal wind is oriented along the
elongated zone of strong temperature gradient asso-
ciated with fronts and is quasi-two-dimensional, CSI
may be triggered in response to slantwise, ageostro-
phic, frontal circulations initiated by confluence/
diffluence acting on a cross-frontal temperature gra-
dient. It is also thought that CSI may be responsible for
eyewall rainbands in some tropical cyclones.
The Parcel Theory of Symmetric
Instability in an Inviscid, Dry
Atmosphere
The analysis of symmetric stability is simplified by
using a parcel approach analogous to that used in
the parcel theory of upright convection. Consider a
Cartesian coordinate system in which there is a
temperature gradient in the y–p plane and that q=qx
of all variables is zero (this choice of an axis of
symmetry is arbitrary; sometimes the y-axis is chosen
to be the axis of symmetry). For simplicity, the
dynamics are described for the Northern Hemisphere.
Consider the quantity
m ¼ u  fy ½1
where u is the x-component of the wind and f is the
Coriolis parameter. In inviscid flow m, the absolute
momentum or pseudo-angular momentum, is con-
served; it is attributed to an infinitesimal tube of air
extending through some point ðy; pÞ infinitely off in
both the þx- and x-directions.
 INSTABILITY / Symmetric Stability 1015

The inviscid momentum equation in the y-direction, or d, and if qy=qz > 0 and qmg =qyo0: If qy=qzo0
with height as the vertical coordinate, is (Figure 1D) or if qmg =qy >0 (Figure 1E), then the
atmosphere is statically unstable or inertially un-
Dv 1 qp

stable, respectively, and not symmetrically unstable.
¼ fu  ¼ f m  mg ½2
Dt r qy Figure 1A depicts neutral stability and Figure 1C
where v is the y-component of the wind, r the density, depicts absolute stability.
p the pressure, D=Dt the total (material) derivative, The thermal wind relation in terms of potential
and the geostrophic absolute momentum temperature is to a good approximation

mg ¼ ug  fy ½3 qug g 1 qy
¼ ½5
qz f y qy
where ug is the geostrophic component of the wind
in the x-direction. Therefore the net force in the The slope of a surface of constant y is therefore
y-direction on a tube is proportional to the difference

f qug =qz
 
between the m of the tube, which is conserved, and the dz
¼ ½6
mg of the environment into which the tube is displaced. dy y g=yðqy=qzÞ
It is assumed for simplicity that the tube does not mix
and the slope of a surface of constant mg is
with its environment. The original value of m of the
tube is just its geostrophic value at its equilibrium  
dz f  qug =qy
point in the y–p plane and that q=qx plane. Therefore ¼ ½7
dy mg qug =qz
there will be a net force in the y-direction on the tube if
it moves into an environment where mg is different It follows that the necessary condition for symmetric
from that of its equilibrium, starting location. Vertical instability is
gradients of mg are associated with thermal wind shear
in the x-direction (i.e., with temperature gradients in ðg=yÞðqy=qzÞ f
the y direction); gradients of mg in the y-direction are Ri ¼
2 o ½8
qug =qz z gþf
associated with geostrophic absolute vorticity.
The inviscid vertical equation of motion is where Ri is the Richardson number for the geostrophic
wind and zg is the geostrophic vorticity, which for
Dw g 0
¼ ð y  yÞ ½4 symmetric flow ðq=qx ¼ 0Þ is qug =qy. In typical
Dt y
synoptic-scale flow in midlatitudes the geostrophic
where w is the vertical velocity, g the acceleration of vorticity is an order of magnitude smaller than f ; then
gravity, y the potential temperature of the environ- the necessary condition for symmetric instability is
ment, and y 0 the potential temperature of the tube. If that Rio1. In the vicinity of fronts where geostrophic
the flow is adiabatic and there is no diffusion of heat vorticity is much larger, Ri must be smaller.
then y0 is conserved following the motion of the tube. It Ertel’s potential vorticity for an atmosphere in
is assumed that the environment is not disturbed by the geostrophic and hydrostatic balance is
tube’s motion so that there is no vertical perturbation
Cp g qy zg þ f
 
pressure-gradient force. Therefore there will be a net 1
Z¼ f  ½9
force in the vertical on the tube if it moves into an r y qz f Ri
environment where y is different from that of its
equilibrium, starting location. It follows from eqn [8] that an equivalent necessary
Whether or not there is a restoring force on the tube condition for symmetric instability is that Ertel’s
that brings it back to its equilibrium point about which potential vorticity for the geostrophic wind is negative
it undergoes a stable oscillation (symmetric stability) (anticyclonic in either hemisphere). Since
or whether is continues to move in the direction of its

f
displacement (symmetric instability) depends on how zg þ f y ¼ zg þ f z  ½10
Ri
the surfaces of mg and y are oriented and what the

direction of displacement is with respect to the where zg þ f y is the geostrophic absolute vorticity

surfaces (Figure 1). Symmetric instability is possible evaluated on an isentropic surface and zg þ f z ¼
(Figure 1B) if the slope of the y surfaces is greater than f  qug =qy is the geostrophic absolute vorticity evalu-
the slope of the mg surfaces and if the tube is displaced ated on a surface of constant height, then nega-
infinitesimally along a plane whose slope is interme- tive (anticyclonic in either hemisphere) isentropic
diate between that of the y surfaces and that of the mg geostrophic absolute vorticity is also an equivalent
surfaces (i.e., along paths a or c, but not along paths b necessary condition for symmetric instability.
1016 INSTABILITY / Symmetric Stability

z + ∆
mg + ∆mg
dz dz ∂ ∂mg
a = > 0, <0
mg dy dy ∂z ∂y
mg

b Ri =
f
ζg + f
(A) y

z
+ ∆
mg + ∆mg
dz dz ∂ ∂mg
> > 0, <0
mg dy dy mg
∂z ∂y
d a
f
b Ri <
(B) c
y ζg + f

z
mg + ∆mg + ∆
mg dz dz ∂ ∂mg
< > 0, <0
a dy dy mg ∂z ∂y
d
f
Ri >
c b ζg + f
(C) y

z − ∆

∂ ⎡ ∂mg ⎤
a <0 ⎢ < 0
∂z ⎣ ∂y ⎦

b
(D) y
z
mg + ∆mg
mg ⎡ ∂ ⎤ ∂mg
⎢⎣ ∂z > 0⎦ ∂y > 0
a

b
(E) y

Figure 1 Stability criteria for symmetric flow ðq=qx ¼ 0Þ in terms of the slope of the mg (solid lines) and potential-temperature y (dashed
lines) surfaces. Displacements in the directions a; b; c, and d are referred to in the text. (A) Neutral stability for displacements along the y
and mg surfaces; otherwise stable for infinitesimal displacements. (B) Symmetric instability: unstable for infinitesimal slantwise
displacements intermediate in slope between that of y surfaces and mg surfaces; otherwise stable. (C) Absolute stability: stable for any
infinitesimal slantwise displacement. (D) Gravitational instability: unstable for any infinitesimal slantwise displacement not along a y
surface. (E) Inertial instability: unstable for any infinitesimal slantwise displacement not along an mg surface. (Adapted with permission
from Bluestein HB (1993) Synoptic-Dynamic Meteorology in Midlatitudes. Vol. II: Observations and Theory of Weather Systems. New
York: Oxford University Press.)

It can also be shown that the necessary condition for
symmetric stability is equivalent to the ellipticity
condition for the Sawyer–Eliassen equation, which
describes the vertical circulation about a front forced
by geostrophic confluence/diffluence and differential
diabatic heating and whose dynamics are governed by
the geostrophic-momentum approximation. Since the
Sawyer–Eliassen equation is a second-order, constant-
coefficient, partial differential equation, the condition
of ellipticity is necessary for it to have unique solu-
tions. Thus, balanced frontal circulations are possible
only if the atmosphere is symmetrically stable.
The Parcel Theory of Slantwise
Convection in an Inviscid, Moist
Atmosphere
The analysis of symmetric instability in a moist
atmosphere is complicated by latent heat release, and

water and ice loading. The governing momentum
equation remains eqn [2]. The governing vertical
equation of motion, on the other hand, is different
from eqn [4] since it must account for latent heat
release, and if there is condensate, for water and ice
loading also. Surfaces of constant entropy that
account for latent heat release and for condensate
loading replace potential temperature in eqn [4]. If
both the environment and the tube are unsaturated
and there is no condensate, then virtual potential
temperature ðyv Þ may be used in place of potential
temperature. If the tube is saturated and the environ-
ment unsaturated, and condensate ignored, then
potential temperature of the tube may be replaced by
equivalent virtual potential temperature ðyev Þ; if both
the tube and the environment are saturated, and
condensate ignored, then the potential temperature of
both the tube and environment may be replaced by
equivalent virtual potential temperature. (Wet-bulb
virtual potential temperature ðywv Þ may be used
instead of equivalent virtual potential temperature.)
For the purpose of illustration consider an atmos-
phere that is unsaturated and has no condensate, but is
moist. Suppose that the distribution of yv ; yev , and mg
is as shown in Figure 2. Since the slope of the surfaces
of constant yv ; are not steeper than the surfaces of
constant mg at low levels, the atmosphere there is
symmetrically stable or even neutral with respect to
unsaturated displacements. However, there are re-
gions aloft where the slope of surfaces of constant yev
are steeper than the surfaces of mg : In these regions, if
condensate loading is ignored then the atmosphere
is symmetrically unstable with respect to infin-
itesimal saturated displacements. Since the stability
depends upon whether or not a tube is saturated or
unsaturated, the symmetric instability condition is
conditional.
Suppose an unsaturated tube at low levels is lifted a
finite distance along a surface of constant yv (e.g., by
the ascending branch of a frontal circulation or more
slowly as a result of quasigeostrophic forcing) until it
reaches its lifting condensation level (LCL) and that
condensate is ignored. If lifted any further, it follows a
surface of constant yev . Thus far the tube is neutrally
buoyant. Owing to the inclusion of the effects of latent
heat, the surfaces of constant yev have slopes different
from that of the surfaces of constant yv : The m of the
tube is greater than that in its environment everywhere
to the right of the original mg surface; therefore
according to eqn [2] the tube is symmetrically stable
because it feels a restoring force that has a component
to the left; were the tube forced no further, it would
become negatively buoyant and move back down and
to the left towards its original equilibrium position.
The tube is symmetrically stable even though it
INSTABILITY / Symmetric Stability 1017
z
mg +∆mg mg
Tropopause
ev, wv
LNB
v
ev +∆ ev,
wv +∆ wv
LFS
v+∆ v
LCL
v constant
y
Figure 2 Idealized example of a vertical cross section in the
Northern Hemisphere, normal to the thermal-wind shear vector,
showing surfaces of constant mg (solid lines), and constant yv ; yev ;
and ywv In this example yv ; yev ; and ywv (dashed lines) increase with
height (gravitational and conditional stability), mg decreases with
increasing y (inertial stability), and yv decreases with y (baroclinic
atmosphere, with colder air at larger values of y). Lifting conden-
sation level (LCL); level of free slantwise convection (LFS); level of
neutral buoyancy (LNB). Below the LCL the slope of the yv surface
is less than that of the mg surface. Note that the slope of yev and ywv
surfaces is greater than the slope of yv surfaces because the lapse
rate of a yv surface is greater than that of a yev or ywv surface, and
yv ; yev ; and ywv decrease with y. (Adapted with permission from
Bluestein HB (1993) Synoptic-Dynamic Meteorology in Midlati-
tudes, Vol. II: Observations and Theory of Weather Systems. New
York: Oxford University Press.)
is saturated and the slope of the surfaces of constant
yev is greater than the slope of the surfaces of constant
mg because the tube has undergone a finite displace-
ment rather than an infinitesimal displacement.
If the tube is lifted further, however, so that
eventually it crosses to the left of the original mg
surface, and it is displaced upward and to the right at a
slope intermediate between that of the yev and mg
surfaces, then according to eqns [2] and [4] it would
continue to accelerate upward and to the right if it
were not forced any more. The level at which it would
first realize symmetric instability is called the level of
free slantwise convection (LFS), in analogy with the
level of free convection (LFC) for upright convection.
Eventually the tube will reach a level at which it
crosses back to the right side of the original mg surface
where the slope of the yev surfaces is now less than that
of the mg surfaces. Above this level, the level of neutral
buoyancy (LNB) for slantwise convection, the atmos-
phere is symmetrically stable. The LNB tends to
be near the tropopause, where y surfaces are more
horizontally oriented owing to the strong static
stability (eqn [6]) in the lower stratosphere and where
mg surfaces are more vertically oriented owing to the
1018 INSTABILITY / Symmetric Stability

lack of horizontal temperature gradient and thermal- of Ug in midlatitudes, the horizontal scale of CSI
wind shear (eqn [7]) at the tropopause. precipitation bands is on the order of 100 km, which is
The amount of kinetic energy it takes to lift a tube to mesoscale.
its LFS is called the slantwise convective inhibition
(SCIN). The potential energy in the environment that
is converted into kinetic energy of the tube while it is Observations of Conditional
symmetrically unstable is called the slantwise convec- Symmetric Instability
tive available potential energy (SCAPE). SCIN and
Slantwise convection may be triggered suddenly after
SCAPE are the analogs to the convective inhibition
a gradual buildup in SCAPE by synoptic-scale or
(CIN) and convective available potential energy
mesoscale processes, or it may be in statistical equi-
(CAPE) in the parcel theory of upright convection. librium so that the SCAPE is nearly zero and constant.
The SCAPE is equivalent to CAPE computed along a
Since the latter is the frequently observed case, it is not
surface of constant mg : In the geostrophic coordinate
easy to use SCAPE as a forecast tool because its
system used in semigeostrophic theory, mg surfaces are
absence does not preclude the possibility that CSI is in
parallel to the geostrophic coordinate, which is
fact being released; the existing observational network
directed opposite in direction to the horizontal tem-
cannot resolve the space and time scales of the
perature gradient. Thus, SCAPE is CAPE computed in
production of CSI and its release when the atmosphere
geostrophic coordinates. In the limit of vanishing
is in statistical equilibrium.
baroclinicity, mg surfaces become vertically oriented Bands of precipitation ahead of warm fronts may
(see eqn [7] when qug =qz ! 0) and SCAPE is identical
be due to CSI. The bands are important in that
to CAPE.
large amounts of precipitation may accumulate in
If the atmosphere is saturated then an equivalent
mesoscale regions, which makes the forecasting of
necessary condition for CSI is that Ertel’s potential
floods and heavy snow difficult; synoptic-scale
vorticity for saturated moist processes is negative
dynamics, on the other hand, can explain only how
(anticyclonic in either hemisphere). If the effects of
lower amounts of precipitation accumulate over
condensate are accounted for then the criteria for CSI
broader regions. The slanted convection in the
are more complicated. The loading depends on what eyewall of some rapidly intensifying tropical cyclones
phase of water substance is present; condensate that
might also be a result of CSI. In this case, the
falls out does not follow air parcel motion and may
temperature gradient is directed toward the center of
evaporate into unsaturated air.
the tropical cyclone.
Symmetrically unstable tubes undergoing slant-
When Ertel’s potential vorticity for moist processes
wise ascent must be compensated for by regions of
is very small, i.e., when the atmosphere is nearly
slantwise-descending air. The regions of slantwise-
neutral with respect to CSI, the atmosphere’s response
descending air modify the environment so as to
to frontogenetical forcing is enhanced. Since frontal
make the tubes less symmetrically unstable, just as secondary circulations and the secondary circulations
compensating subsidence around buoyant clouds
in tropical cyclones themselves can produce preci-
(according to the slice method in the theory of up-
pitation bands it is therefore not always easy to
right convection) warms the environment and lessens
distinguish between bands forced by the secondary
the buoyancy in the clouds (i.e., the CAPE is dimin-
circulations and the bands forced by CSI.
ished). The narrower the slantwise ascending branch
and the wider the slantwise descending branch, the
less is the slantwise acceleration (i.e., the less is the See also
SCAPE). The most unstable configuration is one of
Dynamic Meteorology: Overview. Instability: Inertial
thin, relatively rapidly slantwise-ascending layers
Instability. Mesoscale Meteorology: Cloud and Precip-
of saturated air sandwiched in between thick, less
itation Bands.
rapidly slantwise-descending layers of unsaturated air
that are being cooled evaporatively from precipitation
falling out from above. If the most unstable configu- Further Reading
ration is the one most likely to occur, then CSI
Bluestein HB (1993) Synoptic-Dynamic Meteorology in
precipitation bands should be relatively narrow and
Midlatitudes, Vol. II: Observations and Theory of
widely spaced. Weather Systems. New York: Oxford University Press.
The horizontal scale of CSI precipitation bands Emanuel KA (1994) Atmospheric Convection. New York:
estimated from the horizontal extent of a sloping mg Oxford University Press.
surface (eqn [7]) is Ug =f ; where Ug is the change in Holton JR (1992) An Introduction to Dynamic Meteoro-
geostrophic wind in the layer of CSI. For typical values logy. New York: Academic Press.

Lilly DK (1986) Instabilities. In: Ray PS (ed.) Mesoscale
Meteorology and Forecasting, pp. 259–271. Boston MA:
American Meteorological Society.
Schultz DM and Schumacher PN (1999) The use and
misuse of conditional symmetric instability. Monthly
Weather Review 127: 2709–2732. Boston, MA:
American Meteorological Society.
Wave-CISK
C S Bretherton, University of Washington, Seattle, WA,
USA
Copyright 2003 Elsevier Science Ltd. All Rights Reserved.
Introduction
In 1964, Charney and Eliassen introduced the term
CISK (conditional instability of the second kind) to
describe a positive feedback between deep moist
convection and a large-scale circulation. They hy-
pothesized that hurricane intensification was such a
process, in which surface friction helps to induce low-
level convergence into a vortex, resulting in deep
convection and latent heating that amplify the vortex.
In the late 1960s, Yamasaki and Hayashi first consid-
ered the feedbacks between deep convection and large-
scale equatorial waves. In their models, convection
could intensify (destabilize) the wave in some cases
through purely inviscid processes not dependent on
feedbacks with surface drag or surface thermodynam-
ic fluxes. In 1974, Lindzen termed this destabilization
wave-CISK, to distinguish it from Charney and
Eliassen’s frictional CISK, and put forth perhaps the
most expansive view of the role of wave-
CISK, implicating it in the development of tropical
circulations from squall lines to the Madden–Julian
Oscillation.
Wave-CISK has proved a somewhat slippery hy-
pothesis to test, and has largely fallen from favor
among specialists in convective dynamics. The pre-
dicted instabilities are very sensitive to the represen-
tation of cumulus convection. For simple models the
fastest-growing instabilities have very short wave-
lengths, and are not clearly separable from conven-
tional conditional instability of individual cumulus
clouds. However, wave-CISK is a mode of instability
permitted by many convective parameterizations,
including some used in climate models, so it can be a
useful concept in interpreting model output even if
physically dubious.
The class of convective parameterizations that tend
to lead to wave-CISK instabilities are those that
diagnose convective mass flux based on column-
INSTABILITY / Wave-CISK 1019
Thorpe AJ (1999) Dynamics of mesoscale structure associ-
ated with extratropical cyclones. In: Shapiro MA and
Gronas S (eds) The Life Cycles of Extratropical Cyclones,
pp. 285–296. Boston, MA: American Meteorological
Society.
integrated horizontal moisture convergence. The
Arakawa–Schubert scheme, a typical ‘quasi-equilibri-
um’ closure in which the convective mass flux is chosen
so as to regulate the local convectively available
potential energy, does not support wave-CISK, as
shown by Stark. Neither does the Betts–Miller scheme,
a typical moist adjustment convective parameterizat-
ion, as shown by Neelin and Yu, although Hayashi and
Golder showed that CISK could be excited if the
convective adjustment turns on and off frequently. In
recent years, moisture-convergence-based convective
parameterizations have been criticized for using a
nonlocal measure (moisture convergence) to regulate a
local thermodynamic process (convection), allowing
unrealistic soundings to develop. Furthermore, the
development of wave-CISK is strongly influenced by
the vertical profile of convective heating perturbation
selected by the parameterization. ‘Top-heavy’ heating
perturbations concentrated in the upper troposphere
are most favorable for wave-CISK instabilities.
A Simple Wave-CISK Model
In its simplest form, wave-CISK can be phrased in
terms of a nonrotating inviscid gravity wave interact-
ing with a simple parameterization of moist convec-
tion. Wave-induced perturbations in surface heat flux
and radiative fluxes are neglected. Classically, a linear
stability analysis of a small-amplitude wave is used to
assess the convective feedback.
We consider the mathematical structure of an
extremely simple wave-CISK model based on small-
amplitude (linear) two-dimensional inviscid hydro-
static motions of a nonrotating atmosphere. The
pressure velocity o is assumed to be zero at the mean
surface pressure p ¼ ps . At the tropopause pressure pt ,
a similar boundary condition or a boundary condition
that allows upward-propagating gravity waves to
radiate out of the domain may be applied.
The mean convection is assumed to maintain the
same profile of water vapor mixing ratio qðpÞ every-
where. This is a very strong simplifying assumption on
the column moisture budget, and one of the weakest
1020 INSTABILITY / Wave-CISK
links in simple wave-CISK models. In addition, most
such models have not explicitly attempted to maintain
a consistency between the mean temperature and
moisture profiles that ensures that boundary layer air
will be conditionally unstable, and hence able to
convect, but will not penetrate much above the
tropopause. Such consistency requires a basic state in
which the moist static energy is similar at the tropo-
pause to that in the boundary layer, which relates the
assumed qðps Þ to the assumed troposphere-mean static
stability.
The perturbation convective heating in any column
is assumed to be caused by the conversion of the
converged moisture into rainfall; the resulting latent
heating is redistributed through the column by turbu-
lent convection according to a fixed vertical heating
profile ZðpÞ, which is normalized to have a mean value
of unity averaged over the depth of the atmosphere.
The thermodynamic equation for geopotential pertur-
bation fðx; p; tÞ is eqn [1].
q qf LR
þ so ¼ ZðpÞM ½1
qt qp pCp ðpb
pt Þ
In eqn [1], x is the horizontal coordinate, t is time, s is
the static stability parameter, Mðx; tÞ is the perturba-
tion column-integrated horizontal moisture conver-
gence, R and Cp are the gas constant and isobaric
specific heat of air, and L is the latent heat of
vaporization for water vapor. Using the continuity
equation, the moisture convergence can be written as
eqn [2].
Z ps
qo dp
M¼ qðpÞ ½2
pt qp g
Applying the horizontal momentum and continuity
equations results in a linear, separable equation for
oðx; p; tÞ (eqn [3]).
2 2 2 2
q q o q o LR q M
þs 2 ¼ ZðpÞ 2 ½3
qt2 qp2 qx pCp ðpb
pt Þ qx
With a rigid-lid boundary condition, oðps Þ ¼ oðpt Þ
¼ 0. A radiation boundary condition would relate oðpt Þ
to its vertical derivative. In either case, this equation
admits normal modes of the form o ¼ o ^ ðpÞ
expðik½x
ctÞ. In the case of constant s, Cho and
Pendlebury showed how the complex phase speed c
can be calculated by a transcendental equation most
easily derived by Fourier-analyzing o ^ ðpÞ and the
normalized heating profile ZðpÞ. Unstable modes
ðImðcÞ >0Þ exist for a variety of heating profiles. Use
of a radiation boundary condition instead of a rigid lid
at the tropopause usually has a minor effect on the
unstable modes.
Raymond has given a simple physical description of
wave-CISK by regarding the convective heating as a
series of pulses, each of which acts as a vertically
distributed source of gravity waves. He then showed
that for a wave moving at a particular phase speed, the
gravity wave generated by the pulse heating would
generate low-level horizontal convergence in phase
with the preexisting wave, causing the wave to
amplify.
An important feature of the eigenvalue problem is
that the horizontal wavenumber k cancels out of the
equation. This implies that if an instability is present,
shorter wavelengths will grow fastest (no short-wave
cutoff), since the growth rate is proportional to
ImðkcÞ. This suggests that wave-CISK might manifest
itself at the shortest scale at which the model assump-
tions are still plausible. Since the model assumes an
ensemble convective heating response, this scale
would have to be somewhat larger than the spacing
between convective clouds or cloud groups. However,
on such scales other mechanisms such as cold-pool
dynamics seems to play a more important role in
organizing convection into mesoscale systems.
The short-wavelength behavior is related to the
difficulty of clearly separating wave-CISK in this
model from conventional conditional instability of
individual cumuli. For a given temperature profile,
moistening the near-surface air will render the profile
more conditionally unstable to individual cumuli. Our
simple CISK model exhibits a distorted version of this
same mechanism, rendering dubious the notion of
CISK as a distinct instability of cumulus cloud
ensembles. To see this, we can integrate eqn [2] by
parts as in eqn [4].
ps
qq dp
Z
M¼
o ½4
pt qp g
This expresses the parameterized moisture conver-
gence, and hence the cumulus-induced heating rate, as
proportional to the vertical velocity weighted by
dq=dp . This heating counteracts the adiabatic cooling
associated with rising motion, reducing the effective
static stability of the lower and mid-troposphere. If the
near-surface air is sufficiently humid, this effective
static stability can become negative, promoting short-
wavelength CISK instabilities. Within the framework
of our CISK model, this condition of negative effective
static stability plays the same role as conventional
conditional instability does for growth of individual
cumuli.
The stability analysis is more involved for convec-
tive parameterizations that do not employ a moisture
convergence closure, and only a few such studies
have been published. The complication is that the

right-hand side of eqn [3], which involves the hori-
zontal Laplacian of the heating, is not usually express-
ible purely in terms of o. Although a similar eigenvalue
problem for c can often still be formulated, it usually
must be solved numerically by vertical discretization
and may now have a short-wave cutoff.
Elaborations on the Basic Model
Many elaborations on the above model have been
proposed. In 1970, Hayashi extended a similar model
to continuously stratified motions on an equatorial
beta-plane using separation of variables in the merid-
ional direction. In particular, the equatorial Kelvin
wave has a similar zonal structure and growth rate to a
nonrotating gravity wave. This has led to many
theories that rationalize the tropical Madden–Julian
(intraseasonal) oscillation as a wave-CISK mode.
More sophisticated general circulation model simula-
tions using moisture-convergence-based convective
parameterizations, starting with Hayashi and Sumi in
1986, have also frequently produced intraseasonal
oscillations that have been interpreted as wave-CISK.
In 1979, Davies obtained a short-wavelength
cutoff by assuming a short delay between the mois-
ture convergence and the convection. With a
30–60-minute delay, the fastest growing wavelength
is a few hundred kilometers, corresponding to a typical
size of a mesoscale convective system. In 1987, Lau
and Peng considered ‘positive-only heating’, in which
only upward motion (creating moisture convergence)
produces perturbation heat release. This can be
thought of as a simple way to represent waves suffi-
ciently strong to suppress all convection in their
subsiding branches, but still weak enough to be
approximated by linear dynamics. Such models pro-
duce unstable modes with a propagating narrow band
of ascent surrounded by a broad subsidence region.
This could be considered as a parameterized repre-
sentation of the circulation around a single intense
cumulonimbus.
Additional physical feedbacks have been consid-
ered. In 1990 Wang and Rui considered frictional
wave-CISK, the impact of surface friction and con-
vective heating on an equatorial wave, and found that
the surface drag could stimulate a pattern of convec-
tive heating that helps destabilize an equatorial Kelvin
wave. Mak considered the feedback of cumulus
convection (represented via eqn [2]) with an Eady
model of mid-latitude baroclinic instability. He
showed that the most unstable Eady mode becomes
shorter and intensifies more rapidly when moderate
cumulus-driven latent heating is included; these effects
can be interpreted in part as consequences of a reduced
effective static stability. With sufficiently strong cu-
INSTABILITY / Wave-CISK 1021
mulus heating, his theory also predicted new classes of
boundary-trapped CISK modes. Emanuel found a
baroclinic wave-CISK mode varying perpendicularly
to the wind shear, somewhat akin to symmetric
instability, which can be excited in a broader set of
conditions than classical wave-CISK modes. These
theories await decisive testing against observations
and more sophisticated numerical models.
Current Status of Wave-CISK
Neither observations nor current cloud-resolving nu-
merical model simulations clearly show classical
wave-CISK-like modes. Furthermore, the theoretical
models that predict wave-CISK are based on dubious
parameterizations of cumulus convection. However,
there are many intriguing indications that moist
convection may in fact help destabilize some large-
scale waves through mechanisms not considered in
classical wave-CISK. Some of these include: (i) the
effect of wave-associated surface wind perturbations
on the surface fluxes and boundary layer structure
(wind-induced surface heat exchange, or WISHE); (ii)
radiative feedbacks on the wave associated with
convectively produced anvils or moisture redistribu-
tion; and (iii) feedbacks between the convection and
the humidity profile in the convecting column. The
importance of these convective/large-scale feedbacks
in producing transient variability in the tropics and
parts of the mid-latitudes on all time scales remains an
active and very important topic of research.
See also
Baroclinic Instability. Dynamic Meteorology: Poten-
tial Vorticity. Hamiltonian Dynamics. Instability: Inertial
Instability; Symmetric Stability. Kelvin–Helmholtz Insta-
bility. Lagrangian Dynamics. Quasi-geostrophic
Theory. Rossby Waves. Vorticity.
Further Reading
Charney JG and Eliassen A (1964) On the growth of the
hurricane depression. Journal of the Atmospheric Sci-
ences 21: 68–75.
Cho H-R and Pendlebury D (1997) Wave CISK of equatorial
waves and the vertical distribution of cumulus heating.
Journal of the Atmospheric Sciences 54: 2429–2440.
Davies HC (1979) Phase-lagged wave-CISK. Quarterly
Journal of the Royal Meteorological Society 105:
325–353.
Emanuel KA (1982) Inertial instability and mesoscale
convective system. Part II: Symmetric CISK in a baro-
clinic flow. Journal of the Atmospheric Sciences 39:
1080–1092.
Hayashi Y (1970) A theory of large-scale equatorial waves
genearated by condensation heat and accelerating the
1022 IONOSPHERE
zonal wind. Journal of the Meteorological Society of
Japan 48: 140–160.
Hayashi Yand Golder DG (1997) United mechanisms for the
generation of low- and high-frequency tropical waves.
Part I: Control experiments with moist convective
adjustment. Journal of the Atmospheric Sciences 54:
1262–1276.
Hayashi Y-Y and Sumi A (1986) The 30–40 day oscillation
simulated in an ‘aqua-planet’ model. Journal of the
Meteorological Society of Japan 64: 451–467.
Lau K-M and Peng L (1987) Origin of low-frequency
(intraseasonal) oscillations in the tropical atmosphere.
Part I: Basic theory. Journal of the Atmospheric Sciences
44: 950–972.
Lindzen RS (1974) Wave-CISK in the tropics. Journal of the
Atmospheric Sciences 31: 156–179.
IONOSPHERE
M C Kelley, Cornell University, Ithaca, NY, USA
Copyright 2003 Elsevier Science Ltd. All Rights Reserved.
Introduction
The early Greeks thought that all material was created
from four elements: air, earth, fire, and water. We now
know that the elements are actually combinations of
protons, neutrons, and electrons and that all matter is
made from various combinations of these atomic
building blocks. However, when cataloguing the
possible states of matter the ancients were not so far
off. In our daily lives, the states of matter referred to as
solid, liquid, and gas are clearly related to Aristotle’s
earth, water, and air.
But what about fire? What about a fourth state of
matter? Indeed, there is a fourth state of matter, the
state called ‘plasma’, which could equally well have
been called ‘fire’, since the hotter a flame, the closer it
comes to the plasma state. We dwell for a moment here
on the plasma state itself, since the Earth is surrounded
by just such a medium – a region called the ionosphere
– which is the topic of this article. In fact, we see
throughout this volume that the atmosphere of the
Earth itself includes all four states of matter when one
includes raindrops, snow, ice, and the ionospheric
plasma, in addition to the gaseous component.
The list of earth, water, and air (solid, liquid, and
gas) can be reordered according to the common
knowledge that when a solid is heated it becomes a
liquid and then a gas in processes called ‘change of
phase’. At each phase change, bonds are broken to
form the next phase. More energy is required at each
Mak M (1994) Cyclogenesis in a conditionally unstable
moist baroclinic atmosphere. Tellus 46A: 14–33.
Neelin JD and Yu J-Y (1993) Modes of tropical variability
under convective adjustment and the Madden–Julian
oscillation. Part I: Analytical theory. Journal of the
Atmospheric Sciences 51: 1876–1894.
Raymond DJ (1983) Wave-CISK in mass-flux form. Journal
of the Atmospheric Sciences 40: 2561–2572.
Stark TE (1976) Wave-CISK and cumulus parameterization.
Journal of the Atmospheric Sciences 33: 2383–2391.
Yamasaki M (1969) Large-scale disturbances in the condi-
tionally unstable atmosphere in low latitudes. Papers in
Meteorology and Geophysics 20: 289–336.
Wang B and Rui H (1990) Dynamics of the coupled moist
Kelvin–Rossby wave on an equatorial beta plane. Journal
of the Atmospheric Sciences 47: 397–413.
step, and by the time one gets to the transition from gas
to plasma, quite a lot of energy is required. In this final
step the phase change required actually rips an
electron away from the gaseous atom or molecule,
leaving a positive ion behind. Since electric charge is
conserved, the new state of matter remains neutrally
charged on average (equal numbers of positive ions
and negative electrons), but these constituents may
seldom run into each other and hence have little
chance to recombine to the atomic (gaseous) state. If
this is so, a plasma is born.
How much energy is needed to rip apart atoms? The
response is ‘a few electron volts’. This is surprising at
first, since we are all familiar with the batteries in our
radios and automobiles, which operate at voltages of
1.5–12 V. We must remember, however, that batteries
run on chemical reactions that themselves involve
exchanges of electrons between atoms and molecules,
so the volt is a natural-sized unit for ionic bonds. How
can we relate this unit to temperature? Suppose we
have a pure gas like hydrogen; how hot does it need to
be to become a plasma? Suppose the gas is already hot
enough that the H2 molecules have separated into pure
hydrogen as the bonds are broken due to collisions of
molecules with each other. The proton–electron pair
that makes up a hydrogen atom has a binding energy
requiring 13.5 electron volts (eV) to separate the
particles. Thus to have significant numbers of atom
collisions result in ionization as in eqn [I], the average
energy of the colliding H atoms must be the order of
13.5 eV.
H þ H ! H þ Hþ þ e
½I

Just what does a volt signify? The units of a volt are
joules per coulomb (J C
1). This means that if a 1-volt
battery is capable of storing one coulomb of charge, a
total of one joule of energy is available (enough to lift a
mass of one kilogram to a height of 100 cm on the
Earth’s surface). A typical car battery can supply a
current of 100 amperes (100 coulombs per second) for
an hour, although not continuously, so it stores about
360 000 C. Thus the 12 V car battery stores about
4.3 million joules (12 J C
1  360 000 C). Since an
electron has a change of only 1.6  10 19 C, an energy
of 13.5 eV corresponds to about 2.2  10 18 J. Now
we need to relate energy to temperature. Kinetic
theory shows that the average energy of a particle in
a gas is equal to 32kB T where kB is Boltzmann’s con-
stant (1.38  10 23 J K 1) and T is the absolute
temperature in Kelvin. If we set this expression
equal to 2.2  10 18 J and solve for T, we obtain
T ¼ 106 280 K! Such a high temperature shows why it
is difficult to produce and control plasmas in the
laboratory or in fusion machines.
The Sun is powered by nuclear fusion at its core and
is therefore very hot; hence it follows that much of its
matter is ionized. Gravity controls this fiercely hot
object. The Earth is much cooler, and hence it is not
obvious that there would be a plasma state in its
environs. However, there is a plasma surrounding the
Earth called the ionosphere. The fundamental pro-
duction and loss mechanisms for the Earth’s iono-
sphere are described next. This is followed by a
description of more exotic sources of the plasma
surrounding the Earth, including the solar wind,
magnetic storms, meteors, and the auroras. These
sources are localized in time and space and can be
linked under the umbrella of weather processes in
space or, in short, space weather. Space weather is also
influenced by sources of energy and momentum from
the Earth, the dense atmosphere below, and sources
including waves from severe storms, orographic fea-
tures, and earthquakes, as well as the release of stored
energy via plasma instabilities.
Sources and Fundamental Features
of the Ionosphere
The ionosphere is formed primarily when the most
energetic component of the solar spectrum – the X-
rays and extreme ultraviolet (EUV) light – impact the
illuminated side of the Earth. These high-energy
photons strike the daytime side of the Earth, ionizing
the upper atmosphere and losing energy in the process.
As the beam penetrates the atmosphere, the ionizing
beam becomes weaker and weaker, leaving behind a
layer of ionization. Part of the energy goes into heat as
IONOSPHERE 1023
well as into ionizing the air, resulting in the temper-
ature also rising to values much higher than in any part
of the dense atmosphere below. Life on Earth is thus
protected by its upper atmosphere from these danger-
ous photons, just as the ozone layer absorbs the lower-
energy, but still harmful, ultraviolet component of the
Sun’s spectrum.
We compare and contrast the atmosphere and
ionosphere in Figure 1. The most important atmos-
pheric parameter is temperature, which is plotted
versus height in (A). The key ionospheric parameter is
the number of electrons (which equals the number of
positive ions) per cubic centimeter. This is plotted in
(B) for typical nighttime and daytime conditions.
As anticipated above, the atmospheric temperature
rises from its lowest value near the mesopause (near
200 K) to well over 1000 K in the thermosphere in the
same height range where the daytime ionosphere is
produced. A glance at Figure 1B shows that the
ionosphere does not entirely disappear at night, even
though the sunlight is no longer present to create new
ionization. This is one of the key characteristics of the
Earth’s ionosphere and explains, for example, how
Marconi was first able to send wireless signals across
the Atlantic Ocean at night. To understand why some
of the ionosphere remains through the night, we must
consider the ion chemistry of the region.
At high altitudes (4300 km), production (P) and
loss (L) of ionospheric plasma are both small. The
balance between diffusion and gravitation results in
the so-called hydrostatic equilibrium in which the
plasma pressure (p) is of the form of eqn [1].
h=Hp
p ¼ p0 e ½1Š
In eqn [1], where e is the base of the natural logarithms;
h is height above some reference; p0 is the pressure at
the reference altitude; and Hp is the plasma scale
height, given by eqn [2].
kB T 2kB T
Hp ¼ ¼ ½2Š
ðM=2Þg Mg
In eqn [2], M is the average ion mass and g is the
gravitational acceleration. According to eqn [1], the
pressure falls by a factor of about 2.7 for each altitude
increase of Hp . M is quite close to the average mass of
the neutral atmospheric particles surrounding the
plasma. The factor of 2 comes from the fact that the
average plasma mass is half the ion mass, since the
electron mass is so tiny. The neutral atmosphere
behaves like eqn [1] except that the neutral scale height
Hn is half as large. One conclusion from the above is
that, because the electrons are so light, the ionosphere
extends higher into space than the neutral atmosphere
1024 IONOSPHERE

Neutral gas Ionized gas

Protonosphere
Thermosphere

1000 1000

F Region

Mesosphere 100 E Region

100
Altitude (km)

D Region
Stratosphere

10 10
Day
Night
Troposphere

1 1
0 400 800 1200 1600 103 104 105 106
(A) Temperature (K) (B) Plasma density (cm−3)

Figure 1 Typical profiles of neutral atmospheric temperature (A) and ionospheric plasma density (B) with the various layers designated.
(Reprinted with permission from Kelley MC (1989). Copyright 1989 by Academic Press.)

surrounding it. For reference, Hn is about 50 km and
Hp is about 100 km in the middle ionosphere.
At these altitudes the composition of the atmos-
phere is no longer similar to the surface composition
(which is 79% N2 , 20% O2 1 minor constituents).
The atmosphere is no longer mixed, and lighter atoms
can reach higher altitudes. Also, O2 is photodissoci-
ated into free oxygen atoms. Figure 2 shows the
composition in terms of various atoms, molecules, and
ions versus height for the mid-latitude ionosphere/
thermosphere. We see that oxygen becomes dominant
at 200 km and hydrogen above 700 km. Similarly, the
ionosphere is primarily made up of H1 (with some
He1) at very high altitude, O1 in the height range near
the peak density, and a mixture of O1 1
2 , N2 , and NO
1
in the lower thermosphere. Hydrogen is so light that it
can escape the Earth’s gravity and form the Earth’s
geocorona, a halo of hydrogen analogous to the Sun’s
glowing corona seen during an eclipse. By chance,
hydrogen and oxygen have almost identical ionization
potentials, so charge exchange is a very easy process,
as shown in eqn [II].
H þ Oþ ÐO þ Hþ ½II
Thus, if O1 is surrounded by H gas, after a while an
oxygen ion will give up its charge to form a hydrogen
ion (H1). This explains why O1 ions formed at low
altitudes during daytime become H1 ions at very high
altitudes.
Gravity and pressure are not the only forces with
which the ionosphere must deal. The Earth’s dipole
magnetic field lines force the hydrogen ions to travel
along closed trajectories between the hemispheres,
since following the magnetic lines is easy but moving
across them is not. The particle motion is helical, the

1000 N+ e−
O+ He
He+
Altitude (km)

500 N2
H+ O
300
250 NO+
200 O2+
N+ +
He O2 N2
150 N2+ O Ar
+ NO+ e−
100 O2 O
102 103 104 105 106 107 108 109 1010 1011 1012
Number density (cm−3)

Figure 2 International Quiet Solar Year (IQSY) daytime atmospheric composition. (Reprinted with permission of the MIT Press from
Johnson CY (1969). Ion and neutral composition of the ionosphere. Ann IQSY 5. Cambridge, MA: MIT Press. Copyright 1969 by MIT.)

Plasmapause Plasmasphere
N by [VIb].
5
Earth radii
S
Figure 3 A toroidal region of high plasma density exists around
the Earth on average within the region shown. These magnetic flux
tubes are filled with plasma of ionospheric origin during the day and
discharge only slowly at night.
particles moving in circles around the magnetic field
lines while freely moving parallel or antiparallel to the
direction of the field lines. The result of this motion is
that the entire region, in a toroidal shape (shown in
Figure 3), becomes filled with a hydrogen plasma
during the daytime (whose source is sunlight ioniza-
tion of oxygen coupled with charge exchange). During
the night, this region – called the plasmasphere – starts
to unload downward into the ionosphere by the
reverse process, tending to maintain the oxygen
plasma in the ionosphere during the night with the
whole process starting over the next day. The reason
the plasmasphere abruptly ends at about 4 earth radii
(601 magnetic latitude) is very interesting and is
discussed below.
Refilling from above is not the entire reason that the
ionosphere lasts all night, however. It turns out that a
charged atom cannot easily recombine with an elec-
tron, since in a reaction such as [III] it is very difficult
simultaneously to conserve both energy and momen-
tum, and the reaction rate is very small.
Oþ þ e
! O ½III
However, in reactions [IVa] and [IVb] there are two
end products, and this difficulty does not arise.
Oþ

2 þe ! OþO ½IVa
NOþ þ e ! N þ O ½IVb
Reactions [IVa,b] are called dissociative recombi-
nation and are very fast. This explains why the
molecular ions (seen in Figure 2) at low altitudes
disappear at night, leaving the O1 plasma above as the
distinct nighttime layer (shown in Figure 1). In fact,
reaction [III] is so slow that O1 is actually lost through
a two-step process such as charge exchange [Va]
followed by [Vb].
Oþ þ O2 ! Oþ
2 þO ½Va
Oþ

2 þe ! OþO ½Vb
IONOSPHERE 1025
An alternative route is ion–atom interchange followed
Oþ þ N2 ! NOþ þ N ½VIa
NOþ þ e ! N þ O ½VIb
Both [Va] and [VIb] leave oxygen in an excited state,
which emits both red (630 nm) and green (557.9 nm)
light that is visible from the ground to sensitive
cameras. Such emissions are called airglow and
provide a tool for visualizing the ionosphere, as shown
in the next section.
To summarize thus far: the ionosphere is created
during the daytime by X-rays and EUV from the Sun,
which are absorbed while heating and ionizing the
outer layer of the atmosphere. This heats the gas to
temperatures over 1000 K, explaining why it is called
the thermosphere. The plasma, which is primarily O1
above 200 km, diffuses upward against gravity, reach-
ing high enough that charge exchange with the
geocorona converts the ionosphere to a H1 plasma,
which can escape gravity. The plasma is constrained by
the dipole magnetic field to a toroidal configuration,
filled during the day and emptied at night. Molecular
ions dominate in the lower thermosphere, but they
disappear quickly after sunset, leaving a slowly
decaying O1 layer.
A Day in the Life of the Midlatitude
Ionosphere
The most powerful single tool for ionospheric studies
is an incoherent scatter radar that detects the micro-
scopic fluctuation due to thermal motions in the
ionosphere. The first such instrument, and still the
largest of the 11 now in use worldwide, is near
Arecibo, Puerto Rico. The dish is 1 km in circumfer-
ence and can ‘see’ the ionosphere out to several
thousand kilometers altitude. Figure 4 shows the
plasma content measured by the radar over a full day.
The earlier discussion explains the basic character of
this plot, but not at all its details. We see that the
density is high during the day, even at 100 km. At night
the lower ionosphere rapidly – and almost completely
– disappears, so much so that to see anything at all we
have to change the altitude scale and gray scale. Then
slowly during the night, the density decays to low
values just before sunrise, when the cycle begins again.
But what about the wiggles? Why does the layer move
up and down? And what is the origin of the weak
ionization layers seen at low altitudes?
These effects in large part are due to horizontal
winds and waves in the thermosphere. With a very hot
atmosphere on the daytime side (41000 K) in full
1026 IONOSPHERE
500
450
400
350
Altitude (km)
300
250
200
150
100
Noon 18:00 Midnight
Figure 4 Ionospheric plasma density over Arecibo during a 24-h period (16–17 September 1999; time is Atlantic Standard Time). The
lack of plasma below 250 km at night is due to recombination of molecular ions. The high-altitude plasma and interesting thin layers are due
to ions such as O1, Mg1, and Fe1, which have long lifetimes.
sunlight and a very cool one at night (o800 K), it is not
surprising that strong winds blow from day to night
continuously all over the globe. Unlike the thick lower
atmosphere, the thermosphere has no thermal inertia.
The winds simply blow continuously, acting as a huge
atmospheric thermal tide. Speeds of 200 m s
1
(720 km h
1) are not uncommon. It is very hard to
move the plasma across the magnetic field lines, but
such winds easily move the ionosphere up and down
the magnetic field lines. At Arecibo, the direction of the
magnetic field is at an angle of 451 to the vertical
(pointing northward and downward). When the
wind blows southward, the ionosphere moves up-
ward, much like a ping-pong ball held up against
gravity on an inclined plane by a hair dryer, as shown
in Figure 5.
At high altitude there is very little neutral gas and
recombination is weak. If the wind ceases or blows
north, the ionosphere falls owing to gravity into
regions where reactions [Va,b and VIa,b] can eat away
at it. Thus a southward wind not only elevates the
ionosphere, but it also keeps it high out of reach of the
losses due to the thermosphere. Some of what we see at
night in Figure 4 is due to these winds.
Wind
Figure 5 An illustration showing the analogy between the mid-latitude ionosphere on the right, with atmospheric winds pushing the
ionosphere up the magnetic field, and a light object suspended on an inclined plane.
1.0 × 106
Electron density (cm−3)
5.0 × 105
0
06:00 Noon
The more abrupt changes in height may be due to the
electrical forces that act on the ionosphere. These
electric fields have associated voltages as high as
200 000 V and ionospheric currents as great as
1 000 000 A, yielding power levels of 2  1011 W:
more power than any man-made generator on Earth.
Two major generators provide this electrification and
both involve motion of a conductor across a magnetic
field, in exactly the manner by which generators
convert mechanically rotation of machines into elec-
tric energy. The solar wind is the most powerful of the
two, generating hundreds of kilovolts across the
Earth’s polar regions and causing one of nature’s
most spectacular visual displays, the aurora borealis
and aurora australis (see the next section). The second
generator is the motion of the Earth’s atmosphere,
described earlier. Tides, winds, and gravity waves in
the atmosphere all drive currents and generate electric
fields by the dynamo effect. Because the magnetic field
lines behave like conducting wires, the only voltage
easily allowed is across magnetic field lines. Figure 6
illustrates what happens if a single positive particle is
subjected to orthogonal electric and magnetic fields.
Initially, the particle is accelerated parallel to the
Magnetic field
line
Ionospheric layer
 IONOSPHERE 1027

Ion path
E
lEl
Rolling wheel V=
lBl
B
Electron path

Figure 6 In crossed electric and magnetic fields in vacuum, ions and electrons exhibit the motion shown schematically. The ion path is
specifically shown to be similar to that of a dot on the rim of a rolling wheel.

electric field by the force qE. But once it attains a
velocity, the magnetic force, qV 3 B, deflects it to the
right. Eventually it comes to rest and the cycle starts
over. The trajectory is the same as that of a dot on the
edge of a rolling, nonslipping wheel moving to the
right at a velocity E=B. A negatively charged particle
(electron) starts out in the opposite direction, but is
also deflected to the right, and drifts on average at the
same speed as the positive ions. Since there are equal
numbers of positive ions and electrons there is no net
current, just a net velocity. Above about 150 km,
collisions are so rare that Figure 6 describes the
motion quite well – electric fields are one-to-one
related to the motion of the ionosphere across the
magnetic field lines, whereas winds, gravity, and
diffusion dominate along the direction of the magnetic
field. An eastward electric field over Arecibo, for
example, causes an ðE 3 BÞB
2 drift northward and
upward at a 451 angle. Some of the abrupt height
changes visible in Figure 4 are due to such electric field-
induced motions.
The abrupt height changes might be temporal or
spatial or a combination of both; it is difficult to tell
with a single measurement. But the fact that red light is
emitted in reactions [Va] and [VIb] allows us to
visualize the plasma in two dimensions. The data in
Figure 7 were obtained on the same night using a bare,
back-lit CCD (charge-coupled device) chip illuminat-
ed by a fisheye (all-sky) lens. A narrow (63071 nm)
filter was inserted into the path and the chip was
exposed for 90 s. There were 1024  1024 pixels in the
image, which, at a height of 250 km, covers a circle of
1000 km diameter. The image has been corrected for
the lens effects, vignetting, etc., and projected as if the
viewer were above the Earth looking down rather than
up (hence the map of the Caribbean in its usual
geometry). We see intricate patterns of light and dark
regions, with one of the dark zones positioned right
over the Arecibo Observatory. The ionosphere is
highly structured on this night and is very different
from what would be predicted if only production, loss,
gravity, and diffusion were operating. The sequence of
images taken on this night show that the dark bands
surged poleward from well south of Puerto Rico and
then drifted toward the west. This unexpected be-
havior shows we have much to learn about even the
most well-behaved regions of the ionosphere.
Fire and Ice: The High-Latitude
Ionosphere
Other entries in the Encyclopedia discuss the auroras
and the Earth’s magnetosphere at some length. Here

20:56 LT 22:08 LT 23:12 LT

600

400

200
(km)

200

400

600

600 400 200 0 200 400 600 600 400 200 0 200 400 600 600 400 200 0 200 400 600
(km) (km) (km)

Figure 7 The 630 nm airglow for 17–18 February 1999, superimposed on a map of the Caribbean islands with Puerto Rico in the center.
The regions depleted of airglow (shown black) commenced in the SE and surged to the NW on this night. (Reproduced by permission of the
American Geophysical Union from Kelley et al. (2000) Caribbean Ionosphere Campaign, Year One: airglow and plasma observations
during two intense midlatitude spread-F events. Geophysical Research Letters).
1028 IONOSPHERE

we discuss some of the more striking features of the

ionosphere in this region. First of all, the light
emissions we call the auroras also are due to the
impact of energetic particles, primarily electrons, on
the atmosphere. Some of these leak in from the solar
wind, but most are accelerated from the background
plasma at a height near 5000 km above the Earth. The
Earth has its own ‘cosmic ray’ generator, just as do
pulsars, Jupiter, and other magnetized objects in the
cosmos. Figure 8 shows a view of the Earth from
800 km during a spectacular aurora. Close inspection
of the city lights shows that the aurora reached
Madison, WI, and covered most of the Great Lakes
region.
Since intense light must be accompanied by the
production of plasma, the aurora is a highly dynamic
source of the ionosphere wherever and whenever it
appears. Figure 9 gives some insight into its global
character. The view is from 5000 km and shows rings
of light circling the top (and bottom) of the Earth’s

Figure 9 The sequence begins at 05:29 UT on 2 April 1982

(upper left image) as the NASA/GSFC spacecraft Dynamics
Explorer 1 first views the auroral oval from the late evening side
of the dark hemisphere at low northern latitudes near apogee (3.65
Earth radii altitude) and then from progressively greater latitudes as
the spacecraft proceeds inbound over the auroral oval toward
perigee. The poleward bulge at onset of the auroral substorm is
observed beginning at 06:05 UT (fourth frame). In successive
12-minute images the substorm is observed to expand rapidly in
latitude and longitude. (Photograph courtesy of LA Frank, JD
Craven, and RL Rairden, University of Iowa. Reprinted by
permission of Academic Press from Kelley (1989). Copyright
1989 by Academic Press.)

polar regions like a halo. This auroral oval waxes and

Figure 8 A spectacular view of the eastern half of the United
States during a major auroral display.
wanes in size, sometimes (as seen in Figure 8) reaching
highly populated regions, but usually limited to the
mid-arctic zone.
During certain conditions in the solar wind (when
the interplanetary magnetic field is parallel to the
Earth’s magnetic dipole), the solar wind electric field is
transmitted very efficiently along the Earth’s magnetic
field and throughout both polar ionospheres. Mag-
netic storms often occur at these times and great
auroras result. Since the Earth’s magnetic field is nearly
vertical, near the pole the ðE 3 BÞB
2 drifts are nearly
horizontal. Thus, two huge circulation cells often
occur in the ionosphere. If the magnetic storm contin-
ues for more than an hour or so, the neutral atmos-
phere can also be put into motion. It is remarkable that
the coupled solar wind and ionosphere actually put the
Earth’s atmosphere in motion! Heat is also generated
 IONOSPHERE 1029

1000

900
48
800 42
36
30
700 24
18
Altitude (km)

12
600 6db

500

400

300

200

100
19:00 20:00 21:00 22:00 23:00
Local time

Figure 10 A space weather radar map showing regions of highly turbulent ionospheric plasma over Peru. The dark regions are
analogous to clear air and thunderstorm-related turbulence in the troposphere. (Reproduced by permission of the American Geo-
physical Union from Kelley et al. (1981) Gravity wave initiation of equatorial spread F: a case study. Journal of Geophysical Research
86: 9087.)

by the electric currents, which change the global
circulation.
The polar circulation rips away at the plasma-
sphere, reducing its size and compressing it to lower
latitudes. After the storm and over a few days, the
region refills with cool dense plasma out to about 4
earth radii and 601 magnetic latitude. This altitude–
latitude region is where the solar wind and earthly
wind dynamos have approximately equal control of
the ionosphere–plasmasphere–magnetosphere sys-
tem. During a great magnetic storm the composition
of the thermosphere can be significantly modified,
even worldwide. This creates great negative iono-
spheric storms wherein the ionosphere virtually dis-
appears for a day, even in full sunlight. The
atmosphere changes so much that the Earth acts like
some other planet, one with very little oxygen. The
solar cycle maximum of 2001 has yielded many
exciting scientific discoveries as we continue to
instrument the Earth and its surrounding near-space
regions. We also expect that space weather will
become more relevant to mankind, as discussed next.
Space Weather
There are many aspects of space weather that are
relevant to human habitation, particularly as we
become more dependent on technology and space-
based systems. Some of these aspects involve the
ionosphere and are described here. Others are related
to magnetospheric phenomena, such as the killer
electrons of the radiation belt that create havoc in
satellite systems, and solar proton events that create
severe radiation levels for astronauts building the
International Space Station.
Rapidly changing magnetic fields due to the vast
ionospheric currents flowing in a major auroral event
create electrical voltages at the surface of the Earth,
just as they do in an electrical transformer. Vast power
grids are the perfect detectors of these voltages, which
are unexpected and can thus trigger the unnecessary
shutdown of elements along the grid. During the great
storm of the last solar cycle, the Province of Quebec
went dark for 12 hours because of such a surge. Today
our power grids are even more interconnected, and
predictions of such conditions are becoming both of
more practical importance and of some practical
feasibility. Such predictions are among the first chal-
lenges of the fledgling National Space Weather Pro-
gram in the United States as well as the global
counterpart of this new program.
Some important space weather effects are strictly
due to the Earth’s dynamic atmosphere without help
from the solar wind. The mid-latitude weather dis-
cussed earlier is of this type and is fairly rare. But near
the magnetic equator (where the magnetic field lines
are exactly horizontal), severe convective storms occur
1030 ISENTROPIC ANALYSIS
night after night in some seasons and longitudes. The
ionosphere is so dense at the Equator that such storms
create havoc with communication systems using radio
waves, which must propagate through the ionosphere
from satellite to the ground. In brief, in such storms the
satellite signals ‘twinkle’, just as starlight does passing
through the turbulent lower atmosphere. This creates
deep fades and distortions in satellite signals, which
disrupt communications. The higher the radiowave
frequency, the less does ionospheric turbulence create
problems. However, even at the high frequencies of the
Global Positioning System satellites (41 GHz), iono-
spheric effects can occur.
Figure 10 is a space weather radar map obtained
over the magnetic equator near Lima, Peru. The dark
areas reveal where ionospheric turbulence exists. On
this night severe weather erupted just after sunset and
lasted for several hours, its effects extending to over
1000 km altitude. Radio signals propagating through
this region would be seriously degraded.
Summary
We have learned much about the ionosphere since its
discovery near the beginning of the twentieth century,
when Marconi first skipped radiowaves off it all the
way across the Atlantic. The space age rocket and
satellite probes, along with powerful ground-based
radars, have revealed much about its properties. New
things are still being discovered, but we are also
entering an age of prediction. The US National Space
Weather Program has as a goal the ability to predict the
ISENTROPIC ANALYSIS
J R Holton, University of Washington, Seattle, WA, USA
Copyright 2003 Elsevier Science Ltd. All Rights Reserved.
Introduction
Atmospheric motions are generally three-dimension-
al; air parcels change altitude (and pressure) as they are
advected by the winds. Since vertical velocities asso-
ciated with large-scale motions tend to be small, and
are not directly observed, the computing of the
trajectories of air parcels utilizing conventional mete-
orological analyses on isobaric surfaces is a difficult
task. Isentropic analysis seeks to simplify the
calculation of air parcel trajectories by utilizing the
fact that, outside regions of active precipitation,
large-scale motions in the atmosphere are quasi-
onset of severe space weather, much as the meteorol-
ogy community has come to predict severe storms in
the lower atmosphere. However, the scale of the
interacting system stretches from the Sun to the Earth
and is much more variable than the solar constant
that directly heats the Earth’s lower atmosphere.
In addition, this vast region of space has few observ-
ing stations, and the challenges of prediction are
enormous.
See also
Dynamic Meteorology: Waves. Electricity, Atmos-
pheric: Ions in the Atmosphere. Global Change: Upper
Atmospheric Change. Magnetosphere. Mesosphere:
Polar Summer Mesopause. Radiation (Solar). Satellite
Remote Sensing: GPS Meteorology. Solar Winds.
Thermosphere. Turbulent Diffusion.
Further Reading
Chen FF (1984) Introduction to Plasma Physics and Con-
trolled Fusion. 2nd edn. New York: Plenum Press.
Davies K (1990) Ionospheric Radio. Exeter, UK: Peter
Peregrinus.
Hargreaves JK (1992) The Solar–Terrestrial Environment.
Cambridge: Cambridge University Press.
Kelley MC (1989) The Earth’s Ionosphere. San Diego:
Academic Press.
Rishbeth H and Gariott OK (1969) Introduction to Iono-
spheric Physics, International Geophysical Series, vol.
14. New York: Academic Press.
adiabatic. Thus, changes in the internal energy of
air parcels owing to exchange of heat with the
surroundings are small compared with changes asso-
ciated with adiabatic expansion or compression
caused by pressure changes. Air parcels moving
adiabatically are confined to isentropic surfaces.
Analysis of the motion relative to isentropic surfaces
then reduces the three-dimensional problem of trajec-
tory analysis to a problem of two-dimensional parcel
advection on the isentropic surfaces. For this reason,
isentropic analysis has become a popular technique for
analysis of the transport of long-lived trace constitu-
ents. The method is especially useful in the strato-
sphere where the flow can often be considered to be
approximately adiabatic for time scales of a week or
longer.
 ISENTROPIC ANALYSIS 1031

Isentropic Coordinates of relative vorticity evaluated on an isentropic

surface, and f is the Coriolis parameter. For adiabatic
The first law of thermodynamics for adiabatic condi-
motions P is conserved following the motion on y
tions can be expressed as
surfaces:
D
ðcp ln T
R ln pÞ ¼ 0 ½1
 
q
Dt þ V ry P ¼ 0
. ½4Š
qt
where R, cp , p, and T stand for the gas constant for dry
air, the specific heat at constant pressure, the atmos-
pheric pressure, and the temperature, respectively.
If this expression is integrated from a state at pre- Applications
ssure p and temperature T to a state with pressure In practice, isentropic analysis requires interpolation
ps  1000 hPa, the temperature becomes of meteorological fields from isobaric surfaces, on
which upper-air data is normally reported, to surfaces
y ¼ Tðps =pÞR=cp ½2Š of constant y. A latitude–height section showing the
relationship between isentropic surfaces and isobaric
The quantity y in this expression is referred to as the
surfaces is given in Figure 1. In the troposphere
potential temperature. It represents the temperature
the isotherms slope downward toward the poles,
which an air parcel located anywhere in the atmos-
because temperature decreases with height and with
phere would acquire through adiabatic compression a
latitude. The isentropic surfaces, on the other
pressure of 1000 hPa. Thus, as long as the motion is
hand, slope upward toward the poles in the tropo-
adiabatic, potential temperature is conserved follow-
sphere. This slope is substantial, especially in the
ing the motion, and surfaces of constant potential
winter hemisphere. Note also that for potential
temperature represent isentropic surfaces.
temperatures between about 300 and 350 K, the
Normally, potential temperature increases mono-
isentropic surfaces cross the tropopause, so that air
tonically with height and thus can be used as an
can be exchanged between the low-latitude tropo-
independent vertical coordinate just as pressure is used
sphere and the extratropical lowermost stratosphere
as a vertical coordinate in standard synoptic analysis.
through adiabatic motions. In the stratosphere there is
When the equations of motion are transformed into
a downward slope of the isentropes from the equato-
the isentropic coordinate system (with y as the vertical
rial region to the polar region in the summer hemi-
coordinate) the horizontal pressure gradient force is
sphere and to the midlatitudes in the winter
represented by the horizontal gradient of a quantity
hemisphere. The flattest y-surface tends to occur
called the Montgomery stream function, defined as
near the 350 K level.
C  cp T þ F. Thus,
Isentropic coordinates have occasionally been used
1 in weather forecast models and global climate models.
rp ¼ rC However, partly because of the complexity of the
r

where the gradient of C is evaluated by taking the

partial derivatives with y held constant. 30
In isentropic coordinates an infinitesimal control 560 20
500
Pressure (hPa)

volume of horizontal cross-section dA and vertical

Altitude (km)

100 440 210

200 15
extent dy has a mass dM ¼ r dA dz ¼ s dA dy, where 380 210
s  g 1 qp=qy is the ‘density’ in ðx; y; yÞ space (i.e., it 350
230 10
300 320
is the quantity that when multiplied by dAdy gives 250
500 290 270 5
the mass dM. For adiabatic motions s satisfies the 290
equation 1000
_ 90 _ 60 _ 30 0 30 60 90
qs South Pole Equator North Pole
þ ry . ðsVÞ ¼ 0 ½3Š
qt Latitude

Thus, s is changed locally only if there is a convergence
or divergence of the mass flux in an isentropic layer of
infinitesimal depth.
Motions on isentropic surfaces are governed by
conservation of potential vorticity, defined as
P  ðzy þ f Þs 1 , where zy is the vertical component
Figure 1 Longitudinal average temperature (dashed contours)
and potential temperature (solid contours) for January 1993. The
heavy solid line denotes the tropopause. The shaded area denotes
the lowermost stratosphere, the region in which isentropes span
the tropopause. (Reproduced with permission from Holton JR,
Haynes PH, McIntyre ME et al. (1995) Stratosphere–troposphere
exchange. Reviews of Geophysics 33: 403–439.)
1032 ISENTROPIC ANALYSIS
Figure 2 Synoptic maps of the nitrous oxide mixing ratio on the
465 K potential temperature surface. The left column shows
observations from the Cryogenic Limb Array Etalon Spectrometer
(CLAES) on the Upper Atmosphere Research Satellite and the
right column shows predictions based on trajectories initialized on
14 February 1993. Solid contours on observational plots show
potential vorticity contours. Note that the trajectory calculations
maintain the observed low nitrous oxide mixing ratios in the interior
of the polar vortex as the vortex rotates and becomes elongated.
(Courtesy of Dr Gloria Manney; adapted from Manney GL, Zurek
RW, Lahoz WA, et al. (1995) Lagrangian transport–calculations
using uars data. Pt 1. Passive tracers. Journal of the Atmospheric
Sciences 52: 3049–3068.)
boundary condition where y-surfaces intersect the
ground, the use of isentropic coordinates in numerical
models has been rather limited.
Isentropic analysis is particularly valuable for
studies of trace constituent transport in the strato-
sphere. Radiative heating and cooling are the
only significant diabatic processes in the stratosphere.
The time scale for radiative temperature adjust-
ments is of order 3 weeks in the lower stratosphere
and 1 week in the upper stratosphere. Thus, for
time scales of a week or so motions in the lower
stratosphere can be approximated as adiabatic, and
long-lived trace constituent transport can be approx-
imated by the transport along the isentropes.
An example of isentropic analysis is shown in
Figure 2. The figure compares satellite obser-
vations of nitrous oxide with predictions based on
8-day trajectory calculations. In general the trajectory
model captures the changes in the hemispheric distri-
bution of nitrous oxide during each of the 8-day
periods. If accurate heating and cooling rates are
available then the motion of trace constituents
through isentropic surfaces can be included to improve
the results of such trajectory studies. Isentropic anal-
ysis is likely to remain a preferred methodology for
interpretation of atmospheric trace constituent obser-
vations.
See also
Dynamic Meteorology: Overview. Middle Atmos-
phere: Transport Circulation. Stratosphere–Tropo-
sphere Exchange: Local Processes. Thermodynamics:
Moist (Unsaturated) Air.
Further Reading
Bluestein HB (1993) Synoptic-Dynamic Meteorology in
Midlatitudes. Vol. II. Observations and Theory of
Weather Systems. New York: Oxford University Press.
Holton JR (1992) Introduction to Dynamic Meteorology.
San Diego, CA: Academic Press.
Hoskins BJ, McIntyre ME and Robertson AW (1985) On the
use and significance of isentropic potential vorticity
maps. Quarterly Journal of the Royal Meteorological
Society 111: 877–946.
Salby ML (1996) Fundamentals of Atmospheric Physics.
New York: Academic Press.
Shaw N (1933) Manual of Meteorology. Vol. 3. The Physical
Processes of Weather. Cambridge: Cambridge University
Press.
 ISOTOPES, STABLE 1033

ISOTOPES, STABLE
C A M Brenninkmeijer, Max Planck Institute for
Chemistry, Mainz, Germany Table 1 Minor isotope abundances

Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Element Minor isotopes (%) Main reference materials
2
H H or D: 0.15 V-SMOWa
13
Introduction C C: 1.11 V-PDBb,c, NBS-19c
ðd13 C ¼ 10:95Þ
15
Chemical reactions and phase transitions induce small N N: 0.37 Atmospheric N2
17
but measurable changes in the stable-isotope abun- O O: 0.037 V-SMOW, or
18
O: 0.204 atmospheric O2
dances of the elements in atmospheric gases and 33
S S: 0.76 CDTd
chemical compounds in aerosol. These natural varia- 34
S: 4.22
tions do not affect chemistry or transport but their 36
S: 0.014
37
investigation can deepen our understanding of certain Cl Cl: 24.47 Not available
81
atmospheric processes. The most important underly- Br Br: 49.46 Not available
ing principles are that different sources of trace gases a
Vienna-Standard Mean Ocean Water.
have different isotopic signatures, and that the isotopic b
Original material exhausted.
c
compositions of trace substances are changed by their d
Calcium carbonate.
removal processes. Application of isotopic analysis in Troilite phase of the Canyon Diablo meteorite.
atmospheric chemistry has revealed a rather wide-
spread class of new isotope effects, the causes of which
negligible. Note that atmospheric methane inciden-
are not yet understood. Analytical progress has been
tally has an excess of deuterium owing to releases from
rapid and nowadays very small quantities of subs-
heavy-water nuclear power plants. Its multideuteri-
tances can be analyzed rapidly, thus turning isotope
um-substituted isotopic composition is peculiar and
analysis into an increasingly useful tool in the atmos-
nonrandom and of little practical use.
pheric sciences.
For any two reference materials A and B,
the conversion for a sample X is made according to
Notation eqn [2].

With the exception of hydrogen, variations in stable- 1 þ dB ðXÞ

isotope ratios occurring in nature usually amount to
only several percent. Therefore, the isotope abun-
dances for a particular element are expressed on a
relative scale that is defined by one or more standard
reference materials. For a given sample (SA), the
isotopic value relative to the standard material (ST) is
given in parts per thousand (%); using carbon-13 as an
example with R ¼ 13 C=12 C, eqn [1] gives the relative
isotopic value.
 in Table 1.
RSA

d13 C ¼
1 1000
RST
This definition is based on the ratio of the minor to the
more abundant isotope. It excludes the use of molec-
ular ratios, so that, for instance, d13 CH4 is not defined
in the literature; thus the notation d13 C(CH4) is
unambiguous. For the deuterium content of a sample
of methane the appropriate equation corresponding to
eqn [1] is used. The molecular ratio in the sample,
(CDH3)/(CH4), will be four times larger. Given the low
abundance of deuterium, the fraction of methane
molecules with two or more deuterium atoms is
1 þ dA ðXÞ ¼ ½2
1 þ dB ðAÞ
For mixing a molar fraction f having a delta value da
into a reservoir with db , the resulting isotopic compo-
sition will be given by eqn [3].
dr ¼ f da þ ð1
f Þdb ½3
For convenience, the isotopic abundance of the
elements most used in atmospheric studies are given
Reference materials for isotopic analysis are avail-
½1
able from the International Atomic Energy Agency
(IAEA) in Vienna, and the National Institute for
Standards and Technology (NIST) in Gaithersburg,
MD, USA.
Measurement
Measurement is in most cases based on mass spect-
rometry, by which a precision of up to 0. 01% can be
achieved. This is ample for all atmospheric applica-
tions. The isotopic composition of a sample, at times in
micromole quantities, is repeatedly compared to that
1034 ISOTOPES, STABLE
of a mass spectrometer standard under near-identical
conditions. The analytical principle is electron impact
ionization, magnetic separation and focusing, and
Faraday collectors. Conversion to the appropriate
reference material is basically done using eqn [2], after
ion current corrections have been made. For in-
stance for CO2 , mass 45 represents a combination of
13 16 16
C O O plus 12C17O16O. A recent development
has revolutionized atmospheric applications. Instead
of transferring the sample material under high vacuum
into the mass spectrometer, a flow of helium is used as
carrier gas. This has allowed the coupling of a gas
chromatograph, which can separate atmospheric trace
gases from the air matrix, with the mass spectrometer.
One advantage is that nanomole quantities can be
analyzed. The other is that the speed of analysis is
increased and automation is straightforward. This
on-line analysis already accounts for most of the
atmospheric applications. It is expected that the
number of isotope-based atmospheric studies will
increase rapidly because the analytical effort has
become comparable to the abundance measurement
of, for instance, hydrocarbons.
In a few cases, optical techniques are used. Tunable
diode laser absorption spectroscopy (TDLAS), for
instance, allows comparison of absorption by CDH3
and CH4. The advantage is that no sample preparation
is necessary: the methane simply has to be separated
from air. When using mass spectrometry, the methane
has first to be chemically converted to CO2 and H2. A
TDLAS instrument has been used in aircraft measure-
ment of water vapor isotopes. Occasionally Fourier
transform Infrared (FTIR) spectroscopy is used for
laboratory measurement of CO2 and N2O is-
otopomers. Optical techniques have the major advan-
tage that remote sensing is possible. Infrared
spectroscopy is used for stratospheric water vapor
and ozone isotope measurements.
Kinetic and Equilibrium Isotope
Effects
A distinction is made between kinetic and equilibrium
isotope effects, as for water vapor in contact with the
liquid in an enclosed volume. The depletion of the
vapor in D and 18O can be calculated on the basis of
the vapor pressure differences. When CO2 is in oxygen
isotopic equilibrium with H2O, its 18O content at
251C is 41% above that of the water. This effect can be
calculated like most other (thermodynamic) equilib-
rium effects, using statistical mechanics and the
partition function.
A kinetic isotope effect also occurs during net
evaporation of water. In the absence of turbulent
Table 2 Reaction rate constants at 298 K in 10
15 cm3
molecules
1 s
1
Cl OH
CH4 100 6.4
CDH3 74 5.2
CD2H2 46 3.4
CD3H 23 2.1
CD4 6 0.9
mixing in the air above the liquid, the depletion of the
water vapor is determined by the difference in diffu-
sion speeds of the isotopically substituted molecules
(isotopomers), and can be calculated readily.
Kinetic isotope effects are common in chemical
reactions. The heavier isotopes have stronger bond
strengths, and heavier molecules react more slowly.
Here the calculations are more difficult and transition
state theories with several refinements, including
tunneling effects, are often used. For the atmospher-
ically relevant reaction between chlorine radicals (Cl)
and CH4 , calculations are based on the intermediate
complex Cl–HCH3. The theory and experimental
results do not always conform. The systematics of
isotope substitution for the reactions of CH4 is-
otopomers with OH and Cl are shown in Table 2.
Note that the reaction rate constants are determined
by the large difference between the reactivity of D
compared to H. For Cl the reaction rate constant is
almost entirely determined by the degree of substitu-
tion because the reaction rate for D is much smaller.
For atmospheric applications, the overall reaction
rates are used. Thus, whether, for instance, DCl or HCl
is the product is not of importance.
The fractionation factor k (or a) is defined as the
isotopic ratio in one compound or phase relative to
that of the other. No strict convention is in operation.
Often water vapor is assayed relative to the liquid, and
ko1, (i.e., depletion). The accompanying fraction-
ation constant e is defined as a
1. For kinetic isotope
effects (KIE), the fractionation factor is the ratio
between the reaction rate constants for the minor and
major isotopomers. In most cases a will be positive as
molecules substituted with the heavier isotopes gener-
ally react more slowly. A notable exception is CO, for
which C18O reacts more rapidly with OH than does
C16O, and accordingly ao1. The various isotope
fractionation processes of relevance to atmospheric
processes are shown in Table 3.
Mass-independent Fractionation
Stable-isotope investigations in the atmosphere have
revealed one or more types of fractionation processes
that are not yet understood and have been pooled
 ISOTOPES, STABLE 1035

Table 3 Fractionation processes pheric research has changed this; recently MIF has
been detected for O3 , CO2 , CO, N2O, H2O2 , SO24
,
Process Example
and even O2. A measure for MIF is the excess in 17O
Phase change Evaporation expressed as: D17 O ¼ d17 O
0:52 d18 O.
Diffusion CO2 in leaf stomata
Dissolution CO2 in water
Gravitational settling Ice core studies Atmospheric Applications
Atmospheric escape H2
Chemical reaction CH4 1 OH General
Photolysis N2O
Isotope effects do not affect atmospheric chemistry
under most circumstances. The reasons are that
isotope effects are nearly always small, in the percent
under the name ‘mass-independent fractionation’ range, and the abundances of the minor isotopes are
(MIF). The rationale for this name was that the effect, low. Even though the isotope effects for deuterium
first observed for ozone (O3), was attributed to may be large because of the large relative mass
molecular symmetry effects, totally independent of difference, its low abundance (Table 1) excludes
mass. Retrospectively, the name is an unfortunate effects on chemistry. For chlorine, the less-abundant
choice, but it is often used along with the expression isotope is comparatively rather abundant, but here
‘anomalous fractionation’. Because of its practical and fractionation effects of only a few percent do not make
fundamental importance, MIF is discussed here. much impact on chemistry. Thus, in general, isotope
Figure 1 shows the relationship between d17 O and effects are useful as tracers but have no direct
d18 O for a wide range of substances. The slope of the implications.
mass-dependent fractionation line thus defined is close Given a certain atmospheric trace gas, isotopic
to 0.52. The theoretical basis for this correlation using variations are due to two factors. One is the isotopic
statistical mechanical calculations is well understood. composition of its sources. In principle, the relative
For instance, the equilibration constant for an isotope source strengths can be calculated using eqn [3],
exchange reaction is proportional to the vibrational provided there are only two isotopically distinct and
frequencies, and therefore the reciprocal masses. For well-defined sources. The other factor, which disturbs
diatomic molecules for O isotopes (18O, 17O and 16O) this simple picture, is isotope fractionation in the
one obtains ð1=16
1=18Þ=ð1=16
1=17Þ ¼ 1:89, removal of the gas; however, this does give informa-
the inverse giving 0.53. In fact, because of this tion about the degree of removal. If one sink has a
omnipresent strict mass dependence, almost no atten- distinctly stronger fractionation, its impact can be
tion has been paid to the analysis of 17O in oxygen- detected and, under certain conditions, quantified.
containing compounds for several decades. Atmos- For an isolated, well-mixed amount of gas being
depleted by a loss process (e.g., reaction with OH or
photolysis), the isotopic composition will evolve in
time according to a Rayleigh distillation process.
100
Using m=m0 as the ratio of the actual to the initial
90
O3 mixing ratio, eqns [4a] or [4b] apply.
80
70  ð1
1 =aÞ
60
m
dþ1¼ ½4a
m0

17O (‰)

50
40 Stratospheric CO2
30 
1
  
m
Mass-dependent
20 d ffi 1
l ½4b
10
CO CO2 line a m0
O2
0
− 10 V-SMOW
− 20 Ozone, O3
− 40 − 20 0 20 40 60 80 100 120

18O (‰) Ozone has been analyzed in situ optically and by
balloon-borne mass spectrometry, and via sample
Figure 1 The d18 O and corresponding d17 O values for some
collection. Ozone is formed via Oð3 PÞ þ O2 þ M !
important reservoirs, showing mass-independent fractionation
effects as deviations from the ‘mass-dependent fractionation’ line O3 þ M and has a remarkable isotopic composition
of slope 0.52. Note that molecular oxygen has a small negative throughout the atmosphere. It is considerably
value of
0.15%, not visible on this scale. enriched in 18O relative to molecular oxygen by
1036 ISOTOPES, STABLE
Table 4 Relative reaction rates for the formation of ozone from
16
O(O), 17O(P), and 18O(Q)
Reaction Relative rate
O 1 OO 1 which is dominated by the Northern Hemisphere
P 1 OO 1.03
Q 1 OO 0.93
O 1 PP 1.23
P 1 PP 1.02
Q 1 PP 1.03
O 1 QQ 1.53
P 1 QQ 1.31
Q 1 QQ 1.03
70–110% and by almost the same amount in 17O. In
fact, D17 O ffi 33%. No other substance known on
Earth has this degree of excess of 17O. Via this mass-
independent enrichment (the excess 17O), certain
other trace gases (e.g., stratospheric CO2) exhibit
MIF. The cause for the complex isotopic anomaly of
O3 is only slowly becoming understood. The reaction
rate constants for permutations of the isotopes, and
their temperature and pressure dependence, have been
measured (examples are given in Table 4). Despite the
many experimental data, there is not yet a theoretical
explanation.
The isotopic composition of stratospheric O3 has
also been debated because some extreme enrichments
have been reported, though these have not been
confirmed. Improved measurement of O3 collected
from the stratosphere up to 35 km shows that d18 O is
between 70% and 110%. Laboratory experiments
show that this range of d values corresponds to
temperatures of 270–200 K. Thus, the enrichment of
stratospheric ozone is a measure of temperature.
Similar results are obtained using 17O. Tropospheric
O3 has been collected and analyzed in an urban
environment: the d18 O values ranged between 80%
and 100% and were not correlated with the mixing
ratios between 5 and 90 nmol/mol. The delta values
agree with laboratory experiments.
Carbon Dioxide, CO2
Troposphere
Quantifying fluxes of CO2 between the terrestrial
carbon reservoirs is of great importance, and isotopic
analysis is useful for this. The combustion of fossil
fuels has decreased d13 C of atmospheric CO2 from a
preindustrial value of about
6 % to presently about

7.8% V-PDB. The preindustrial value is determined
by the equilibrium fractionation (
9% at 151C)
between gaseous CO2 and ocean bicarbonate
13
C ffi 2%). The fractionation associated with the
net flux into the terrestrial biosphere, which favours
12
CO over 13CO by about 25%, is compensated by the
return flux. The bicarbonate is buffered by the large
carbonate reservoir. The activity of the biosphere,
biota, induces a seasonal cycle of up to 1% at high
northern latitudes at the surface. The seasonal mini-
mum is in spring when CO2 from respiration, oxida-
tion, and fossil fuel emission have peaked. Fossil fuels,
and plant organic matter have d13 C values that
average approximately
25%. To what degree the
effect of El Niño can be extracted from the two or three
globally accurate isotope time-series is debated, but
dissolution into the oceans depends on temperatures
and wind and causes a small fractionation. NIST has
produced gaseous standards for the required high-
precision isotope analysis of atmospheric CO2. A
useful application of d13 CðCO2 Þ has been the identi-
fication of a large Northern Hemispheric terrestrial
carbon sink in 1992 and 1993.
18
O(CO2) is determined by the exchange with the
main water reservoirs. Isotopic exchange with cloud
droplets is far too slow in comparison to the atmos-
pheric residence time of CO2. No exchange with water
vapor is possible because bicarbonate formation is
required. However, exchange with soil water, leaf
water, and ocean water is important. In this, CO2
strives toward isotopic equilibrium accompanied by a
temperature-dependent enrichment of about 40%.
Accordingly, the d18 O value of CO2 is close to 40%
(V-SMOW, Vienna-Standard Mean Ocean Water).
Note that the standard V-SMOW-CO2 is often used,
thus giving a value of near zero. Because leaf water is
enriched owing to the equilibrium and kinetic isotope
effects accompanying stomatal evaporation, CO2 in
equilibrium with leaf water is more strongly enriched.
The 18O isotope effect of emission of CO2 by fossil fuel
combustion (18O 5 23.5%) has only a small influence
owing to the size of other large exchange fluxes. When
CO2 from two sources is mixed, each with a given
18
O/16O ratio, one can calculate the (low) abundance
of C18O18O. This calculated ratio does not necessarily
correspond with the real C18O18O ratio, because it
would require scrambling of 18O between the CO2
molecules, which does not occur (no gaseous ex-
change). Analytical problems in determining the ratio
of C18O18O to C16O16O have prevented practical
application of this generally overlooked effect.
Stratosphere
The seasonality of d13 C and CO2 mixing ratios
propagates with delay and attenuation into the stra-
tosphere. This allows accurate stratospheric measure-
ments to be used as a clock. In the stratosphere, CO2

acquires MIF from O3 via quenching of O(1D),
according to O3 þ hn ! O2 þ Oð1 DÞ followed by
Oð1 DÞ þ CO2 ! CO3n ! Oð3 PÞ þ CO2 . D17 O in-
creases accordingly from zero to 14%. This interesting
signal is a new gaseous tracer for air entering the
troposphere from the stratosphere, such as O3 and
14
CO.
Water, H2O
Troposphere
The International Atomic Energy Agency coordinates
the global monitoring of the isotopic composition of
precipitation for hydrological applications (isotope
hydrology). There are large seasonal and geographical
differences. Water evaporating from the oceans is
depleted, but the condensing water is again enriched
relative to the vapor. With increasing distance into the
continents, D and 18O decrease (distance effect)
following the mentioned Rayleigh fractionation.
With increasing latitude and decreasing temperatures,
vapor and precipitation cause further depletion. In
Antarctica, dD can be as low as
400%, and d18 O
reaches –50 %. Isotopic analysis of ice cores is one of
the main tools in paleoclimatology. Global precipita-
tion D and 18O isotope values define the ‘meteorolog-
ical water line’ dD ffi 8d18 O þ 10%. Hailstones have
been assayed layer by layer for study of their formation
process. Relatively little attention has been paid to
tropospheric water vapor analysis.
Stratosphere
Because of the many roles of atmospheric water
through latent heat, cloud albedo, heterogeneous
chemistry, and the radiation budget, efforts are made
to use isotopic analysis. In the stratosphere, gas-phase
chemical interactions affect the isotopic composition.
Through cooling and concomitant condensation, wa-
ter vapor reaching the tropopause attains dD ¼

670% and d18 O ¼
82%. During transport into
the middle atmosphere, isotope exchange and addi-
tion of H2O from methane oxidation take place.
Exchange of oxygen is via the HOx family of reactions
involving oxygen and ozone. These processes enrich
the isotopically depleted vapor imported from the
troposphere. Above 40 km, dD reaches about

400%, and stays constant at that level. d18 O
increases by as much as 100%. There is a clear
deviation from the meteoric water line. Furthermore,
stratospheric H2O obtains MIF largely from O3 via
H þ O3 ! OH þ O2 and the subsequent hydrogena-
tion of OH. The oxygen isotope exchange between
OH and H2O has a rate constant of 6  10
17cm3 s
1
ISOTOPES, STABLE 1037
at 250 K and plays no significant role. d17 O increases
from zero at the tropopause to 10% at 40 km.
Methane, CH4
The isotopic compositions of atmospheric methane
and its sources are shown in Figure 2; a large depletion
is observed. The actual atmospheric composition
deviates from the average source composition because
of the isotopic fractionation mainly in the sink
reaction of CH41OH. Also, the smaller soil sink
tends to enrich the methane that is left in the
atmosphere. Note the large difference for deuterium.
Despite the considerable spread in the source values,
inverse modeling using the isotopic composition has
helped to further constrain the methane budget. Both
d13 C and dD show seasonal cycles with a total
amplitude of approximately 0.5% depending on
latitude and can be resolved with difficulty. There are
presently few data on deuterium available, but the
advances in isotope mass spectrometry that have been
mentioned will change this. Interannual variations in
the Southern Hemisphere have been associated with
biomass burning. Moreover, at times small increases in
d13 C have been detected that were attributed to the
reaction of Cl 1 OH; as Table 5 shows, this reaction
has a characteristic large 13C kinetic isotope effect.
The strong increase in atmospheric CH4 induced an
isotopic disequilibrium in which the isotopic compo-
sition came closer to that of the averaged source.
Analysis of CH4 extracted from firn air reveals a clear
minimum dD value about 75 years ago. After a
weakening in the increase of CH4 over the last decade,
the equilibrium was approached more closely and dD
increased again.
Stratospheric chemistry produces large isotope
changes in CH4 because of the greater role of Cl and
O(1D) as sinks. Table 5 shows that O(1D) has a large
isotope effect, which was not expected on the basis of
the rapid kinetics of this reaction. The isotopic
composition of CH4 in the stratosphere can be
successfully modeled using a 2D model, incorporating
the fractionation factors from Table 5. Such results are
of further use for calculating dD of H2O and d13 C of
CO.
Hydrocarbons and Methyl Chloride
The recent introduction of coupled gas chromatogra-
phy–isotope mass spectrometry has allowed the anal-
ysis of nanomole amounts of substances, bringing the
low-abundance trace gases within reach of isotopic
analysis. Values obtained for ethane, ethene, and
propene in background air in New Zealand yielded
1038 ISOTOPES, STABLE

Biomass
Present atmosphere burning

−100

D (‰ V-SMOW)

KIE
Natural gas

−200

Ruminants Average source

−300
Wetlands

Rice

−400
−100 − 80 − 60 −40 −20 0

13C (‰ V-PDB)
13 2
Figure 2 Overview of the C and deuterium (D H) isotope ratios of the main methane sources. The arrow indicates the kinetic isotope
effect.

d13 C values of
22% to
29%, which is typical for
organic matter. Methyl chloride was strongly depleted
at
43.5%. Samples collected south of Japan,
between 51 and 351 N showed the following d13 C
values: ethane, propane, n-butane, n-pentane, i-bu-
tane, and i-pentane between
30 and
23%; ethyl-
ene and propylene between about
30 and
10%;
acetylene between
20 and
2%; methyl chloride
between
40 and
30%.
Carbon Monoxide, CO
The shorter lifetime of CO results in larger concen-
tration and isotopic changes (Figure 3). Concer-
ning source signatures, CO from high-temperature
combustion processes adopts the d18 O value of
atmospheric oxygen (123.5%), without much frac-
tionation. This gives a useful clear signal for the
technological source. Also distinctive is the low d13 C
value of CO derived from CH4 oxidation. This value is
believed to be about
51%, composed of the
47%
of CH4 further lowered by the kinetic fractionation of
4% in CH41OH (Table 5). There are problems in
closing the isotope budget for 13CO in the Southern
Hemisphere using this information, because atmos-
pheric d13 C(CO) values appear rarely to fall below

32%. One has to assume either that additional
isotope fractionation occurs in the formation of CO
from the photolysis of formaldehyde, or that the yield
of CO from CH41OH, is lower than the generally
accepted 85%. For 13C there is a pressure effect. The
KIE maximizes at 5% at 1013 hpa, and turns into a
negative effect at 300 hpa. The annual average d13 C
value in the Southern Hemisphere is approximately

29%; in the mid-latitude Northern Hemisphere it is

27%. The annual cycles are dominated by the CH4
source effect, not the KIE of CO1OH.
For 18O there is a negative isotope effect of nearly

10%, almost independent of pressure. Generally,
the farther away from CO sources, the more negative
the d18 O becomes, reaching
10% in the lowermost
stratosphere. There are only few stratospheric data for
CO, but air samples collected during ozone hole
conditions in the Antarctic lowermost stratosphere
Table 5 Kinetic isotope fractionation factors (KIE) for reaction
with CH4 at 296 K
13
C D
OH 1.004 1.294
Cl 1.065 1.50
O(1D) 1.013 1.06

CH4 oxidation

18O (‰ V-SMOW)
KIE
−60 − 50 − 40 − 30
Isoprene/terpenes

13C
13 18
Figure 3 Overview of the C and O isotope ratios of the main carbon monoxide sources. The arrow indicates the kinetic isotope effect.
Observational values for some Northern Hemisphere locations are shown. The spread in the data is mostly due to the seasonal cycle.
(NMHC, non-methane hydrocarbons.)
yielded CO with d13 C as low as
43%. Modeling
confirmed that CO from CH41Cl was the cause of
this. Another application of the large KIE in CH41Cl
is the estimation of free chlorine during Arctic low-
ozone events at the surface. Small downward excur-
sions in d13 C(CO) were observed during such events.
After MIF had been detected for O3 and CO2 , it also
was discovered in CO. MIF in CO has two causes. One
is that unsaturated organic compounds react with O3.
CO from these reactions is a small source of CO
enriched in 18O and exhibiting MIF derived from O3.
However, the main cause lies in the important reaction
CO1OH-CO21H. Assuming for the moment that
C18O reacts 10% faster than C16O, C17O should react
about 5% faster if the normal widespread mass-
dependent fractionation applies. In contrast, C17O
reacts nearly as rapidly as C16O. This then results in an
excess of 17O in atmospheric CO. There is not yet a
theoretical explanation for MIF in CO. Because all
major CO sources are mass-dependent, d17 O(CO) is a
unique signal indicating the ‘aging’ of CO by reaction
with OH.
Nitrous Oxide, N2O
Despite its long lifetime and concomitant small
isotopic and abundance variability, the isotopic com-
ISOTOPES, STABLE 1039
30
Technological
Savanna 20
burning
Forest
burning
10
0
Ocean
− 20 −10 0 10
Alert
− 10
Spitsbergen
Izaña
− 20 Baring head
Anthropogenic
Scott base
NMHCs
− 30
(‰ V-PDB)
position of N2O has been studied intensively. The
d15 N and d18 O values of tropospheric N2O show little
scatter around approximately 7% and 21% relative to
atmospheric N2 and O2 , respectively. The main
sources of N2O are based on microbial nitrification
and denitrification in soils and the ocean, and are
generally depleted relative to the atmosphere. The
range of d values is considerable. Soil gases form the
most depleted source type, with averages of about

13% and 10%, respectively. A source identified
with enrichment was N2O from denitrification in
upwelling deep water in the Arabian sea. Enriched
N2O was also found in the Pacific. These sources do
not account for the atmospheric enrichment. The
enrichment of atmospheric N2O has been explained
on the basis of its main stratospheric sink, i.e.,
photolysis. N2O that escapes photolysis and is reim-
ported into the troposphere causes enrichment relative
to the average sources. Theory and experiment of
photolysis show a qualitative agreement. Photolysis
experiments over the entire range of wavelength of
interest still has to be performed. Analysis of strato-
spheric N2O samples confirms that enrichment in-
creases with altitude.
Increased interest in N2O was sparked by the
awareness that in this linear molecule, NNO, frac-
tionation through photolysis is not identical for the
two different N atoms. Instrumental innovation in
1040 ISOTOPES, STABLE
Remaining N2O fraction (%)
10 20 40 60
0.10
0.08
ln (
15N/1000+1)
0.06 14N15NO
0.04 Average
(15N14NO + 14N15NO)
0.02
15N14NO
0.00
− 2.5 −2.0 −1.5 − 1.0 − 0.5
ln (remaining N2O fraction)
Figure 4 A Rayleigh-type of plot for the change in the 15N isotopic
composition of N2O subjected to photolysis at 193 nm.The enrich-
ment for 14N15NO is considerably larger than for 15N14NO.
mass spectrometry now allows the detection of the
isotopic ratios for both positions using only small
amounts of sample. Figure 4 shows the result of
N2O photolysis at 193 nm for 15N. The fraction-
ation factors for 15N14NO and 14N15NO are
10.9% and 35.7% respectively. Thus 14N15NO is
less likely to be photolyzed. The same applies for
14 14 18
N N O, for which at this wavelength the frac-
tionation factor is 17.3%. For the other important
stratospheric reaction, O(1D)1N2O, only few data
are available.
Possible gas-phase sources of N2O have been
proposed in the literature, and the occurrence of
MIF, albeit at a low level ðd17 O ffi 1%Þ, has intensified
this interest. No confirmation has been provided, and
the cause of MIF in N2O remains unknown.
Molecular Oxygen, O2
In view of the sheer size of this reservoir, no usable
isotope effects were originally contemplated to occur.
Notwithstanding, it has been established that even
atmospheric oxygen itself possesses a small degree of
MIF, d17 O ¼
0:15%, which offers interesting appli-
cations. The 18O isotopic enrichment of O2 relative to
ocean water (V-SMOW) is 23.5%. The cause for this
enrichment is well understood. Respiration processes
in plants and soils favor the use of the lighter of the
isotopic O2 molecules. In addition to this is that the
leaf water in plants, which forms the substrate for the
oxygen appearing in photosynthetic products, is
enriched. These processes keep atmospheric oxygen
enriched relative to the very large reservoir of terres-
trial water. Because of this exchange between the
large terrestrial water reservoir and atmospheric
O2 , its isotopic composition should be strictly mass-
80 100 dependent.
However, a very small fraction of O2 is continuously
converted to O3 , which exhibits MIF. Thus a small
excess of 17O is withdrawn. In the stratosphere, this
MIF signal is transferred from the O3 pool via O(1D) to
become incorporated into CO2. The ensuing step
in the process is that CO2 reenters the troposphere,
where it isotopically exchanges with H2O. In this
way, CO2 loses its MIF signature continuously to the
large reservoir of H2O. Another process transferring
MIF is the reaction O(1D)1H2O, which also leads
0.0
to a net loss of 17O. Both processes remove a small
excess of 17O from the O2 reservoir. Given the long
lifetime of atmospheric oxygen of roughly 1000 years,
it acquires a deficit of 17O, resulting in the value
d17 O ¼
0:15%. Although this value is small, there
are applications in limnology and oceanography in
which atmospheric oxygen can be distinguished from
photosynthetic oxygen.
Aerosols
The oxidation of SO2 by O3 or H2O2 in the liquid
phase leads to a small degree of MIF in atmospheric
sulfates. This can possibly be applied to trace the
oxidation pathway of SO2. There is one report of
H2O2 possessing MIF, for which the reason is not
known. Sulfur isotope measurements on marine sul-
fate aerosol particles support the hypothesis that
dimethyl sulfide can be a source of non-sea-salt sulfate.
Some research has also been dedicated to chlorine
isotopes in aerosols.
See also
Aerosols: Physics and Chemistry of Aerosols. Carbon
Dioxide. Evolution of Atmospheric Oxygen. Evolu-
tion of Earth’s Atmosphere. Global Change: Ozone
Trends. Land–Atmosphere Interactions: Trace Gas
Exchange. Methane. Planetary Atmospheres: Jupiter
and the Outer Planets; Mars; Venus. Stratospheric
Ozone Recovery. Stratospheric Water Vapor. Trop-
ospheric Chemistry and Composition: Carbon Mon-
oxide.
Further Reading
Gonfiantini R, Stichler W and Rozanski K (1995) Standards
and intercomparison materials distributed by the Inter-
national Atomic Energy Agency for stable isotope meas-
urements. In: Reference and Intercomparison Materials

for Stable Isotopes of Light Elements, IAEA-Techdoc-
825, pp. 13–29. Vienna: IAEA.
Kaye JA (1987) Mechanisms and observations for
isotope fractionation of molecular species in
planetary atmospheres. Reviews of Geophysics 25:
1609–1658.
Kaye JA (ed.) (1992) Isotope Effects in Gas-Phase Chemis-
try. Proceedings of a symposium by the Division of
Physical Chemistry, 201st National Meeting of the
American Chemical Society, Atlanta, Georgia. ACS
symposium Series 502. Washington DC: American
Chemical Society.
ISOTOPES, STABLE 1041
Thiemens MH (1999) Mass-independent isotope effects in
planetary atmospheres and the early solar system. Science
283: 341–345.
Richet P, Bottinga Y and Javoy M (1977) A review of
hydrogen, carbon, nitrogen, oxygen, sulphur, and chlo-
rine stable isotope fractionation among gaseous mole-
cules. Annual Review of Earth and Planetary Science 5:
65–110.
Röckmann T, Brenninkmeijer CAM, Saueressig G, et al.
(1998) Mass-independent oxygen isotope fractionation
in atmospheric CO as result of the reaction CO1OH.
Science 281: 544–546.