Notes

Notes on Atmospheric Physics
Arnaud Czaja1
Physics Department &

Grantham Institute for Climate Change,
Imperial College, London.
May 5, 2017
1 Contact details: Huxley Building, Room 726; email: a.czaja@imperial.ac.uk.

Office Hours: Thursdays, 11.30-12.30; Fridays, 1-2pm.
2
Contents
1 An overview of the atmosphere 1

1.1 Atmospheric composition . . . . . . . . . . . . . . . . . . . . . 2
1.2 Mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2.1 Pressure as a measure of mass . . . . . . . . . . . . . . 4
1.2.2 Measures of water in air . . . . . . . . . . . . . . . . . 5
1.3 Main features of the atmosphere . . . . . . . . . . . . . . . . . 8
1.4 What drives atmospheric winds, weather patterns, etc? . . . . 9
1.5 Further reading . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.6 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2 Radiative heating and cooling of the atmosphere 21

2.1 Concepts and definitions . . . . . . . . . . . . . . . . . . . . . 21
2.1.1 Intensity of radiation . . . . . . . . . . . . . . . . . . . 21
2.1.2 Blackbody radiation . . . . . . . . . . . . . . . . . . . 23
2.1.3 Shortwave and longwave radiation . . . . . . . . . . . . 24
2.1.4 Irradiance . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.1.5 Kirchoff’s law . . . . . . . . . . . . . . . . . . . . . . . 26
2.1.6 The “solar constant” . . . . . . . . . . . . . . . . . . . 27
2.1.7 Radiation balance and emission temperature . . . . . . 28
2.2 A simple model of the greenhouse effect . . . . . . . . . . . . . 30
2.3 Beer’s law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.4 Schwarzchild’s equation . . . . . . . . . . . . . . . . . . . . . . 33
2.4.1 Derivation . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.4.2 Physical interpretation . . . . . . . . . . . . . . . . . . 34
2.4.3 Final form . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.5 Some applications of Schwarzchild’s equation . . . . . . . . . . 35
2.5.1 No scattering . . . . . . . . . . . . . . . . . . . . . . . 35
2.5.2 Infrared radiation by an isothermal atmosphere . . . . 36
2.5.3 Remote sensing of temperature . . . . . . . . . . . . . 36
2.6 Radiative heating and cooling rates . . . . . . . . . . . . . . . 38
2.6.1 Shortwave heating . . . . . . . . . . . . . . . . . . . . 39
i
ii CONTENTS
2.6.2 Longwave cooling . . . . . . . . . . . . . . . . . . . . . 41

2.6.3 Net radiative heating rates . . . . . . . . . . . . . . . . 49
2.7 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3 Radiative-convective equilibrium 55
3.1 Radiative equilibrium . . . . . . . . . . . . . . . . . . . . . . . 55
3.2 Convection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.2.1 The “parcel’s equation” . . . . . . . . . . . . . . . . . 57
3.2.2 Potential temperature . . . . . . . . . . . . . . . . . . 59
3.2.3 Stability of temperature profiles to vertical displace-
ments of air parcels . . . . . . . . . . . . . . . . . . . . 60
3.3 Radiative-convective equilibrium . . . . . . . . . . . . . . . . . 63
3.3.1 Stability of the radiative equilibrium temperature profile 63
3.3.2 The Tropopause . . . . . . . . . . . . . . . . . . . . . . 63
3.4 Dynamical effects of moisture? . . . . . . . . . . . . . . . . . . 64
3.4.1 Dry and moist adiabatic lapse-rates . . . . . . . . . . . 64
3.4.2 Application of “dry and moist adiabats thinking” . . . 67
3.5 Radiative-convective equilibrium and the real world . . . . . . 69
3.6 Sloping convection . . . . . . . . . . . . . . . . . . . . . . . . 71
3.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
3.8 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
4 Atmospheric motions 81
4.1 Equations of motions . . . . . . . . . . . . . . . . . . . . . . . 81
4.1.1 Forces acting on a parcel of air . . . . . . . . . . . . . 82
4.1.2 Material derivative . . . . . . . . . . . . . . . . . . . . 82
4.1.3 Rotating frame of reference . . . . . . . . . . . . . . . 84
4.1.4 Coriolis and centrifugal forces . . . . . . . . . . . . . . 86
4.1.5 Mass conservation . . . . . . . . . . . . . . . . . . . . . 88
4.2 Scale analysis of the equation of motions . . . . . . . . . . . . 89
4.2.1 Vertical momentum equation . . . . . . . . . . . . . . 89
4.2.2 Horizontal momentum equation . . . . . . . . . . . . . 91
4.2.3 The thermal wind relation . . . . . . . . . . . . . . . . 92
4.3 The vorticity view . . . . . . . . . . . . . . . . . . . . . . . . 95
4.3.1 The geostrophic flow, vorticity and divergence . . . . . 95
4.3.2 Predicting the vorticity of the flow: the vorticity equation 97
4.3.3 Rossby waves . . . . . . . . . . . . . . . . . . . . . . . 100
4.4 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
4.5 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
CONTENTS iii
5 Climate change 111

5.1 Anthropogenic radiative forcing . . . . . . . . . . . . . . . . . 112
5.2 Response of the atmosphere to a sudden doubling of CO2 . . . 112
5.2.1 Fast response: atmospheric processes only . . . . . . . 114
5.2.2 Slow response: the atmosphere interacts with the up-
per ocean . . . . . . . . . . . . . . . . . . . . . . . . . 118
5.2.3 Very slow response: the atmosphere interacts with the
ocean circulation . . . . . . . . . . . . . . . . . . . . . 121
5.3 Climate change in realistic models and observations . . . . . . 123
5.4 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
5.5 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
6 Appendices 129
6.1 Radiative transfer . . . . . . . . . . . . . . . . . . . . . . . . . 129
6.1.1 Radiation pencils vs. photon showers . . . . . . . . . . 129
6.1.2 Conservation of intensity . . . . . . . . . . . . . . . . . 131
6.1.3 Application to blackbody radiation . . . . . . . . . . . 131
6.2 Thermodynamics of moist air . . . . . . . . . . . . . . . . . . 131
6.2.1 Entropy of cloudy air . . . . . . . . . . . . . . . . . . . 131
6.2.2 A general formula for the Brunt-Vaisala frequency? . . 132
6.3 Dynamics of rotating fluids . . . . . . . . . . . . . . . . . . . . 133
6.3.1 Formula for D/Dt in a change of frame of reference . . 133
6.3.2 Kelvin’s identity . . . . . . . . . . . . . . . . . . . . . 134
6.3.3 The vorticity equation? . . . . . . . . . . . . . . . . . . 135
iv CONTENTS
Welcome to Atmospheric Physics!

I used to live in the US and I remember watching a Brian Green’s episode
of “The Elegant Universe” about quantum mechanics. He entered the “Quan-
tum café”, asked for an orange juice and the waiter answered: “yes, maybe”.
The more I work on the physics of the oceans and the atmosphere, the
more I realise that climate is like the Quantum café. Take the winds. You
think it’s straightforward. Well, winds on Earth are governed by the Taylor-
Proudman theorem: “there can be no variations of the steady state winds in
the direction of the axis of rotation”. A lab demo of this is just spectacular.
And this is the world we live in, the atmosphere whose composition we
are altering, the climate whose heat balance we are perturbing. Within the
current state of understanding, we can only make a few scientific predictions:
“The radiative forcing due to human activities is large”, “Sea level will keep
rising” are two examples. Not very satisfactory isn’it? To say more with
confidence, we need to use our senses –build new senses (instruments)– to
collect observations of the climate as it is now with future generations in
mind; and, for our own pleasure and maybe a chance to see the future more
clearly, to use our intelligence to unravel how the atmosphere, the oceans,
the cryosphere and the biosphere work and interact. Please join in.
CONTENTS v
Practical things
These notes contain only the basic information discussed in the lectures
(the latter are where emphasis is on the physical interpretation and schemat-
ics). My aim in writing them is to provide you with a support and a clear
knowledge of what is examinable (=what is in the notes). The formal “Aims
and objectives” for the course follows next page. In terms of reading:
• I recommend the excellent textbook by Wallace and Hobbs (‘’Atmo-

spheric Sciences: an introductory survey”) as a companion for the
course (the library has many copies).
• You might also enjoy reading “Clouds in a glass of beer: simple ex-
periments in atmospheric physics” by Craig Bohren (cheap paperback
Dover edition), as well as the textbook by John Marshall and Alan
Plumb entitled “Atmosphere, Ocean and Climate Dynamics: An in-
troductory text” and the older but concise and clear “The Physics of
Atmospheres” by John Houghton. I have also included further refer-
ences in some chapters.
Each chapter contains a set of problems, whose solutions will be provided

as we go along. Some sections in the notes are highlighted with a ? which
indicates that they are a little more challenging.
There are a couple websites which I would like to emphasize:
• https://earth.nullschool.net This is a wonderful website depicting

the state of the atmosphere and the (surface) ocean in nearly real time.
The data comes from a global operational forecast system from the US
(which is constrained by many observations) and satellite observations
for the ocean. The graphics are stunning and there is so much to learn
and wonder spending time on it (believe me it beats Youtube).
• http://www.ecmwf.int/s/ERA-40 Atlas/docs/ An excellent source

of quick plots for the mean atmospheric state. This climatology has
been developed at a big European centre for weather forecast and cli-
mate in Reading (ECMWF).
vi CONTENTS
• http://www.sp.ph.ic.ac.uk/aczaja/EP ClimateModel.html This is

a simple climate model which I will at times use during the lectures. It
is well documented and cheap to run (either in Matlab or in Python,
thanks to the hard work of an undergraduate student Joe Marsh Ross-
ney).
Please do not hesitate to come to Office Hours (Thursdays, 11.30-12.30;

Fridays, 1-2pm) for further help or to give me feedback on the course. You
are also welcome to make any suggestions by email at a.czaja@imperial.ac.uk.
CONTENTS vii
Aims and Objectives for the Course
Aims To provide students with an understanding of the physics behind the

structure, the dynamics and the energetics (radiative transfer, thermodynam-
ics) of the Earth’s atmosphere (emphasis on troposphere and stratosphere).
Objectives By attending the course, the students should:
• be able to describe the basic structure of the Earth’s atmosphere and
the climate system
• be able to use fundamental thermodynamics to derive expressions for
the variation of temperature, pressure, and density with height
• understand the concept of potential temperature and how it relates to
stability, buoyancy frequency and temperature lapse rate
• understand the concept of radiative-convective equilibrium
• know the components of the Earth radiation balance
• understand the concepts of optical depth, radiation intensity, irradi-
ance, and transmission of radiation
• be familiar with Schwarzschild’s equation of radiative transfer and be
able to solve it for both solar and thermal radiation streams under
simple conditions
• be able to derive a simple model of the greenhouse effect
• be able to compute radiative heating rates given irradiances
• know the forces acting on a parcel of air and apply Newton’s 2nd Law
to deduce the equations of motion for a compressible gas on a rotating
planet
• know how to apply scale approximations to the equations of motion
(e.g., hydrostatic and geostrophic approximations, Rossby number)
• understand why vorticity is a useful concept for the study of atmo-
spheric motions
• understand the effect of water on the radiative, thermodynamic and
dynamical aspects of Atmospheric Physics
• understand the concept of radiative anthropogenic forcing and the basic
response of the atmosphere to this forcing
viii CONTENTS
Chapter 1
An overview of the atmosphere
key concepts: well mixed gases, “dry” and “moist” air, measures of water
vapour in air, top-of-the-atmosphere (TOA), global budgets of mass, heat
and angular momentum.
Before we dive into quantitative analysis of the atmosphere it is important

not to loose sight of some of the big questions. Here are a few which I
mentioned in the introduction lecture (ppt slides on Blackboard):
• The atmosphere is our common environment. It is the fluid we all

breath. We say that we are connected with the internet but we are
actually physically connected because we constantly share and recycle
air molecules through our lungs. When the surface winds come from
the South, I am breathing air which was a day or two before breathed
in and out by someone in Spain or maybe Africa. This is because
typical north-south velocities in a weather system are on the order of
10ms−1 so an air parcel covers approximatively 10◦ of latitude in one
day (≈ 105 s).
• Randomness. The tropical Pacific ocean is one area of the globe with
the most observations (ocean and atmosphere) because it is the site
of a major reorganisation of the wind, temperature and precipitation
patterns known as El Nino. Every few years, heat builds up in the
western Pacific ocean and is suddenly released eastward and poleward,
shifting entirely the tropical precipitation pattern. The atmosphere
is “rung” by this shift and generates waves propagating towards the
Northern and Southern Hemispheres, perturbing the weather systems
there. El Nino is also the climate phenomenon with the best theo-
ries. Or so we thought. In the summer of 2014, all El Nino experts
1
2 CHAPTER 1. AN OVERVIEW OF THE ATMOSPHERE
predicted that the largest event ever recorded will develop during the
following winter. It simply did not happen! No El Nino event at all
(see McPhaden, 2015). Imagine telling your friends you’re an expert at
something, predicting the largest anomaly ever seen...and things go on
perfectly normally. This shows that there is a lot more to understand.
Maybe, fundamentally, deterministic predictions of the coupled atmo-
sphere - ocean system are impossible. Maybe this system is a bit like a
quantum mechanics sytem, with only probablistic statements possible.
• Observations. In fifty or a hundred years, we will still need to check that

our numerical models of the climate are accurate. The ones we have and
use now have been only tested over a short period of time (true global
observations of the atmosphere only started with the satellites launched
in the 1970s –in the ocean this type of coverage simply does not exist
below about a 1000m) and one should not be overconfident regarding
their accuracy (see point above). People in fifty or a hundred years will
not be able to travel back in time and make these observations. It is
our duty to do so. Even if like me, you are not someone developing
instruments, you can help those who do by finding the most useful
quantity to observe. And by using the data in your own way, you help
maintain the observational network.
1.1 Atmospheric composition

The most abundant substance in the atmosphere is diatomic nitrogen (N2 ),
which accounts for 78% of the air molecules we breath. Most of the nitrogen
on Earth is actually stored in the atmosphere (3.9 × 1018 kg), with the Lito-
sphere (Earth’s crust) coming second (≈ 2 × 1018 kg) The large atmospheric
reservoir of nitrogen reflects the outgassing from the Earth’s interior in the
earliest stage of its history and the great stability of the N2 molecule.
Next in abundance comes “free” oxygen (O2 ), which represents 21% of
atmospheric molecules. The Earth is unique in having so much of its atmo-
sphere made up of diatomic oxygen, and there is little doubt that this reflects
the presence of life early in its history (it is believed that oxygen started to
accumulate in the atmosphere about 2Gyr ago, when the production of O2
by bacteria exceeded the consumption of O2 by iron ions dissolved in the
oceans).
The percentages given above assume that any given sample of air has the
same composition. In practice, this is only true for gases whose residence
time in the atmosphere is long compared to the time it takes for atmospheric
1.1. ATMOSPHERIC COMPOSITION 3
motions to mix (from a few days to a few months). This is for example the
case for N2 , O2 as well as argon (Ar, ≈ 0.9% of air molecules) and carbon
dioxide (CO2 , ≈ 0.04% of air molecules), the next two most abundant species
after N2 and O2 . Water vapour has a highly variable distribution (with
concentrations which can be greater than that of Ar locally) depending on
time and location because it can be quickly removed from the atmosphere
through rainfall. The height at which mixing by motions is not vigorous
enough to maintain a uniform composition is about 100km (the turbopause).
For the troposphere (the lowest layer of atmosphere where temperature
decreases with height, roughly from the Earth’s surface to a height z =
10km) and stratosphere (the layer above the troposphere where temperature
increases with height, from about z = 10km to z = 50km), which will be
the focus of the course, it is convenient to simplify atmospheric composition
by considering “dry air”, a mixture of N2 , O2 , Ar , CO2 and other trace
gases, and “moist air” (water vapour). The primary reason for this is phase
change: as we’ll see in Section 1.5, there is a net heat gain by the atmosphere
through the hydrological cycle (latent heat) whereas this does not occur for
other species (N2 , O2 , etc, although they are also exchanged between the
atmosphere and the Earth’s surface). It is thus important to keep track
of local concentrations of water vapour. A useful measure of the “distance
to equilibrium of phases” is given by relative humidity (RH), the ratio of
the vapour pressure e of a sample to the vapour pressure in thermodynamic
equilibrium of phases (eeq (T ), a sole function of temperature T from the
Thermodynamic year 2 course):
e
RH ≡ (1.1)
eeq
NB: This is just a definition. In thermodynamic equilibrium RH = 1 but
this condition is rarely met in the atmosphere (for example, it is in the core
of deep clouds but not in the accompanying donwdrafts which are too dry for
e to match eeq (T ) and thus are air masses with RH < 1).
At a given temperature T and volume V , the pressure of “dry air” Pd
obeys the ideal gas law to an excellent approximation,
Pd V = Nd kB T (1.2)
as does water vapour,

eV = Nv kB T (1.3)
In these two equations, kB is Boltzmann’s constant while N denotes the num-
ber of molecules (the subscripts d and v will be used throughout the course
for dry air and water vapour, respectively). Note that the total pressure P
of a given sample of air is simply the sum of Pd (the partial pressure of dry
air) and e (the partial pressure of water vapour), as result known as Dalton’s
law,
P = Pd + e (Dalton’s law) (1.4)
Atmospheric pressures are usually expressed in hP a where 1hP a = 100P a
(you might also find pressures expressed as millibar (1mb = 10−3 bar), in
which 1bar = 105 P a).
Because of the very large number of molecules in the atmosphere, it is
convenient to rewrite the ideal gas law as,
Pd V kB
= T (1.5)
Nd µd µd
P P
in which µd is the mass of a “dry air molecule” (µd = Ni µi / Ni where µi
is the mass of molecule i of which there are Ni in the sample considered –the
sum is carried over i = N2 , O2 , Ar , CO2 , etc). Introducing the specific volume
of dry air αd , and the gas constant for dry air Rd = kB /µd = 287J kg −1 K −1 ,
this becomes,
Pd αd = Rd T (1.6)
Likewise, for water vapour,
eαv = Rv T (1.7)
with Rv = kB /µH2 O = 461J kg −1 K −1 .
1.2 Mass
1.2.1 Pressure as a measure of mass
In layers of air of large horizontal extent, and in particular for the global
horizontal average, there is an approximate balance between gravity and the
vertical pressure gradient force,
∂P
ρg = − (hydrostatic equation) (1.8)
∂z
Note the minus sign, which expresses that pressure must decrease with height
to be able to oppose the downward acceleration due to gravity.
One interesting use of this equation is to integrate it in the vertical as,
Z +∞
P (z) = ρgdz (1.9)
z
1.2. MASS 5
in which we have used the fact that pressure vanishes at sufficiently large
heights. The discussion in section 1 showed that this is a good approximation
for z 8km and this loosely defines the “top-of-the-atmosphere” (TOA
throughout the course). The corresponding layer of air is still very thin
compared to the Earth radius so that one can approximate g in the integral
by its surface value g = 9.81ms−2 ,
Z +∞
P (z)/g = ρdz (1.10)
z
This shows that atmospheric pressure can be thought of as a mass mea-

surement since ρdz is simply the mass per unit area sandwiched between
heights z and z + dz. A couple of straightforward applications of this equa-
tion are worth mentioning. An order of magnitude for the surface pressure
is Ps = 1000hP a while for the tropopause it is 100hP a. This shows that the
troposphere contains about ' (1000 − 100)/1000 = 90 % of the mass of the
atmosphere. Conversely, since the pressure P in (1.10) is the total pressure
(P = Pd + e), and that e at the Earth’s surface is typically 10hP a in the
global and annual mean, water vapour contributes to ' 10/1000 = 1 % of
atmospheric mass.
Technical sidenote: units of pressure. Pressure is usually expressed in hP a =

100P a in atmospheric sciences. You might also find the use of millibars (mb,
1mb = 10−3 bar where 1bar = 105 P a).
1.2.2 Measures of water in air

A given sample of air is described, besides its temperature and pressure, by
its mass of dry air md (see previous section), water vapour mv , liquid water
ml and ice water mi . It is common practice to introduce ratios of these
quantities:
mv
qv ≡ (specific humidity) (1.11)
md + mv + ml + mi
ml
ql ≡ (specific liquid water content) (1.12)
md + mv + ml + mi
mv
qi ≡ (specific ice water content) (1.13)
md + mv + ml + mi
md
qd ≡ (specific mass of dry air) (1.14)
md + mv + ml + mi
In terms of size, because there is so little amount of water in the air, qd qv ,

and typically qv (ql , qi ). Note that qd = 1 − (qv + ql + qi ). The total specific
amount of water is denoted by qt ≡ qv + ql + qi .
Sometimes, mass mixing ratio, rather than specific humidity is used. The
difference is that, for example for water vapour (mixing ratio rv ), mixing ratio
involves taking the ratio of mv to md rather than mv to mv + md + ml + mi ,
i.e.,
mv
rv ≡ (mass mixing ratio) (1.15)
md
The air density ρ is defined according to,
md + mv + ml + mi
ρ= (1.16)
V
in which V is the total volume occupied by the sample (the sum of the
volumes occupied by the gas, liquid and solid phases). As a result, the
density of dry air ρd = md /V = (m/V )(md /m) = ρqd , and, likewise, the
density of water is ρt = (mv + ml + mi )/V = (m/V )(mv + ml + mi )/m = ρqt .
The total amount of water vapour in an atmospheric column, or total
precipitable water (TPW), is
Z ∞
TPW = ρqv dz (1.17)
0
To get a feel for the surprising result to come below, let’s use the simple
model ρ = ρs e−z/Hs and qv = qs e−z/Hq in which Hs is the scale height, Hq
a scale height for moisture (Fig. 1.1) and ρs , qs refer to surface density and
specific humidity, respectively. One can then estimate that,
Hs Hq
T P W ≈ ρs qs H with H = (1.18)
Hs + Hq
Expressed in mm of water per unit area by dividing this quantity by the

density of water, we find typically that the atmosphere holds something
like 20mm of precipitable water in vapour form (for ρs = 1.2kgm−3 , qs =
10g/kg, Hs = 7km, Hq = 3km). Observed values (Fig. 1.2) are indeed
within that range. This is somewhat surprising since it can rain much more
than that in a matter of a few hours but it can be rationalized by thinking
of storms (the cyclones at our latitudes and even more so hurricanes) as very
efficient machines collecting water vapour over very large distances, condens-
ing it, and “dumping” it as rain: this is the “dehumidifier view” of cyclones.
1.2. MASS 7
Figure 1.1: Annual mean specific humidity (qv , in g/kg) as a function of

latitude and pressure (here expressed in dbar, 1bar = 105 P a). Note the
sharp decrease with decreasing pressure so that the scale height for moisture
is much less than it is for pressure. This is because of condensation and rain
which removes moisture from the atmosphere.
Figure 1.2: Mean total precipitable water (in mm) averaged over the oceans
for November 2013. This map was produced using passive microwave mea-
surement from satellites.
1.3 Main features of the atmosphere

Based on the ppt slides for this chapter (to which you are referred to for
illustrations), the main features of the atmosphere are:
• a well mixed structure up to ≈ 100km in terms of constituents, with a
near exponential decay of pressure and number densities with height.
The associated scale is on the order of 8km for the well mixed layer1 .
• a rich temperature structure, with, in some regions, temperature de-

creasing with height and poleward, but in some regions temperature
increasing upward and poleward (Fig. 1.3, top panel). The simplest
view (global average as a function of height) is schematized in Fig. 1.4,
introducing the troposphere, the stratosphere and the mesosphere. The
course will focus on the first two of these where 99.9 % of the mass of
the atmosphere resides.
• the presence of strong zonal (=along a latitude circle) time mean jets
with windspeeds in excess of 30m/s. These are mostly found going
from west to east (e.g., the tropospheric Jet Stream) but also exists
seasonally from east to west (mesospheric jets) –see Fig. 1.3, bottom
panel. At the Earth’s surface, westerlies are found poleward of 30◦ of
latitude, and easterlies (“Trade winds”) are found equatorward of that
latitude. The atmosphere is in a state of “superrotation”, an air parcel
in the tropospheric Jet Stream coming back to its initial position in
about 23h, not 24h! We’ll prove in Chapter 4 that these jets are in
“thermal wind balance”, meaning that their variations with height are
constrained by the horizontal temperature gradients.
• the presence of smaller time mean velocities in the North-South direc-

tion (a few ms−1 ). These are predominantly seen in the Hadley cell
at low latitudes, with rising motions near the equator and descending
motion along ≈ 30◦ . Such “meridional cells” (in the latitude-height
plane) also exist in the stratosphere and mesosphere but the associated
mass transport is much weaker than that of the Hadley cell.
• its convective nature (Fig. 1.5) on scales ranging from a few km to

thousands of km (planetary scale). Updraft motions are associated
1
From Boltzmann’s principle we would expect the ratio of distribution of a molecule
of mass m at height z1 and z2 to obey n1 /n2 = emg(z1 −z2 )/kB T where g is gravity and T
temperature. This provides a different scale height for each molecule according their mass
(kB T /mg), which is not observed below 100km (the “turbopause”). It is observed above
100km.
1.4. WHAT DRIVES ATMOSPHERIC WINDS, WEATHER PATTERNS, ETC?9
with phase change and the formation of rain, snow and other hydrom-
eteors. The convection involves mostly upward/downward motions in
the Tropics, but sloping (i.e., upward and poleward, downward and
equatorward) motions at higher latitudes as we’ll discuss in Chapter 3.
• The fundamental role of water vapour. Not only does it affect atmo-
spheric motions through its effect on buoyancy (condensational heat-
ing, evaporative cooling add or remove buoyancy to air parcels, as we’ll
see in Chapter 3), but water vapour is also the main greenhouse gas
(as we’ll see in Chapter 2). Because the oceans occupy 70% of the
Earth’s surface and because surface evaporation depends on surface
temperature, water vapour couples the state of the oceans to that of
the atmosphere.
• The atmosphere is only one component among many (oceans, cryosphere,

biosphere, the deep Earth, etc) setting the Earth’s climate.
• The atmosphere has a mind of its own. The “butterfly effect” was
introduced by MIT’s meteorologist Ed Lorenz to illustrate the sensitiv-
ity of the atmospheric state to initial conditions. Predictability beyond
a week or so arises from slower changes in boundary conditions (sea
surface temperature, sea ice, vegetation cover, etc). In addition, the
atmosphere is turbulent, with energy transfers towards small scale but
also, more surprisingly, towards large scales. This makes the standard
definition of weather (=state of the atmosphere at a given time) and
climate (=statistics over a long enough time period) a bit ambiguous.
One should really add a “grey zone”, the low frequency variability of
the atmosphere, i.e., fluctuations which can persist for longer than a
week (e.g., blocking conditions associated with long lived cold spells in
the UK like occurred in 2009-2010). These are not “weather”, nor are
they “climate”.
1.4 What drives atmospheric winds, weather

patterns, etc?
A few simple ideas are worth mentioning:
• There is an asymmetry between the radiation received from the Sun and
that emitted by the Earth (surface + atmosphere). Photons emitted
by the Sun have wavelengths smaller than a few microns while photons
Figure 1.3: Seasonal and zonal (i.e., averaged along a latitude circle) mean
atmospheric temperature (top panel, in degree Celcius) and zonal wind (bot-
tom panel, in ms−1 with a contour interval of 10ms−1 , W indicating west
to east winds and E east to west winds) as a function of height/pressure
(vertical axis) and latitude (horizontal axis). Figure taken from Wallace and
Hobbs’ textbook.
Figure 1.4: Global, annual mean atmospheric temperature as a function of

height/pressure.
Figure 1.5: Global composite infrared map on 9 March 2004. White is cold
on this map and, in most regions, indicates the presence of upper level clouds.
Notice the “spotty” nature of the convection in the Tropics and the “wavi-
ness” in middle and high latitudes. You can find many of those maps (as
well as animations) on the MetOffice website.
emitted by the Earth have wavelength larger than a few microns (Fig.
1.6). As a consequence, an atmospheric layer exchanges radiation with
other atmospheric layers and the Earth’s surface (and these exchanges
tend to cancel out), but there is no two-way exchange with Space and
the atmosphere cools radiatively in the infrared (Fig. 1.7)
• In addition, solar photons are in comparison much less absorbed by the

atmosphere than terrestrial photons. This means that, to zero order,
the atmosphere can be thought of as transparent to solar radiation,
the latter being primarily absorbed by the Earth surface. So the at-
mosphere is heated from below by the Earth’s surface, and it cools
radiatively to Space (previous point). This is a very unstable situa-
tion, a bit like a pan of water boiling on a cooker: the atmosphere is
in a state of global convection.
• The radiative cooling to space is relatively uniform spatially but incom-

ing solar radiation peaks at low latitudes as a result of the spherical
shape of the Earth and the large Earth-Sun distance. This means that
in addition to the heating from below and cooling aloft, there is also
a net cooling at high latitudes and net heating at low latitudes (Fig.
Figure 1.6: Planck function Bλ (see Chapter 2) for a body at T = 5780K

(red) and T = 255K (blue) in a log-log scale. The red dashed curve rescales
the red one by a factor π(Rsun /1AU )2 /2π where Rsun is the Sun’s radius, to
account for the different solid angles associated with solar (π(Rsun /1AU )2 )
and terrestrial (2π) radiation.
Figure 1.7: A schematic of the radiative exchanges for an atmospheric layer.

(Left) All exchanges are represented: the layer absorbs radiation from above
and below (including the Earth’s surface), and it emits to other layers above
and below (small arrows). In addition it also emits radiation to Space (large
arrow) but does not absorb solar radiation. (Right) Assuming the inter-layers
and surface exchanges nearly cancel, which is not a bad approximation if
temperature variations are sufficiently weak and the atmosphere is sufficiently
opaque, this leaves a net loss of energy to Space. Picture taken from Wallace
and Hobbs’ textbook.
1.8). Laboratory experiments with rotating tanks cooled at the out-

side and heated at the inside, to mimic the equator-to-pole contrast in
heating, indicate that the resulting motion can be irregular with clear
qualitative similarities with the atmosphere.
The above arguments are rough but they give a feel for the key role of
radiation as a driver of atmospheric motions and weather systems (a full
estimate of the various energy fluxes is given in Fig. 1.9). The laboratory
experiments mentioned in the last bullet point also indicate the very strong
constraint imposed by the rotation of the Earth (rotation rate Ω). It is only
when the latter is fast enough that the simulated flows bear a qualitative
resemblance to the atmosphere. This is because the flow not only transports
heat from the equator to the pole, to balance the deficit highlighted in Fig.
1.8, but also transports atmospheric angular momentum (L). The latter is
simply the azimuthal velocity (u + ΩR cos φ), R being the Earth radius, φ
latitude and u the west-to-east velocity relative to the rotating Earth, times
the distance to the axis of rotation (R cos φ):
L = R cos φ(u + ΩR cos φ) (1.19)
Atmospheric angular momentum has thus a contribution from the Earth’s

solid body rotation, or planetary contribution (ΩR2 cos2 φ), and a contri-
bution from relative motions (uR cos φ). In practice, the former dominates
over the latter. For example at the latitude of the subtropical Jet Stream
(≈ 30◦ ), one has u ≈ 30ms−1 while ΩR cos(30◦ ) ≈ 400ms−1 . Note that in
this derivation, L is angular momentum per unit mass, and that I have used
throughout R + z ≈ R where z is the height of an atmospheric ring above
the Earth’s surface.
Integrated over the whole mass of the atmosphere L is approximatively
constant, i.e., ZZZ
∂
ρLdV ≈ 0 (1.20)
∂t
This implies an intriguing compensation between the Tropics, in which the
surface winds are westward (Trade winds) and thus where the atmosphere
is gaining angular momentum (friction accelerates low levels in the sense
of the Earth’s rotation), and higher latitudes, where the surface winds are
eastward and thus where the atmosphere is loosing angular momentum. As
we shall see in Chapter 4, the Tropics and extra-tropics are coupled through
the propagation of a certain type of waves called Rossby waves. The latter
are excited mostly from midlatitudes by the storm we experience daily and
as they propagate equatorward they transport angular momentum poleward.
Figure 1.8: Global mean radiative fluxes as seen from satellites, in W m−2 , as
a function of the sin of latitude. The top continuous curve is the incoming
solar radiation, while the lower continuous curve also includes the amount
being reflected by the Earth and is thus lower. The dashed line is the infrared
energy emitted by the atmosphere (plus a contribution from the Earth’s
surface). The red + sign indicates net energy gain and the blue − sign
indicates net energy loss.
Figure 1.9: Recent observations of the Earth’s energy budget. All quoted
numbers represent W m−2 . Solar fluxes are in beige and infrared fluxes in
pink. From Stephens et al. (2012).
Thus, the storms we experience daily across the UK provide the mechanical
coupling between vastly distant parts of the Earth.
Finally, note that this compensation between surface easterlies and west-
erlies can only be approximate since torques are also exerted as a result of the
pressure contrast across the major mountains of the Earth like the Himalayas
and the Rockies.
1.5 Further reading

-Stephens et al., 2012: An update on Earth’s energy balance in light of the
latest global observations, Nature Geosciences, 5, 691-696.
-Lovelock, 1979: Gaia, a new look at life on Earth, Oxford University Press.
This book (and the many sequels) offers a fascinating discussion of atmo-
spheric composition and demonstrates the broadness of the subject.
-McPhaden, M., 2015: Playing hide and seek with El Nino, Nature Climate
Change, 5, 791-795.
1.6 Problems
Q1 In the (zonal mean) stratosphere and mesosphere, where (latitude, alti-
tude, season) are found: (i) the coldest temperature (ii) the warmest tem-
perature (iii) the strongest westerly wind? You might find useful to refer to
the slides for Lecture 1.
Q2 Why are deserts more likely to be present in the sub-tropics than in

either the tropics or mid-latitudes?
Q3 These two subquestions are independent of each other.

(i) At 25 km altitude, where atmospheric pressure is Po ≈ 25hP a and
temperature is To ≈ 220K, the mass mixing ratio of ozone is 10 parts
per million. Compute (a) density and (b) partial pressure of ozone
stating any assumptions made. Data: Rd = 287JK −1 kg −1 .
(ii) Express the fraction of water vapour molecules in ppm (part per mil-
lion) in a sample of air where the partial pressure of water vapour is
e = 10hP a and that of dry air Pd = 1000hP a. Compare the number ob-
tained with the current fraction of carbon dioxide molecules (400ppm).
Q4
1.6. PROBLEMS 19
Table 1.1: Data for Q5. The number in parentheses refer to the molecular
mass of the main atmospheric constituents.
Planet Major atmospheric Mean lower Mass Radius
component temperature (K) kg km
24
Venus CO2 (44) 750 4.87 × 10 6,051
Earth N2, O2 (29) 280 5.97 × 1024 6,371
Mars CO2 (44) 250 6.42 × 1023 3,397
Jupiter H2 (2) 123 1.90 × 1027 71,490
(i) One kg of air of specific humidity q1 = 10g/kg is mixed with one kg of

air of specific humidity q2 = 5g/kg. What is the specific humidity q3
of the air mass after mixing? You may assume the conditions are such
that no phase change occurs.
(ii) In the jargon of Thermodynamics, is specific humidity an intensive or
an extensive variable?
Q5 (i) Show that in an isothermal atmosphere (T = To ) the pressure decays

exponentially with height with a scale Hs = kB To /mg (the scale height), in
which g is the gravity of the planet and m is the mass of an atmospheric
molecule. (ii) Estimate Hs in the lower atmosphere for each of the planet
listed in Table 1.1.
Q6 Show that the specific volume α of a sample of air can be approximated

as,
α ≈ αd (1 − qt ) (1.21)
in which αd = Rd T /Pd is the specific volume of dry air (Rd being the gas
constant for dry air, Pd the partial pressure of dry air and T temperature).
You might want to start by writing that α = (V +Vl +Vi )/(md +mv +ml +mi )
in which V is the volume of the sample occupied by the gas phase, Vl and Vi
that of the liquid and ice phases, respectively.
Q7 By inspection of a surface pressure map (for example from the on-line

ERA40Atlas2 ), work out whether the Rocky mountains tend to increase or
decrease atmospheric angular momentum.
Q8 A simple view of the Tropics is that it is primarily independent of lon-

gitude, i.e., rings of air flow upward at the equator in the ascending branch
2
See the web link on Blackboard.
of the Hadley cell, and flow poleward at upper levels, conserving their angu-
lar momentum. Estimate the implied zonal velocity u (relative to the solid
body rotation of the Earth) at 30◦ of latitude. How does it compare with
the observed velocity? Hint: you may assume that the relative velocity u of
the ring is very small near the ground.
Q9 Using the numbers in Fig. 1.9, discuss whether the atmosphere can be
reasonably described as transparent to solar radiation.
Chapter 2
Radiative heating and cooling

of the atmosphere
Key concepts: solid angle, irradiance, radiation balance, greenhouse effect,

Beer’s law, Schwarzschild’s equation, infrared cooling to Space.
2.1 Concepts and definitions

2.1.1 Intensity of radiation
The intensity of electromagnetic radiation at wavelength λ measures the en-
ergy crossing a unit area perpendicular to the direction of propagation per
unit time, per unit wavelength interval, and per unit solid angle (Fig. 2.1).
The “per unit area” and “per unit time” are familiar, but the “per unit wave-
length interval” and “per unit solid angle” is less so. The “per unit wave-
length interval”, or “per wavelength” for short, is needed to account for the
spectrum of electromagnetic radiation, theRtotal energy flux being computed
as the integral over all wavelengths (i.e., Fλ dλ). In absence of scattering
or absorption, the intensity of radiation, denoted by Iλ , is conserved (see the
Appendix if you’re interested to read more about this).
The “per unit solid angle” is included to represent the 3D nature of
electromagnetic radiation. The radiation reaching P in Fig. 2.1 in a given
direction is made of an infinitesimal cone of rays, or “radiation pencil”, filling
a certain fraction of the sky. This fraction is measured by solid angle Ω,
exactly like an angle is a measure of length on a unit radius circle (Fig. 2.2).
Solid angle is expressed in steradians (sr) and the maximum solid angle
attainable (the amount of space filled by the sky if we were floating in the
air) is 4π ≈ 12.5sr (an hemisphere is 2π).
21
22CHAPTER 2. RADIATIVE HEATING AND COOLING OF THE ATMOSPHERE
Figure 2.1: The intensity of radiation at wavelength λ is denoted by Iλ and

measures the energy flow along the direction of propagation per unit area
(dS) and time (dt), per unit wavelength (dλ) and unit solid angle (dΩ).
Figure 2.2: From vector Calculus (year 1), the area element on a sphere of
unit radius r = 1 is dΩ = sin θdθdφ. This is the infinitesimal solid angle
which we will use throughout in this chapter.
2.1. CONCEPTS AND DEFINITIONS 23
Figure 2.3: Example of calculation of solid angle. The total solid angle of
the Sun at a distance r is the sum over the angle θ (from 0 to α) of all the
small solid angles 2π sin θdθ indicated by the black shell.
Example: Compute the solid angle of the Sun (radius R), as seen from a
distance r to its center (Fig. 2.3). The Sun R 2πisRseen by an observer on the
α
sphere of radius r with a solid angle Ω(r) = 0 0 sin θdθdφ = 2π(1 − cos α)
where the angle α satisfies tan α = R/r (note the right angle magenta triangle
defining α in Fig. 2.3). It covers a solid angle Ω ≈ 2π very close to the source
(α = π/2) and Ω ≈ π(R/r)2 for r R since 1 − cos α ≈ α2 /2 when α is
small.
2.1.2 Blackbody radiation

As taught in Year 2 (Thermodynamics and Statistical Physics), blackbody
radiation is the radiation emitted by a body in thermal equilibrium. It is
isotropic and only depends on the temperature T characterizing the equilib-
rium, not on the nature of the material making the body. Its intensity, at a
given wavelength λ is given by the Plank function,
2hc2 λ−5
Bλ (T ) = hc/λk T (2.1)
e B −1
which has units of W m−2 per wavelength per solid angle.
2.1.3 Shortwave and longwave radiation

We saw in section 1.4 that there is a small overlap of the Planck functions
associated with terrestrial (infrared) and solar (visible) emissions (Fig. 1.6
–note the scaling of the Planck function by the solid angle of the Sun as
seen from the Earth which is, from the calculation in section 2.1.1, equal
to π(R/r)2 ). It is, as a result, common practice to separate “longwave” and
“shortwave” radiations. Typically, 4µm is used as the separation between the
two. As we shall see the difference in wavelength will lead to important differ-
ences with respect to the role played by scattering and absorption/emission
in the conservation of radiation intensity for longwave (scattering negligible,
absorption/emission important) and shortwave radiation (scattering impor-
tant, absorption important, emission negligible).
2.1.4 Irradiance
The energy of radiation passing through an horizontal plane, per unit area of
that plane, per unit wavelength is called the monochromatic irradiance Fλ .
It requires integrating Iλ over solid angle (2π at most for either upward or
downward hemispheres) and taking into account the angle between the beam
and the normal to the horizontal plane.
Consider for example the geometry in Fig. 2.4 and take the horizontal
plane to be the x, y plane. The net downward radiation across the horizontal
plane is made of several “radiation pencils”, each coming from different angles
θ with the z direction and the polar angle φ in the horizontal plane. It is
thus a matter of summing over all these pencils, each of infinitesimal solid
angle dΩ, and projecting onto the vertical Iλ → Iλ cos θ. Thus,
Z
Fλ = Iλ cos θdΩ (2.2)
or using dΩ = sin θdθdφ,

Z 2π Z π/2
Fλ = Iλ cos θ sin θdθdφ (2.3)
φ=0 θ=0
Figure 2.4: Geometry for the calculation of irradiance, given a pencil of

monochromatic radiation (green cone) of intensity Iλ propagating downward
(green dashed arrow) across the horizontal plane at an angle θ.
If the radiation is isotropic (i.e., independent of θ and φ), then the integral
simplifies to
Z π/2
Fλ = 2πIλ cos θ sin θdθ = πIλ (2.4)
θ=0
This result is exact for Blackbody radiation, for which Iλ = Bλ in (2.1), and
so Fλ = πBλ R(if one were to integrate the latter over all wavelengths, one
would obtain πBλ dλ = σT 4 in which σ is Stefan-Boltzmann’s constant).
NB: The irradiance just defined is, strictly speaking, a downward irradiance
(we counted solid angles from above). We’ll denote it by Fλ↓ in the following.
We could also have computed the energy flux received from below, producing,
for Iλ = Bλ , the same result: Fλ↑ = πBλ (“upward irradiance”).
2.1.5 Kirchoff ’s law

In practice, most components of the climate system do not behave like black
bodies, or only do so over a limited range of wavenumbers. For example
the spectrum of most gases is made of sharp spectral lines, not a contin-
uum; in addition, gases not only absorb radiation but also transmit it, unlike
blackbodies. We’ll see indeed in section 2.3 that the atmosphere is almost
transparent to infrared radiation in the 10−12µm, the so-called “atmospheric
window” region (this is the spectral region chosen to make infrared pictures
of the Earth since in this region infrared radiation either originates from the
Earth’s surface or from cloud tops –see for example Fig. 1.5).
Surprisingly, even though the atmosphere does not emit according to
Bλ (T ), one can relate atmospheric emission and absorption to Bλ (T ) by
using Kirchoff’s law. To see this, define the emissivity of a particular body
at temperature T (e.g., sample of air) according to,
λ ≡ Iλ (emitted)/Bλ (T ) (2.5)
and define its absorbtivity as,
αλ ≡ Iλ (absorbed)/Iλ (incident) (2.6)
Kirchoff’s law states that, remarkably, irrespective of what the body is made,
and irrespective of the nature of the radiation (isotropic or not),
λ = αλ (Kirchoff’s law) (2.7)
NB: For a blackbody, λ = αλ = 1 for all wavelengths. Note also that

Kirchoff’s law only applies to systems in thermodynamic equilibrium with
their immediate surroundings (“local” thermodynamic equilibrium). This

condition is satisfied in the troposphere and the stratosphere but less so
above these regions.
2.1.6 The “solar constant”

We now apply the concept of radiation intensity and radiation pencils to
estimate the amount of solar energy received per unit time by the Earth.
This quantity is referred to as the solar constant So (in units of W ), although
it fluctuates on many timescales (e.g., the decadal solar cycle)1
So we consider a point P on Earth at latitude φ and longitude γ (Fig.
2.5) and first estimate the flux of energy received at that point. We will then
sum over all points to obtain an expression for So . Because the Sun is so far
away from the Earth, we are going to treat the solar radiation received at
P, at a given wavelength λ, as made of only one pencil of radiation. Indeed,
as the calculation in section 2.1.1 showed, the Sun covers a solid angle angle
δΩ ≈ π(Rs /1AU )2 where Rs is the Sun’s radius and 1AU (astronomical unit)
is the mean Earth-Sun distance, that is a very small solid angle. This is just
a mathematical way to say that the Sun covers only a small fraction of the
sky. In addition, because the Sun is so far away form the Earth, we also
approximate the angle made between this pencil and the local vertical as the
latitude φ in a plane of constant longitude (Fig. 2.5). This is just saying that
we are treating the impinging solar radiation as a plane wave. The Note that
Fig. 2.5 only shows this projection in a plane of constant longitude. You
would need to view the figure from the North pole to see the same effect in a
plane of constant latitude, the result being that the angle between the pencil
and the local vertical in that plane is approximatively π/2 − γ if 0 ≤ γ ≤ π).
With these approximations, the flux of energy per wavelength Φλ received
from the Sun at P is,
Φλ ≈ Iλ cos φ sin γδΩ, (2.8)
so that the power Pλ per unit wavelength integrated over the Earth is simply,
Z
Pλ ≈ Φλ R2 cos φdφdγ (2.9)
Note that in this expression the latitude φ is integrated from South to North
pole (−π/2 and π/2, respectively) but the longitude is only integrated over
1
Once you realise how amazingly active the Sun is this is not a surprise. Have a look at
the fantastic movies NASA has made from the SDO mission (e.g., google “thermonuclear
art SDO”).
Figure 2.5: Schematic for the calculation of the Solar constant. Note that
the diagram is (obviously!) not on the correct scale.
the day hemisphere, i.e. from 0 to π. Inserting (2.8), we obtain,

Z
Pλ ≈ Iλ δΩ R2 cos2 φ sin γdφdγ = (Iλ δΩ)πR2 (2.10)
The above relation shows that, under our approximation of parallel im-
pinging solar radiation, it all looks as if the Earth was a disk of radius πR2 .
Integrating over wavelength, we define the Solar constant as,
Z
So ≡ Iλ δΩdλ ≈ 1, 361W m−2 (2.11)
so that the net power received by the Earth is So πR2 .

NB: This is a calculation valid for equinoctial conditions, or annual mean
conditions, only. One would need to include the tilt of 23.5◦ of the ecliptic
plane to estimate the total energy gained at the solstices.
2.1.7 Radiation balance and emission temperature

A simplified energy balance for the Earth is given in Fig. 2.6. A fraction
αP (the planetary albedo) of the Solar radiation discussed in the previous
subsection is reflected. At equilibrium, the same amount of power must be
Figure 2.6: Radiation balance of the Earth. The incoming solar radiation
is treated as a parallel beam impinging on a spherical Earth. The outgoing
infrared radiation is isotropic.
lost by the Earth. The emission temperature is defined as the temperature

required to achieve this, were the Earth a perfect blackbody in the infrared:
πR2 So (1 − αP ) = 4πR2 σTe4 (2.12)
leading to
1/4
So (1 − αP )
Te ≡ (2.13)
4σ
NB: note the factor of 4 coming from the geometry of the problem (plane
radiation impinging a sphere as opposed to radial emission).
The idea of radiative balance at the TOA is an idealization. Global
conservation of energy requires that any imbalance be reflected in a change
in heat content. This, in practice, is dominated by oceanic heat storage
(largest heat capacity in the climate system) which fluctuates on very long
timescales (decades and longer) because of the slow ocean dynamics. Thus
the TOA net radiative fluxes are not expected to vanish on timescales shorter
than at least a few decades. We will come back to this in Chapter 5 when
discussing anthropogenic climate change.
For Earth annual average, αP = 0.3 so that Te = 255K or −18◦ C (very
cold!). The Earth’s surface temperature is about 288K, or about +15◦ C.
Thus, by contrast with the present model which omits entirely the atmo-
sphere, we can say that the atmosphere is responsible for a ≈ 288−255 = 33K
increase in surface temperature. The way this works is disentangled in the
simple model coming next. Before doing so, it is worth mentioning a couple
of other interesting aspects of this model:
• it suggests that the bulk of the infrared radiation seen from Space orig-
inates from the atmosphere itself rather than from the surface because
255K is found typically at an altitude of 5km above the Earth’s surface.
• the Planck function for a blackbody at 255K is centered near 15µm

(Fig. 1.6). This happens to be a wavelength at which the CO2 molecule
absorbs strongly radiation, hence the strong “leverage” of CO2 on cli-
mate.
2.2 A simple model of the greenhouse effect

We consider a 0D model of radiative balance (averaged over the whole Earth’s
surface area and expressed in W m−2 ) and go a little beyond the previous
section by explicitly introducing the surface temperature Ts (Fig. 2.7).
In the shortwave, the solar flux impinging at the TOA is still So (1−αP )/4
which we denote by Fo . We further assume that only a fraction Tsw of
this radiation reaches the surface, to account for absorption by atmospheric
molecules and aerosols. (we’ll call Tsw the transmissivity of the atmosphere
in the shortwave in the following). From Kirchoff’s law, if some radiation
is absorbed it must also be emitted. However, at terrestrial temperature
the emission in the range of wavelengths where the bulk of solar radiation
resides is negligible, and λ Bλ (T ) αλ Iincoming . As a result, even though
αλ = λ for wavelengths λ in the shortwave part of the spectrum, we can
safely neglect the emission of shortwave radiation by the atmosphere.
In the longwave, we take the atmosphere to be at constant temperature
Ta and denote by Fa the longwave flux it emits upward and downward. The
surface is treated as a blackbody, thus emitting Fs = σTs4 upward. A frac-
tion TLW of this radiation reaches the top-of-the-atmosphere (we’ll call this
fraction the transmissivity of the atmosphere in the longwave). Fig. 2.7
summarizes the energy flows.
We can predict what the surface temperature should be solely from energy
conservation arguments. At equilibrium, the net flow of energy across any
surface must be zero. Applying this at the Earth’s surface yields (see Fig.
2.7):
Fo Tsw + Fa = Fs (2.14)
2.2. A SIMPLE MODEL OF THE GREENHOUSE EFFECT 31
Figure 2.7: A simple model of the greenhouse effect. TOA denotes the
“top-of-the-atmosphere” where pressure vanishes. The shortwave fluxes are
indicated in black, the longwave ones in blue.
while, at the “top-of-the-atmosphere”, it produces:
Fo = Fa + Fs Tlw (2.15)
By assumption Fs = σTs4 while, using (2.13), Fo = σTe4 . As a result,

after elimination of Fa from the above two energy conservation equations, we
obtain,
1 + Tsw 1/4
Ts = Te ( ) (2.16)
1 + Tlw
This equation is remarkably simple and shows that the surface temper-
ature and the emission temperature differ only by a factor proportional to
the transmissivities in the shortwave and the infrared. If there were no at-
mosphere, Tsw = Tlw = 1 and Ts = Te = 255K. With an atmosphere, and
if Tsw > Tlw , the surface temperature will then exceed Te . In the Earth
atmosphere, Tlw ≈ 0.2 while Tsw ≈ 0.9, leading to Ts ' 286K.
The agreement of this prediction with the observed Ts = 288K is fortu-
itous because of the extreme simplicity of the model (isothermal atmosphere,
surface treated as a blackbody). The key point though is that because the at-
mosphere is more transparent to shortwave than it is to longwave (Tsw > Tlw ),
there is a “recycling” of energy towards the surface:
1 + Tsw
Fo Tsw + Fa = Fo > Fo (surface heating) (2.17)
1 + Tlw
The added heating leads to a larger surface temperature –this effect is called
the greenhouse effect.
2.3 Beer’s law

NB: The “atmospheric window”, absorption, emission and scattering of ra-
diation by atmospheric molecules and aerosols is dicussed in the ppt slide for
this chapter.
Consider a monochromatic beam of wavelength λ and of intensity Iλ . We

want to derive the change in intensity (dIλ ) of this beam along its direction
of propagation (measured by the coordinate s). This change is caused ei-
ther because some air molecules scatter the radiation (net loss of radiation
intensity along the path but no change in radiation intensity integrated over
all directions), or absorb it (net loss of radiation intensity along the path
but no change in radiation intensity in other directions). If we denote by qa
the mass of these molecules per unit mass of air, ρqa is the mass of these
2.4. SCHWARZCHILD’S EQUATION 33
molecules per unit volume and ρqa ds is the mass of those molecules per unit
area perpendicular to the direction of propagation.
Beer’s law states that,
dIλ = −Iλ kλ ρqa ds (2.18)
in which kλ (in m2 kg −1 ), called the extinction coefficient, measures the in-

tensity of absorption or scattering of radiation of wavelength λ,
kλ = βλ (absorption) + σλ (scattering) (2.19)
Equation (2.18) can be integrated along the path of the beam as,
Rs
kλ ρ(s0 )qa (s0 )ds0
Iλ (s) = Iλ (so )e− so (2.20)
In this expression so is a starting point where the radiation intensity is Iλ (so ).

It is convenient to introduce the following quantities,
Rs 0 0 0
Tλ (so , s) = e− so kλ ρ(s )qa (s )ds (transmittance) (2.21)
Z s
τλ (so , s) = kλ ρ(s0 )qa (s0 )ds0 (optical depth) (2.22)
so
Note that both τλ > 0 and 0 ≤ Tλ ≤ 1 are non dimensional functions (i.e.,
they are just numbers, without SI units). One measures the opacity of the
atmospheric layer over a given path (τλ , the larger the more opaque the layer;
hardly any radiation escapes a layer with optical depth much greater than
unity), while the other measures its transparency (Tλ , the closer to unity the
more transparent the layer –see for example the discussion of the greenhouse
effect earlier in this chapter).
With these notations, (2.20) can be rewritten more compactly as:
Iλ (s) = Iλ (so )Tλ (so , s) (2.23)
expressing that if only extinction is considered, the intensity at s is that at

so times the transmittance along the path (so , s).
2.4 Schwarzchild’s equation

2.4.1 Derivation
Beer’s law only considers removal of radiation from a beam. Radiation can
however also be added to the beam due to the emission from the layer or
to radiation incident from another direction being scattered into the beam.
The additional elements also depend on the amount of radiatively active con-
stituent ρqa ds so Beer’s law can be modified to include them by introducing
a source function Jλ ,
dIλ = −kλ ρqa dsIλ (s) + kλ ρqa dsJλ (s) (2.24)
This can be simplified by introducing again the optical depth –see eq.
(2.22) in which so is simply taken as a reference location,
dτλ = kλ ρqa ds (2.25)
Doing so allows a change of variables (s → τλ ) and (2.24) can be rewritten
as,
d(Iλ eτλ )/dτλ = Jλ eτλ (2.26)
Integrating from so to s, and noting that τλ (so , so ) = 0, we obtain,
Z τ
−τ 0
Iλ (s) = Iλ (so )e + Jλ (τ 0 )e−(τ −τ ) dτ 0 (2.27)
0
with τ 0 = τλ (so , s0 ) and τ = τλ (so , s) (to simplify notations).
2.4.2 Physical interpretation

Let us step back a little from these calculations. The first term on the r.h.s.
of the previous equation is readily interpreted as the radiation emitted at so ,
a fraction e−τ making to s, in agreement with Beer’s law. The second term,
the integral, reflects the contribution to the radiation at s emitted from all
layers between so and s. To see this more clearly, consider the following.
Measuring the location of any of these layers by s0 with so ≤ s0 ≤ s, we know
from (2.24) that they emit an amount of radiation Jλ (s0 )kλ ρqa ds = Jλ (s0 )dτ 0 .
From the definition of transmittance, only Tλ (s0 , s)Jλ (s0 )dτ 0 reaches s. This is
0
nothing else than Jλ (s0 )e−(τ −τ ) dτ 0 since Tλ (s0 , s) = Tλ (so , s)/Tλ (so , s0 ). Thus
we can as well rewrite (2.27) as,
Z τλ (so ,s)
Iλ (s) = Iλ (so )Tλ (so , s) + Tλ (s0 , s)Jλ (τ 0 )dτ 0 (2.28)
0
2.4.3 Final form

Eq. (2.28) has a clear physical interpretation but, for practical purposes, the
initial form (2.27) can be put to better use. Start from:
0
dTλ (s0 , s) = e−(τ −τ ) dτ 0 (2.29)
2.5. SOME APPLICATIONS OF SCHWARZCHILD’S EQUATION 35
(NB: here s is constant and the derivative applies to the variable s0 , or τ 0 ).

After division by ds0 , this reads simply:
Z s
dTλ (s0 , s)

0
Iλ (s) = Iλ (so )Tλ (so , s) + Jλ (s ) 0
ds0 (2.30)
so ds
This is the most convenient form of Schwarzchild’s equation to use because:
(i) the contribution from atmospheric layers to the radiation intensity at s
(the integral term) is expressed using the natural coordinate s0 rather
than the mathematical function τ 0 .
(ii) this form suggests that the contribution to intensity at s from the
emission by a layer at s0 (i.e., Jλ (s0 )) is weighted by the gradient of
the transmissivity between s0 and s (i.e., dTλ (s0 , s)/ds0 ). As the section
below shows, this turns out to be very useful to gain a bit more intuition
about how radiative transfer works.
2.5 Some applications of Schwarzchild’s equa-

tion
2.5.1 No scattering
So far the exact form of the source function Jλ has not been discussed. When
scattering is important as a source of radiation along a particular line-of-sight
(or “pencil of radiation”), the problem is mathematically difficult as one must
integrate over all solid angles a probability function that photons have been
scattered into that line-of-sight. So it would be useful to know when we can
neglect this complexity.
In the infrared part of the spectrum, scattered radiation can typically
be neglected compared to the radiation emitted directly in a given line of
sight so the approximation is usually adequate. In the shortwave part of the
spectrum, except for a line-of-sight directed at the Sun, all pencils of radiation
are made of scattered radiation, and so the complexity has to be met fully.
If scattering can be neglected, i.e., σλ = 0 in (2.19) (so typically for
the infrared radiation, based on the prevous discussion), the Schwarzschild’s
equation takes a particularly simple form. Indeed, in that case, the absorp-
tivity of the layer of thickness ds is,
αλ = Iλ βλ ρqa ds/Iλ = Iλ kλ ρqa ds/Iλ = kλ ρqa ds (2.31)
while its emissivity is,
λ = Jλ kλ ρqa ds/Bλ (2.32)
Using Kirchoff’s law (2.7), λ = αλ which leads to 1 = Jλ /Bλ and so

Schwarzschild’s equation can be simply rewritten as,
Z s
dTλ (s0 , s)

0
Iλ (s) = Iλ (so )Tλ (so , s) + Bλ (s ) 0
ds0 (2.33)
so ds
2.5.2 Infrared radiation by an isothermal atmosphere

Consider the case of an isothermal atmosphere and neglect scattering. The
downward infrared radiation measured at a distance s from the top-of-the-
atmosphere (s = so ) can be computed from (2.33) as follows.
First, acknowledge that for downward infrared radiation, Iλ (so ) = 0, so
that the first term on the r.h.s of (2.33) vanishes. For the integral term,
one can take Jλ = Bλ outside it since the atmosphere is isothermal (say
Bλ = Bo ). We thus have,
Iλ (s) = Bo [Tλ (s, s) − Tλ (so , s)] = Bo [1 − Tλ (so , s)] (2.34)
since the transmissivity between s and s is unity.
Apply this result to absorption/emission by water vapour for example.
We can very reasonably neglect the amount of water vapour above the
tropopause, so the transmittance, measured from so (the top-of-the-atmosphere)
to s will look like schematized in Fig. 2.8 (red curve): unity until the
tropopause is reached and then decaying towards the surface. Conversely,
the radiation intensity is initially zero until the tropopause is reached, and
then increases towards the Earth’s surface. Note that the intensity at the
Earth’s surface approaches Bo as the transmittance of the atmosphere de-
creases (Tλ → 0). This is expected since in this limit the troposphere behaves
like a Blackbody. The spectral line (or wavelength λ) is then said to be “sat-
urated”.
2.5.3 Remote sensing of temperature

A useful application of Schwarzschild equation is the measurement of at-
mospheric temperature from satellites. For a passive instrument, we would
consider the upward radiation of longwave radiation emitted by the atmo-
sphere and the surface. We thus consider the full expression (2.33), with so
being at the Earth’s surface and Iλ (so ) = Bλ (SST ) (assuming we are over
the oceans, which are excellent blackbodies in the longwave –SST is the sea
surface temperature). The path variable s thus increases upwards and the
transmittance Tλ (so , s) is measured from the sea to any point in the atmo-
sphere above. We are interested in the radiation received by a satellite, i.e.,
at very large s.
2.5. SOME APPLICATIONS OF SCHWARZCHILD’S EQUATION 37
Figure 2.8: Simple application of Schwarzchild’s equation to downward long-

wave radiation. The lower atmosphere (blue region) is assumed to be ra-
diatively active at this wavelength but not the layer above it. In the latter
Iλ = 0 and Tλ (so , s) = 1.
Imagine that there is a strong absorption of infrared radiation at wave-

length λ at a distance sa from the sea, and very little absorption elsewhere
(Fig. 2.9). The transmittance Tλ (s0 , s) would then be nearly zero for s0 ≤ sa
(the (s0 , s) layer would then include the absorbing layer), while it would be
close to unity for s0 ≥ sa (the (s0 , s) layer would then be above the absorbing
layer). As a result, the derivative dTλ (s0 , s)/ds0 would have a “narrow bell”
shape centred at s0 = sa (Fig. 2.9). The intensity measured at s (say aboard
a satellite) would then be
Z s
dTλ (s0 , s)

0
Iλ (s) = Jλ (s ) 0
ds0 ≈ Jλ (sa ) (2.35)
so ds
where I have used the fact that Tλ (so , s) = 0 in eq. (2.33) in this example, and
I have approximated the narrow bell shaped derivative as a Dirac function.
The source function Jλ (sa ) is a strong function of temperature and so, after
calibration, the previous equation provides an estimate of temperature of the
atmosphere at s = sa .
In practice, rather than measuring temperature at a precise location,
the measurement provides temperature over a weighted layer (i.e., the bell
shaped derivative is not exactly a Dirac function. In addition one could
more realistically consider absorbers with a uniform mixing ratio and use
the different strength of absorption lines –see some examples in the slides for
Figure 2.9: Same as Fig. 2.8 but for upward longwave radiation in the
presence of a localized absorbing layer.
this chapter). Typical wavelengths used are the 15µm and 4.3µm bands of
CO2 and the microwave (5mm) band of O2 .
2.6 Radiative heating and cooling rates

The difference between the radiation incoming and outgoing from the sides of
a given volume of air is, by conservation of energy, a heating rate. Although
conceptually simple, the calculation of this heating rate is made difficult in
practice because of the need to integrate the radiation coming/outgoing from
all sides of the sample. In the case of the global atmosphere, the sides in
question are spheres of constant radius, or, in the Cartesian geometry adopted
here for simplicity, planes of constant height. We will thus concentrate in
this section on the heating of a layer sandwiched between height z and z +dz.
From section 2.1.4, the total flux of radiation across an horizontal plane
due to a beam of intensity Iλ was defined as the irradiance Fλ . To distinguish
between radiation coming from above and below, we will separately consider
Fλ↑ and Fλ↓ . The heating rate Qλ of an infinitesimal layer of air sandwiched
between height z and z + dz is thus (Fig. 2.10),
d
Qλ = (Fλ↓ − Fλ↑ ) (2.36)
dz
Because Fλ is in units of W m−2 per wavelength, Qλ is in units of W m−3 per
2.6. RADIATIVE HEATING AND COOLING RATES 39
Figure 2.10: Schematic of the various terms in the heat budget of an infinites-
imal layer sandwiched between z and z + dz. The heat gained per unit time,
area and wavelength is Fλ↑ (z) + Fλ↓ (z + dz). Likewise, the heat lost per unit
time, area and wavelength is: Fλ↑ (z + dz) + Fλ↓ (z). Thus the net heat gained
per unit area, time and wavelength is Fλ↑ (z)−Fλ↑ (z+dz)+Fλ↓ (z+dz)−Fλ↓ (z) =
(dFλ↓ /dz − dFλ↑ /dz)dz. Dividing this expression by dz gives the heating per
unit volume, wavelength and unit time.
wavelength. The total heating rate due to radiation is thus Qrad ,

Z +∞
Qrad ≡ Qλ dλ (2.37)
0
which has units of W m−3 . As we shall see, and consistent with the numbers
in Fig. 1.9, the heating due to shortwave absorption is more than offset
by radiative cooling due to longwave emission so that in the net, Qrad < 0
(cooling).
2.6.1 Shortwave heating

Let’s take advantage of the hard work in section 2.1.6, where we showed by
surface integration on a sphere that the solar energy received by the Earth
was, at a given wavelength λ, equal to,
Pλ = (Iλ δΩ)πR2 (2.38)
In this expression, both the Iλ and R are strictly speaking referring to quan-
tities at the top-of-the-atmosphere since we did not account for atmospheric
absorption and scattering and we were, in this section, only interested in the
energy available as a whole by the Earth. But we can re-interpret this result
as saying that, per unit horizontal area at a height z, the irradiance is,
Fλ↓ (z) = Iλ (z)δΩπR2 /4πR2 = Iλ (z)δΩ/4 (2.39)
Using Beer’s law, we have Iλ (z) = Iλ,T OA e−τλ (z) where τλ (z) is the optical
thickness of the atmosphere from the top-of-the-atmosphere to a height z.
Hence we write,
Fλ↓ (z) = Fλ,T

↓
OA e
−τλ (z)
(2.40)
↓
where Fλ,T OA = Iλ,T OA δΩ/4.
The contribution to Qλ coming from Fλ↓ is then,
d ↓
Qλ = [F e−τλ (z) ] (2.41)
dz λ,T OA
↓ −τλ (z)
= (−Fλ,T OA e )(−ρqa kλ ) (2.42)
= Fλ↓ ρqa kλ (2.43)
in which we have used (2.22). Fig. 2.11 gives a schematic of the vertical
variations of Fλ↓ and ρqa = ρa , as well as a scale for the optical depth.
The downward radiation decreases monotonically as we go downward, as
expected, and the density of absorber is assumed to be exponential-like.
As can be seen, Qλ , the product of these two2 , peaks at a height where
the optical depth is close to unity. Physically, well above the level of unit
optical depth, the incoming beam is virtually undepleted, but the density is
so low that there are too few molecules to produce significant absorption and
heating. Likewise, well below the level of unit optical depth, there are a lot
of molecules to produce absorption and heating, but there is not much left
to absorb as the beam has been mostly depleted. You are invited to prove
this result mathematically in Q5 below.
Detailed calculations, using a “line-by-line radiation code” and including
the contribution to Qλ from Fλ↑ , are shown in Fig. 2.12 (focus here on the
right hand side, i.e., shortwave heating rates). The heating rates are given in
units of K/day, i.e., Qλ /(ρcp ) is plotted rather than Qλ . Absorption by ozone
in the stratosphere dominates the heating rate, on the order of 5 − 10K/day.
After this and O2 at high altitude (above the mesopause), the next most
important absorber of solar radiation is water vapour, which contributes to
a relatively uniform heating of the troposphere on the order of 0.5K/day.
2
We are neglecting here the temperature and pressure dependence of kλ .
Figure 2.11: Schematic of heating rate due to solar absorption of radiation of

intensity Iλ . See text for explanations. Figure from the Wallace and Hobbs’
textbook.
NB: Why divide by cp and not cv to express heating rates in K/day? This is
an interesting question. In the 1st law of Thermodynamics, dU = δQ+δW =
δQ − P dV , the heating δQ is not a state function, i.e., it depends on the path
considered. Radiative cooling/heating does not involve adding/substracting
mass, and mass is constant within a given pressure layer under the hydro-
static approximation. Thus, we can consider that an infinitesimal layer of air
(thickness dz) of fixed mass has a constant pressure (the pressure variations
within the layer are of order dP P so the pressure of the layer is constant
at ≈ P ). This suggests to use instead dU = δQ − d(P V ) + V dP which,
after using P V = N kB T leads to (CV + N kB )dT ≡ CP dT = δQ + V dP . At
constant pressure, a given amount of heat will lead to a temperature change
after division by Cp . At constant volume, the same amount of heat would
lead to a larger temperature change since CV < Cp .
2.6.2 Longwave cooling

It is impossible to neglect the upward irradiance of longwave radiation (this is
the physical mechanism allowing the Earth to cool to Space, see Chapter 1).
Figure 2.12: Global mean longwave (left panel) and shortwave (right panel)
heating rates in K/day as a function of altitude showing contributions of the
major gases. After D. Andrews’ textbook.
Nor is it possible to neglect the downward infrared irradiance responsible for

the greenhouse effect. We’ll thus compute both terms in this section. Besides
this, the other important difference between the calculation here and that in
the previous section regards the transformation from Iλ to Fλ . As illustrated
in fig. 2.13 this simply reflects the well defined source of shortwave radia-
tion (the Sun) as opposed to the more diffuse sources of longwave radiation
(Earth’s surface, atmospheric molecules, clouds, etc). Note that the different
pencils in Fig. 2.13 have also different path lengths. To treat this effect
explicitely without enhancing the complexity too much we’ll use the “par-
allel plane approximation”, which treats the atmosphere as homogeneous in
the horizontal direction: one can then use for all pencils the optical thickness
measured along the direction z, after taking into account the increase in path
length by the factor 1/ cos θ (see Fig. 2.13).
In the calculations below we’ll neglect scattering of longwave radiation
and will accordingly use Jλ = Bλ (section 2.4.2) in Schwarzchild equation,
i.e., Z τ
0
Iλ (s) = Iλ (so )e−τ + Bλ (τ 0 )e−(τ −τ ) dτ 0 (2.44)
0
with τ = τλ (so , s) and τ 0 = τλ (so , s0 ).
Upward irradiance
Let’s first consider the contribution to Fλ↑ arising from emission by the Earth’s
surface, i.e., the term corresponding to Iλ↑ (so ) = Bλ (Ts ) where Ts is the
Earth’s surface temperature. Taking into account the 1/ cos θ term, the con-
↑
tribution Fλ,surf to the upward irradiance from the Earth’s surface is,
Z 2π
↑
Fλ,surf (z) = Bλ (Ts )e−τ (z)/ cos θ cos θdΩ (2.45)
0
in which τ (z) = τλ (0, z) now measures the optical depth of the air column
from the surface to height z, and θ is the angle made by each pencil of
radiation with the vertical. Consistent with the parallel plane approximation,
we assume that Ts does not vary too much, which allows us to take the Bλ (Ts )
term outside the integral. After doing this, and introducing µ = cos θ, we
have, Z 1
↑
Fλ,surf (z) = 2πBλ (Ts ) e−τ (z)/µ µdµ (2.46)
0
R1 ◦
A useful approximation is that 2 0
e−τ /µ µdµ ≈ e−τ / cos(53 ) = e−1.66τ (“diffuse
approximation”). As a result,
↑
Fλ,surf (z) ≈ πBλ (Ts )e−1.66τ (z) (2.47)
Figure 2.13: Schematic of the “radiation pencils” to be considered when com-

puting irradiances at point P for the shortwave (left) and longwave (right)
part of the spectrum. In absence of scattering, only one radiation pencil (the
one covering the small solid angle δΩ subtented by the Sun) would need to
be considered for the shortwave part of the spectrum –but many need to
be taken into account for infrared radiation. The distance ds along a path
making an angle θ with the vertical is related to dz by ds = dz/ cos θ, hence
the factor 1/ cos θ in the calculations in this section.
The total upward irradiance at z is then,

↑
Fλ↑ (z) ≈ πBλ (Ts )e−1.66τ (z) + Fλ,lay (z) (2.48)
where the second term on the r.h.s is the contribution of the atmospheric
lower hemisphere (or layer), to the irradiance. Let’s see how we would esti-
mate it if we were climate modellers, equipped with a (fast) computer. For
a given direction, i.e. a given slanted path, we would calculate the integral
term on the r.h.s of (2.44). Because of the plane parallel approximation,
the result will not depend on the azimuthal direction, but only on z and θ.
Writing it as Φ↑ (z, θ),
Z τ (z)/ cos θ
↑ 0
Φ (z, θ) ≡ B(τ 0 )eτ dτ 0 (2.49)
0
where τ 0 measures the optical depth along the slanted path, the contribution
of the atmospheric layer to the total irradiance would then be,
Z 1
↑
Fλ,lay (z) = 2π Φ↑ (z, µ)e−τ (z)/µ µdµ. (2.50)
0
Downward irradiance
Because of the negligible amount of photons emitted by the Sun at wavelength
≥ 4µm, the “surface term” for downward infrared radiation can be set to zero.
Hence the downward irradiance at a height z only reflects the emission from
the atmospheric column sandwiched between the “top-of-the-atmosphere”
and z. The result would thus simply be the analog of (2.50) for downward
rather than upward integration. We can simply use the previous result and
write, Z 1
Fλ↓ (z) = Fλ,lay
↓
(z) = 2π Φ↓ (z, µ)e−τ (z)/µ µdµ. (2.51)
0
in which τ (z) = τλ (T OA, z) is now the optical depth measured downward
from the “top-of-the-atmosphere”.
An illustration of the calculation is provided using the result in section
2.5.2 for an isothermal atmosphere at T = To . Based on this the result, the
downward irradiance is,
Z 1
↓
Fλ (z) = 2π Bλ (To )(1 − e−τλ (z)/µ )µdµ ≈ πBλ (To )(1 − e−1.66τλ (z) ). (2.52)
0
The contribution of this irradiance to the heating rate can then be estimated
from (2.36) and an illustration is provided in Fig. 2.14 for an idealised water
Figure 2.14: A simple calculation of radiative cooling due to water vapour for
an isothermal atmosphere at T = To = 255K. The specific humidity is taken
as qa = 0.005e−z/3km and the background density is taken as ρ = 1.2e−z/8km .
An extinction coefficient kλ = 0.03m2 /kg was used over the 5µm ≤ λ ≤
50µm. The heating rate (integrated over λ, in K/day) is displayed as a
function of height, as are the optical depth and associated transmissivity.
The Matlab code is available on Blackboard.
vapour profile (i.e., the calculation assumes an exponential qa and a broad

range of wavenumbers –see caption). A mid-tropospheric peak in cooling is
predicted, with a magnitude on the order of 1K/day. The formula can also
be applied for a “CO2 case” by making the mixing ratio qa uniform as a
function of height, by narrowing the range of wavenumbers and increasing
the opacity in that range. As seen in Fig. 2.15, the peak of cooling now moves
upward, with very little cooling below 5km (this is because the irradiance is
“saturated” at these levels and so does not vary with height).
Realistic numerical calculations of upward (red), downward (blue) and the
net (upward-downward, green) are shown in Fig. 2.16. These were carried
out for typical summertime conditions over the Northern Hemisphere (only
three infrared absorbers were included, water vapour, carbon dioxide and
Figure 2.15: Same as Fig. 2.14 but for an idealised CO2 case. The mixing
ratio is now taken as qa = 400ppm and an extinction coefficient kλ = 1m2 /kg
was used over the 12.5µm ≤ λ ≤ 17.5µm. The heating rate (integrated over
λ, in K/day) is displayed as a function of height, as are the optical depth
and associated transmissivity. The Matlab code is available on Blackboard.
Figure 2.16: Calculations of upward (red), downward (blue) and net (green)
irradiances for summertime conditions in the Northern Hemisphere. Cour-
tesy of Wenyi Zhong and Jo Haigh from the SPAT group. The data and the
Matlab code used to generate the plot are available on Blackboard.
ozone). Note that the irradiances are integrated over all wavelengths and are
displayed as a function of pressure. The general shape is that of decreasing
intensity as one goes up in the atmosphere for the upward irradiance: the
upward infrared radiation is absorbed in the atmosphere and is re-emitted
less strongly at upper (lower temperature, lower number density) than at
low levels (high temperature, high number density). Conversely, the general
shape for the downward irradiance is that of increase as one goes to lower
levels: downward infrared is more effectively emitted at low levels where the
temperature and the number density are higher.
Heating rate
The net heating rates will reflect the competition between heating due to ab-
sorption of the radiation emitted by other atmospheric layers and the ground,
and the cooling due to the emission of longwave radiation. Detailed calcula-
tions of the resulting heating rates, integrated over wavelengths, are shown in
Fig. 2.12 (left panel). The first thing to note is that, except for O3 near 25km,
longwave heating rates are negative, on the order of several K/day. Infrared
radiation thus cools the atmosphere globally, which opposes the heating due
to absorption of shortwave radiation (previous subsection). Inspection of

Fig. 2.16 suggests that the cooling reflects the fact that downward irradi-
ance (blue curve) diverges more (i.e., cools atmospheric layers) than upward
irradiance (red curve) converges (i.e., heats atmospheric layers).
The primary reason why net cooling is found in the infrared is that the
radiation that a given atmospheric layer exchanges with the ground and
other layers tend to cancel out, but there is no cancellation when it comes
to exchange of radiation with Space. Restating the arguments first put for-
ward in Chapter 1 (Fig. 1.7), a given atmospheric layer emits upwards and
downwards towards layers above and below, but it also receives radiation
from them. Likewise, it emits towards the Earth’s surface but also receives
upwards radiation from the Earth’s surface. Each of these exchanges approx-
imately cancel out. In addition, the atmospheric layer in question also emits
infrared radiation to Space but it does not receive such radiation in return.
Hence a net cooling is expected, the so-called “cooling to Space” approxi-
mation (clearly visible in Fig.2.16 as the increase of the green curve towards
lower pressure). As Fig. 2.12 shows, this is mostly due to water vapour in
the troposphere, and carbon dioxide in the stratosphere and the mesosphere.
2.6.3 Net radiative heating rates

The compensation between infrared cooling and shortwave heating is imper-
fect in the troposphere, where a net cooling of about 1K/day is suggested in
Fig. 2.12. This is expected since the troposphere receives energy from the
oceans and the land (heat and moisture exchange). Likewise in the meso-
sphere (height above ≈ 50km in Fig. 2.12) a net radiative heating is sug-
gested (this is opposed by molecular diffusion of heat upward at these levels
–the density of air is so small there that molecular diffusion becomes a major
player in the heat budget). In the stratosphere however the compensation
between infrared cooling and shortwave heating is good: the stratosphere is
close to a radiative equilibrium.
One way to visualize this competition is proposed in Fig. 2.17. In the
right panel is shown the downward net shortwave flux: largest at the TOA
and decreasing slighty towards the Earth’s surface as a result of absorption
by atmospheric gases, aerosols and clouds. On the left panel is shown the
net longwave flux, which is upward at all levels: at the TOA it must be
nearly equal in length to the solar flux since the Earth is approximatively in
radiative balance. At the surface its length must be less than the downward
solar flux since the Earth’s surface itself is in equilibrium and cools by surface
evaporation and heat transfer, in addition to emitting infrared radiation. At
a given level, the air cools through radiative exchanges in the infrared (the
Figure 2.17: Schematic of the compensation between longwave cooling and

shortwave heating. The arrows represent the net fluxes, i.e., the sum of
upward and downward radiative fluxes in each frequency band. The fluxes are
shown at the top-of-the-atmosphere (TOA), the tropopause and the Earth’s
surface. For the magnitude of arrows shown, the portion of air above the
tropopause is in radiative equilibrium while the troposphere cools in the net.
This is a good approximation for the troposphere-stratosphere system. See
text for explanations.
upward arrows diverge in the direction of propagation) and heats up through

radiative exchanges in the shortwave (the downward arrows converge in the
direction of propagation).
2.7 Problems
Q1 Calculate the ratio of the solar radiation incident on northward and

southward facing slopes (i.e., the latter faces the Sun in the Northern Hemi-
sphere), each angled α = 20◦ to the horizontal, if the solar zenith angle θS
(the angle between the vertical and the line of sight to the Sun) is: (i) 30◦
(ii) 60◦ . See Fig. 2.18 for a schematic.
2.7. PROBLEMS 51
Figure 2.18: Geometry for Q1.
Q2 Calculate the intensity I of solar radiation at the top-of-the-atmosphere

(TOA), given the irradiance F = 1361W m−2 at zero zenith angle. The radius
of the Sun is Rs = 7 × 108 m while the Earth-Sun distance is d = 1.5 × 1011 m.
Q3 Consider monochromatic radiation passing through a gas with absorption

coefficient βλ = 0.01m2 kg −1 . (i) What fraction of the beam is absorbed in
passing through a layer containing 1kgm−2 of the gas? (ii) What mass per
unit area would the gas layer have to have in order to absorb half the incident
radiation?
Q4 Suppose the gas in the previous question is present with uniform mass
mixing ration q = 10−3 in an atmosphere in hydrostatic equilibrium. Take
surface pressure to be Ps = 1000hP a. (i) Show that the optical depth mea-
sured from the top of the atmosphere is linearly proportional to pressure.
What is the constant of proportionality? (ii) Estimate the pressure of the
level that is one optical depth from the top of the atmosphere.
Q5 Show that, for a beam of monochromatic solar radiation Fλ incident

vertically on an atmosphere in which the mixing ratio of the radiatively

active gas is independent of height and the density decreases exponentially
with height, the heating rate per unit volume is greatest at unit optical depth.
Q6 Show that, for the situation in the previous question, the heating rate
per unit mass is greatest near the top of the atmosphere.
Q7 Now consider infrared radiation traveling upwards (optical depth mea-

sured from Earth’s surface) in an atmosphere where the total optical depth
at the wavelength of the radiation considered is τ∞ = 5. (i) What fraction of
the monochromatic intensity emitted by the ground is absorbed in passing
through the layer of atmosphere extending from optical depth 0.2 to 4.0?
(ii) What fraction of the intensity emitted to space is emitted by the layer
between 0.2 and 4.0 optical depth? You may assume the atmosphere to be
isothermal for this part, and at the same temperature as the Earth’s surface.
Q8 Past exam question (2005, No 3).
(i) Derive an expression for the emission (or effective) temperature of the
Earth as a function of solar constant and albedo.
(ii) Estimate numerically the sensitivity of the effective temperature to fluc-

tuations in albedo. (Solar constant So = 1370W m−2 , planetary albedo
αP = 0.3 and Stefan-Boltzmann constant σ = 5.67 × 10−8 W m−2 K −4 ).
(iii) Over the last 10 years the global mean temperature has risen by about
0.2◦ C. If clouds have an albedo of 0.8 and the surface has an albedo of
0.1, what change in the cloud percentage cover could account for this
temperature rise? State any assumption you make.
(iv) If the transmissivity of clouds increased by 10 % but the cloud cover

stayed the same what would the new equilibrium temperature be?
(v) Why is the observed surface temperature larger than the effective tem-
perature?
(vi) In which direction must the observed net radiation at the surface be
and why?
Q9 Figure 2.19 shows an estimate of the flux of solar radiation impinging on

the Earth (the solar constant So ). Show that it is in good agreement with a
pen and paper prediction assuming only knowledge of the radius of the Sun
2.7. PROBLEMS 53
Figure 2.19: Composite measurement of the Solar “constant” So from Space

through time (in W m−2 ). From Joanna Haigh’s briefing paper published by
the Grantham Institute for Climate Change at Imperial College in 2011.
rsun = 6.96 × 108 m, the Sun’s emission temperature Tsun = 5780K, and the
mean Earth-Sun distance d = 1AU = 1.5 × 1011 m.
Q10 A sunphotometer is an instrument designed to measure the optical

thickness of the atmosphere due to scattering and absorption of solar radia-
tion by air molecules and aerosols. At the ground, two measurements of the
incident solar radiation at a wavelength λ = 1µm, denoted by Iλ , are made.
These are made at two different solar zenith angles θ1 = 20◦ and θ2 = 40◦ .
(i) Show that the ratio rλ = (Iλ )1 /(Iλ )2 of the two intensities satisfy,
ln rλ = τλ (sec θ2 − sec θ1 ) (2.53)

R +∞
in which τλ = 0
kλ ρqa dz is the “column” optical depth.
(ii) The ratio of the two intensity measurements is rλ = 1.12. Compute the
column’s optical depth using the data above, stating any assumptions
made.
Chapter 3
Radiative-convective
equilibrium
Key concepts: radiative equilibrium, fluid parcel, environment, buoyancy,

Brunt-Vaisala frequency, radiative-convective equilibrium.
The preceding chapters showed that the atmosphere cools through (net)
radiative processes (Chapter 2) and is heated from below through surface
evaporation and sensible heat fluxes (Chapter 1). This situation destabilizes
the atmosphere because, at same pressure warm air is lighter than cold air,
and thus rises, while cold air sinks. The effect is to generate a convective
cell which restores the equilibrium by carrying heat upward. In this chapter
we study simple models of this interaction between radiation (destabilizing)
and convection (stabilizing).
3.1 Radiative equilibrium

Mathematically, the state of radiative equilibrium is such that the net ra-
diative heating (i.e., the sum of shortwave and longwave parts) is zero at all
heights. Using the result from Chapter 2, this reads,
Z +∞ Z +∞
d
Qrad = Qλ dλ = (Fλ↓ − Fλ↑ )dλ = 0 (for all z) (3.1)
0 dz 0
Since the longwave emission depends on temperature, there must be a par-

ticular choice of this variable allowing (3.1) to be satisfied. It is referred to
as the “radiative equilibrium” temperature.
Figure 3.1 (left panel) provides an example of a radiative equilibrium
temperature calculation. The input for such calculation is the distribution
55
56 CHAPTER 3. RADIATIVE-CONVECTIVE EQUILIBRIUM
Figure 3.1: Approach to radiative (left) and radiative-convective (right) equi-

librium in a time dependent caculation (Manabe and Strickler, 1964). Each
curve represents temperature (x-axis) vs pressure (y-axis). Dashed curves
refer to a cold start while continuous lines refer to a warm start. The thick
lines indicate the equilibrium temperature profile in each case. In the right
panel, they also locate the region of minimum temperature, the model’s
“tropopause”.
of radiatively active constituents as a function of height, and the incident

solar radiation at the TOA. (note that in this example, clouds were omitted
and water vapour as well as ozone and carbon dioxide distributions were
specified). When starting for example from a cold isothermal state at T =
160K (dashed lines), one observes a systematic warming at all levels as time
increases, reaching a steady state after about 300 days (thick continuous
curve). This warming is pronounced near the surface and at upper levels.
Note that the same equilibrium is reached if one starts from a warm state
T = 360K (continuous lines).
What happens in these experiments is that the shortwave part of the heat-
ing rate is essentially fixed (it does not depend on temperature). The long-
wave cooling, however strongly depends on temperature through Schwarzchild’s
equation (Chapter 2). In the cold start for example, a given layer of air does
not emit enough longwave radiation to offset the shortwave heat gain and,
3.2. CONVECTION 57
as a result, it warms. By warming up, it emits more in the infrared (the

blackbody radiation term in Schwarzchild’ equation) and so the net warming
decreases. When the equilibrium temperature is reached, the layer cools as
much by infrared than it gains heat from shortwave absorption. No further
temperature change can arise.
An interesting feature of Fig. 3.1 is that it displays the time evolution
of the temperature profile. Whether starting from a cold or a warm state,
it takes several months to reach the final equilibrium temperature profile.
This suggests that atmospheric motions such as weather disturbances, con-
vective cloud systems, and others, can be considered as adiabatic since their
timescale is much shorter than a few months.
3.2 Convection
If we were to perturb slightly the radiative temperature equilibrium, say by
allowing a small upward velocity at low level, and wait long enough, would
the same temperature profile be found? In other words, is the radiative tem-
perature equilibrium stable to “dynamical” as opposed to thermal changes?
Unfortunately, the answer is a clear no, and this is the main reason why
weather and climate predictions are difficult to make! The radiative temper-
ature profile is not stable to displacements of the layers: convective motion
will develop and change the final temperature. We first study the basics
of this instability and return to the problem of the radiative temperature
equilibrium later in the chapter.
3.2.1 The “parcel’s equation”

The key concept to understand convection is that of buoyancy whom you
might remember from Archimede’s principle. In Atmospheric sciences this
concept is tied to that of an air parcel, a small enough sample of air that
it describes locally what happens in the atmosphere, but made of a large
enough number of molecules that we can apply thermodynamics to it.
Consider such parcel initially at a reference height zo in the atmosphere.
We ignore horizontal variations. From Newton’s second law, the motion of
the parcel obeys,
dwp X
= Fi (3.2)
dt i
where Fi refers to any force per unit mass acting on the parcel of upward
velocity wp . The forces in question here are simple gravity and the pressure
gradient force,
dwp dP
= −g − αp (3.3)
dt dz
in which αp is the volume per unit mass (or specific volume) of the parcel
and P is the pressure. We wish to establish whether, once displaced initially,
the parcel will come back to zo (stable case) or move away from it (unstable
case).
We will carry out a thorough investigation of all the forces acting on a parcel
of air in chapter 4 (including Coriolis and centrifugal forces, as well as trans-
port of momentum by the flow). As it turns out, the approximation (3.3)
will be shown to be accurate.
The key concept to answer this question is that of the “environment”,
i.e., the atmosphere before we displaced the parcel. In the question at hand
(stability of the radiative equilibrium state), the environment is simply a state
of rest (no motion) with the temperature equal to the radiative equilibrium
temperature. The environment (subscript e) thus satisfies,
dPe
0 = −g − αe (3.4)
dz
Note that this equilibrium is the “hydrostatic balance” (1.8) introduced in
Chapter 1.
To make further predictions regarding the parcel’s fate, we will make two
assumptions:
(i) the pressure field is not modified by the parcel’s motion, so the P in
(3.3) and in (3.4) are equal
(ii) the parcel does not acquire any heat as it moves (consistent with the
short timescale of this process compared to radiative processes, see Fig.
3.1) and the ascent does not involve irreversibilities such as mixing with
surrounding air, friction, etc. As a result the entropy of the parcel is
conserved during the ascent (second law of Thermodynamics).
As a result of (i), we can rewrite (3.3) as,
dwp (αp − αe )
=g (3.5)
dt αe
This equation shows that if the specific volume of the parcel is greater
than that of the environment, the parcel will exhibit upward acceleration
(dwp /dt > 0). Likewise, if the parcel has a smaller specific volume than the
3.2. CONVECTION 59
environment, it will be accelerated downwards. The r.h.s in (3.5) is called

the buoyancy force B,
(αp − αe )
B=g (3.6)
αe
3.2.2 Potential temperature

Up to now the derivation has been fairly general. We restrict from now on
the discussion to the case of dry air so that the pressure P = Pd and the
equation of state for the parcel is simply the ideal gas law Pd αd = Rd T (see
Chapter 1). This allows to rewrite (3.5) as1 ,
dwP (Te − Tp )
= −g (3.7)
dt Te
The goal of this subsection is to introduce a quantity called “potential tem-
perature”, which is going to allow us to solve (3.7) by using the second
assumption above (isentropic ascent).
Potential temperature is simply meteorologist’s jargon for entropy. The
concept is best illustated with the following calculation. Consider an air
parcel at upper levels where its temperature is T and its pressure is P .
Suppose we bring this parcel adiabatically to the surface where the pressure
is Pref = 1000hP a. What would be its new temperature? We’ll call the
latter potential temperature, with the variable θ, to express that it is the
temperature an air parcel at T, P would have if it were brought adiabatically
to the surface.
To find a formula for θ, use that in an adiabatic process, entropy is
conserved. For dry air, this reads,
cp,d dT /T − Rd dP/P = 0 (3.8)
since the specific entropy of dry air is sd = cp,d ln T − Rd ln P (in this ex-
pression cp,d is the specific heat capacity at constant pressure). This can be
integrated from P to Pref and T to θ,
Z T Z P
Rd
dT /T = dP/P, (3.9)
θ cp,d Pref
1
In doing so one must neglect the difference between the pressure of the parcel (Pp ) and
that (Pe ) of the environment: otherwise the ratio of specific volumes αp /αe would include
a term Pp /Pe . It can be shown that this is valid as long as the velocity of the parcel is
less than the speed of sound, which is valid for the atmospheric applications considered
here (i.e., Pp ≈ Pe for the motions considered in this course).
Figure 3.2: Annual and zonal mean distribution of potential temperature

θ as a function of latitude and pressure. Figure obtained from the on-line
ERA40 Atlas.
leading to
−Rd /cp,d
P
θ=T (potential temperature) (3.10)
Pref
For a parcel initially near the tropopause with P = 250hP a, T = 210K, we
find θ = 312K!! So although it is true that the in-situ temperature high
up above our head is colder, bringing these parcels to us would raise their
temperature tremendously because of the work of compression they would
experience during the adiabatic descent.
Figure 3.2 illustrates the zonal mean (i.e., averaging along a latitude
circle) and time mean distribution of θ. Stratospheric air is seen as the
region of the atmosphere where θ is very high (θ ≈ 500K), that is very
buoyant. The downward doming of θ surfaces near low latitudes reflect, at
a given pressure level, the higher temperature in these regions compared to
higher latitudes.
3.2.3 Stability of temperature profiles to vertical dis-

placements of air parcels
Replacing T in (3.7) by θ (note that the pressure terms cancel), we obtain,
dwP (θe − θp )
= −g (3.11)
dt θe
3.2. CONVECTION 61
At first sight, not much is gained. Upon reflection though we see that we
have replaced T , a non conserved variable during the ascent, by θ which is
conserved (you can easily check that ds = 0 implies dθ = 0). In particular,
if we start the parcel at zo with the same temperature as the environment,
θp (zo ) = θe (zo ) while at z, θe (z) = θe (zo ) + dθe /dz(z − zo ). This shows that,
since the parcel conserves its entropy during the ascent, i.e., θp (z) = θp (zo ),
d2 (zp − zo ) g dθe
2
+ (zp − zo ) = 0 (3.12)
dt θe dz
in which we have used wp = dzp /dt, zp being the height of the parcel.
This equation shows that the fate of the parcel is governed by the sign of
the quantity N 2 , defined as,
g dθe
N2 ≡ (the Brunt-Vaisala frequency) (3.13)
θe dz
When N 2 > 0, (3.12) is an harmonic oscillator with angular frequency N :
the parcel undergoes stable oscillations around z = zo . When N 2 < 0, the
parcel displacement is unstable: if initially displaced upwards, it will keep
going upwards, being constantly accelerated upwards by the buoyancy force
(and likewise if initially displaced downwards, the parcel will keep acceler-
ating downwards). The Brunt-Vaisala frequency is clearly a very important
variable describing the atmosphere (see section 3.5).
A couple of important comments:
(i) From Fig. 3.2 we see that the atmosphere has, in the mean, N 2 > 0.
So it is stable to vertical displacements of air parcels.
(ii) From the plot, one can estimate roughly that N = [(9.81/315) × (330 −
300)/5km]1/2 ' 10−2 s−1 . This corresponds to a period of oscillation of
about 10 minutes.
(iii) The fact that the time mean and zonal mean picture has N 2 > 0 does
not mean that the atmosphere is always everywhere, and at all times,
stable. Indeed the deep anvil clouds that one sometimes experiences
on a summer day (Fig. 3.3) are the manifestation of a region of N 2 <
0. This arose because parcels at low levels increased their entropy
tremendously through heating with the hot Earth’s surface while, aloft,
they cooled through infrared radiation. This state of affair generates
high θp at low levels and low θe at upper level and thus a positive
buoyancy for air parcels from (3.11). NB: clearly moisture effects must
be invoked to discuss this aspect fully, and this is postponed to section
3.4.
Figure 3.3: An intense “anvil cloud”. Surface heating throughout the day
increased the entropy of air parcels at low levels while infrared radiative
cooling decreased that of air parcels at upper levels (this latter effect is
admittedly weaker than the heating due to heat exchange with the Earth’s
surface). Note that air parcels seem “stuck” at the top and spread laterally.
This reflects that even though they acquire large buoyancy through surface
heating, this buoyancy was still less than that of the environmental air at
upper levels (the tropopause here). Thus in this example the spreading allows
you to “see” the tropopause!
3.3. RADIATIVE-CONVECTIVE EQUILIBRIUM 63
3.3 Radiative-convective equilibrium

3.3.1 Stability of the radiative equilibrium tempera-
ture profile
We now return to Fig. 3.1 (left column) and the stability of the radiative
temperature equilibrium. At the surface, the temperature is T ≈ 330K with
a pressure P = 1000hP a. Higher up, there is a minimum temperature of
about 180K, reached at a pressure ≈ 300hP a. Application of (3.10) with
Pref = 1000hP a shows that the surface potential temperature is 330K while
at 300hP a it is ≈ 254K. Hence θ decreases with height (N 2 < 0) and the
radiative equilibrium temperature profile is unstable!
What this says is that radiative processes destabilize the atmosphere and
that consideration of radiation only cannot provide a realistic prediction for
the temperature structure of the atmosphere. The right panel in Fig. 3.1
is an attempt to represent the effect of convection, by bringing artificially
the atmosphere into a state with N 2 = 0 (i.e., neutral to vertical displace-
ments of air parcels –see section 3.4.2 below). Independently of the type
of convective parameterization used, we will refer to this state as one of
radiative-convective equilibrium.
The radiative - convective equilibrium state has a weaker change of tem-
perature with height, i.e., a weaker “lapse-rate” Γ in the right compared to
the left panel,
dT
Γ≡− (definition of lapse-rate) (3.14)
dz
It also has colder surface temperature (about 290K), and higher temperature
aloft. This reflects that by carrying high θ parcels upwards and replacing
them by cold θ parcels going downward, convection carries heat upward.
Denoting the associated heating rate per unit volume as Qconv , we extend
the mathematical definition of radiative equilibrium definition in eq. (3.1)
to that of radiative-convective equilibrium:
Qrad + Qconv = 0 (“Radiative-convective” equilibrium) (3.15)
3.3.2 The Tropopause

An interesting feature in Fig. 3.1 is the presence of a clear discontinuity
in the lapse-rate (from being positive to negative) around 300hP a. This
discontinuity is very pronounced in the radiative equilibrium (left panel),
with a sharp transition between the lower (Γ > 0) and upper atmosphere
(Γ < 0). In the radiative - convective equilibrium state, the discontinuity
is still seen, but is less pronounced, the layer between 10 and 25km being
nearly isothermal (Γ ≈ 0).
Returning to Chapter 1 (Fig. 1.4), we identify this discontinuity in lapse
rate as the tropopause. The previous paragraph makes it clear that it arises
primarily as a result of radiative constraints: the absence of significant ab-
sorption of solar radiation below about 10km leads, in pure radiative equi-
librium, to a region of positive lapse-rate (see the simple calculations in Q4
below). Conversely, in a region where there is absorption of solar radiation,
pure radiative equilibrium leads to negative lapse-rate (so that this region
can cool by emitting more infrared radiation at its top than at its bottom).
The “dynamics” (convective adjustment here) merely modulates the exact
value of the height at which this transition happens.
Returning to Fig. 3.1, the fact that the two temperature profiles (left and
right panels) differ below the tropopause is expected from the fact that the
radiative calculation has N 2 < 0 there. Above the tropopause, the radiative
equilibrium temperature is very stable (a region of Γ < 0 has a very large
N 2 ), so no instability is expected and the difference between the two tem-
perature profiles is more surprising. What we are seeing in action here is the
interaction of convection with radiation: changing the tropospheric temper-
ature through convection affects the upward infrared radiation reaching the
stratosphere, and thus perturbs radiative heating rates in this region. The
calculation shows that above ≈ 25km this effect becomes negligible and the
radiative and radiative-convective equilibrium temperature profiles are the
same.
3.4 Dynamical effects of moisture?

3.4.1 Dry and moist adiabatic lapse-rates
We have managed to get a long way by simply considering the buoyancy of
dry air. However, as a buoyant air parcel rises, its internal energy (temper-
ature) decreases because of the work of expansion done by the parcel on the
surrounding air. As a result, phase change can occur. As we shall see, the
latent heat released at this stage increases significantly the buoyancy of the
parcel. Thus, in addition to important radiative effects, moisture also plays
a key role in the dynamics.
To isolate clearly this role we are going to consider the case of cloudy, as
opposed to pure dry, air. We will restrict ourselves to a mixture of dry air,
water vapour and liquid water, the two latter phases being in thermodynamic
equilibrium. This means that the vapour pressure e is solely a function of
3.4. DYNAMICAL EFFECTS OF MOISTURE? 65
temperature, given by the Clausius-Clapeyron equation (Thermodynamcs,

year 2):
deeq ∆s
e = eeq (T ) with = (3.16)
dT p.b ∆α
Note that in this equation, ∆ refers to a change across the phase boundary
(subscript p.b), i.e., ∆s = sv − sl = lv /T (lv being the latent heat of vapor-
ization introduced in Chapter 1) and ∆α = αv − αl ≈ αv (a given mass of
water vapour occupies a much larger volume than the same mass of liquid
water). So we can approximate it as,

deeq lv
≈ (3.17)
dT p.b αv T
Before proceeding, a few comments need to be made:

• The assumption that gas and liquid phases are in equilibrium is very
important. This can occur at the core of a cloud but, in general, we
must acknowledge that the two phases do not coexist in a state of
equilibrium (net evaporation or net condensation occurs). So we are
considering the effect of moisture in a very idealised case (cloudy air).
• The phase change occurring as a parcel ascends is adiabatic. What

is said here is that latent heating is internal to the parcel. It is not
heat taken from the environment, as occurs for example when a cold
continental air mass flows over a warm ocean. So we can still assume
that the parcel conserves its entropy during ascent.
The calculation below adresses the question: what is the lapse rate when
the environment is neutral to vertical displacement of air parcels? That
is, what is Γ when N 2 = 0. In such cases, the buoyancy experienced by the
parcel is zero throughout its displacement –see (3.12)– so that αp = αe . Since
the environment satisfies the hydrostatic balance itself αp ∂P/∂z = −g, and
since the pressure of parcels and environment are the same by assumption,
this means that the parcel also satisfies the hydrostatic balance,
dP
α = −g (3.18)
dz
(in this equation, and from now on, the subscript p has been dropped). To
see how we can use this to predict the lapse-rate, let’s first start with the dry
case. For an adiabatic displacement, ds = 0 leads to
cp,d dT /T − Rd dP/P = 0 (3.19)

After dividing by dz and using the ideal gas law and the hydrostatic balance,
we obtain,
g
Γd = (“dry adiabatic lapse - rate”) (3.20)
cp,d
In the case of moist air, we accept the result that the entropy of cloudy
air is approximately given by,
lv qv
s = sref + cp,d ln(T /Tref ) − Rd ln(Pd /Pref ) + (3.21)
T
in which qv (specific humidity) was introduced in chapter 1 (the appendix
provides a derivation of this expression for s if you’re interested). Using again
the hydrostatic equilibrium (being careful that now the P in the hydrostatic
balance refers to the total pressure P = Pd + e), applying ds = 0 to this
equation leads to2 ,

αd lv dT α g
d T d
1+ =− − (lv qv /T ) (3.22)
αv cp,d T dz α cp,d cp,d dz
The first term on the r.h.s is similar to the case of dry air, the factor in
parenthesis being close to unity. The second term on the r.h.s is new and
clearly reflects the effect of phase change. In a realistic range of temperature,
this term can be approximated as,
T d lv dqv
− (lv qv /T ) ≈ − (3.23)
cp,d dz cp,d dz
Likewise, the second term in parenthesis on the l.h.s is small compared to

unity so that,
g lv dqv
Γm ≈ + (“moist adiabatic lapse - rate”) (3.24)
cp,d cp,d dz
Because specific humidity decreases with height (water vapour condensing

to liquid water in ascending air, liquid water evaporating in descending air),
the second term on the r.h.s is negative so that the moist adiabatic lapse
rate is weaker (typically 6.5K/km) than the dry adiabatic one. Physically,
as a parcel rises and performs work against the environment, its temperature
decreases. In the cloudy air case, the decrease is however not as pronounced
as in the dry air case because of the release of latent heat during condensation.
2
the Appendix provides a general derivation of N 2 which shows that, even for cloudy
air, N 2 = 0 implies ds = 0.
3.4. DYNAMICAL EFFECTS OF MOISTURE? 67
3.4.2 Application of “dry and moist adiabats thinking”

Parameterisation of convective instability
In either the dry or moist case, the state of neutrality to vertical displace-
ment of air parcels is one in which no energy can be extracted from the
“environment”, i.e., air parcels cannot gain kinetic energy through the work
of buoyancy forces (identically zero). This is what motivates the param-
eterisation of convection in Fig. 3.1 (right panel): radiation generates an
unstable temperature profile but convection uses rapidly the energy stored
in that profile to generate motions so that there is never any “left over” en-
ergy available. Setting Γ to be equal to Γm in the calculation in Fig. 3.1
(right panel), and thus imposing N 2 = 0 throughout the calculation, is a
very powerful and simple way to represent this state of affair.
Conditional instability
The calculations above for a dry and a cloudy parcel are exact when the
environment is neutral (N 2 = 0). When N 2 6= 0, they would have to be
modified by replacing g by gαp /αe , since the parcel is now experiencing a
buoyancy force3 . This correction is small. For example, for dry air, αp /αe =
θp /θe is at most 300 ± 10K/300K, i.e., a 3% error. Thus in practice, we can
consider that the temperature lapse rate of a dry parcel during its adiabatic
displacement follows approximatively Γd . Likewise for a moist parcel and
Γm . This is very useful, as the exemple below shows.
Consider the temperature profile in Fig. 3.4 (black), whose lapse rate Γ
is somewhere in between dry and moist adiabats (i.e., Γm ≤ Γ ≤ Γd ). If we
were to lift an air parcel from the surface upward (for example as a result
of a large scale convergence of air near the surface), would it continue to go
up? Here we take into account moisture and do not assume that the parcel is
already in equilibrium of phases (i.e., the relative humidity is less than unity
at the surface).
Starting at the surface at A, we displace the parcel upward following a
dry adiabat since no liquid water is yet present4 . For the case shown, this dis-
placement is stable since the temperature decreases faster with height along
a dry adiabat (blue) than it does for the environment (black). Nevertheless,
3
This is because we can only use αp ∂P/∂z = −g when αp = αe since it is only the
environment which is in hydrostatic balance when N 2 6= 0.
4
Some of you might rightly question the use of a dry adiabat for the isentropic ascent
of a mixture of dry air and water vapour considered here. Doing the calculation properly
would only change the lapse rate slightly but would require using an exotic quantity called
“virtual potential temperature”. For simplicity I decided not to do this.
Figure 3.4: An idealized temperature profile (black) with dry (blue) and
moist (green) adiabats superimposed. The LCL is the “lifting condensation
level” where the temperature equals the dewpoint temperature (i.e., where
the first drop of liquid water is produced by condensation). The LFC is the
“level of free convection”, beyond which the parcel’s displacement would be-
come unstable. In the lower panel, the trajectory of the sample is shown in
the temperature (T ) - vapour partial pressure (e) space. The displacement
from A to B is on the vapour side of the phase diagram, and follows approx-
imatively a line of constant e. From then on, the displacement is along the
phase boundary (purple curve).
3.5. RADIATIVE-CONVECTIVE EQUILIBRIUM AND THE REAL WORLD69
at some point the temperature of the parcel has dropped enough that con-
densation occurs. This happens at a temperature Td (dewpoint temperature)
such that:
eeq (Td ) = e (3.25)
in which eeq is the equilibrium vapor pressure (a function of temperature only,
given by integration of the Clausius-Clapeyron equation –see eq. (3.16) in
section 3.4.1) and e the vapor pressure in the sample. The level at which this
occurs is called the “lifting condensation level” (point B in the figure). From
this point onward our sample is in thermodynamic equilibrium of phases (i.e.,
the relative humidity is unity) so further adiabatic lifting will proceed along
the moist adiabat (green curve). For the case shown, the parcel is still less
buoyant than the surrounding from B to C, the latter being the point where
the parcel’s temperature equals that of the environment. Thus the deplace-
ment is overall stable from A to C. Beyond C however, the displacement
is unstable, since the parcel’s temperature exceeds that of the environment
(and its pressure equals that of the environment).
In summary, if the parcel has enough kinetic energy when it starts at the
surface, it could reach C at which point buoyancy forces will accelerate it
further upward. This type of instability is called “conditional instability”.
This situation occurs frequently in the atmosphere, and the LCL is readily
identified as the cloud base. Deep convective clouds are those where enough
energy was available at low levels to reach the LFC. We’ll return to these
ideas in the lecture on clouds (powerpoint file on Blackboard).
3.5 Radiative-convective equilibrium and the

real world
The main message from all the above is that there is a competition between
radiative processes, which tend to destabilize the atmosphere, and dynamical
processes, which tend to stabilize the atmosphere. The 1D case illustrated in
Fig. 3.1 showed that radiation acts relatively slowly (it took several months
to achieve radiative equilibrium on the left panel). Convection is expected to
be much faster (a moderate updraft of 10cm/s would reach the tropopause in
typically 10km/10cms−1 ≈ 1day). As a result, the prediction is that convec-
tion should dominate and we do not expect to observe unstable temperature
profiles.
Figure 3.5 displays the observed ratio (Γm − Γ)/Γm . It is clearly seen that
in the Tropics (equatorward of 30◦ of latitude), and above the near surface
layer (P ≤ 900hP a), observed lapse rates are within 10 % of the moist adi-
Figure 3.5: Observed tropospheric lapse rate, expressed as percentage of the

moist adiabatic value (i.e., (Γm − Γ)/Γm ). From Stone and Carlson (1979).
abatic value. (They are slightly greater than Γm below 500hP a and slightly
smaller above this level which reflects the curvature of the moist adiabats
and the more linear temperature variations –see the ppt slide discussed in
the lecture for this section). This is suggestive of a strong role of convec-
tive processes at low latitudes and has deep implications for a variety of
problems. For example, it suggests that any change in surface temperature
will be quickly felt throughout the atmosphere through adjustment to a new
(warmer) moist adiabatic profile. This is indeed one of the prediction of the
Intergovernmental Panel on Climate Change (IPCC) to which we will come
back later in the course. Another interesting idea is that since observations
suggest Γ ' Γm in the Tropics, this could be used to parameterize convection
in climate models (these do not resolve individual clouds, and not even their
aggregation into clusters), like was done in Fig. 3.1. This is what all climate
models do...but you’ll be surprised (interested?) to know that there is to
present day no theory which explains why it should be so.
In mid and high latitudes, observed lapse rates are weaker than moist adi-
abatic values (Γ ≤ Γm ), i.e., atmospheric motions stabilize radiative effects
beyond what is expected from a consideration of their tropical counterparts.
Thus in addition to the vertical convection discussed above, something else
must be stabilizing the atmosphere. We all know this something else: the
weather systems we experience daily at the latitude of England. We show
next that these motions can also be thought of as resulting from a convective
3.6. SLOPING CONVECTION 71
instability, but with 3D motion of air parcels rather than 1D (purely vertical)
being involved. We’ll go deeper in their dynamics in Chapter 4.
3.6 Sloping convection

To explain why the observed lapse rates are significantly smaller than moist
adiabatic in mid and high latitudes compared to low latitudes (Fig. 3.5), we
start by looking at the distribution of potential temperature θ (Fig. 3.2). The
latter shows a striking distinction between the Tropics and higher latitudes.
To within a few tens of degree of latitude off the equator, θ-surfaces are
nearly flat while they slope upward in the extra-tropics. This distribution
reflects the presence of marked horizontal temperature gradients away from
the equator (we will see in Chapter 4 that this difference in slopes of θ-
surfaces between low and high latitudes arises because of the effect of rotation
on motions). By consideration of the energetics associated with parcels’
displacements in the latitude-height plane, rather than purely in the vertical,
we can get a sense as to why storms help to create larger values of N 2 in
midlatitudes. As it turns out the basic physics is well captured by ignoring
the effect of moisture on buoyancy, so we will use the framework in section
3.2.2.
Starting from the “parcel’s equation” (3.5) in the form,
dwp (θp − θe )
=g (3.26)
dt θe
we can say something about displacements of air parcel’s in the latitude-
height plane (Fig. 3.6). For example, imagine swapping point A with either
point B, C or D. We note that:
• A ↔ B: this is the situation discussed in section 3.2 when the Brunt-
Vaisala frequency is positive (θ increases upward, θ(A) < θ(B)). Thus
work would have to be done against the buoyancy force to achieve this
displacement. This is a stable situation.
• A ↔ C: this displacement is along a θ surface so the parcels experience
no buoyancy force (θe = θp ).
• A ↔ D: this time θ(A) > θ(D) so the displacement would be unstable,
i.e., the buoyancy force would do work on the parcels.
• A ↔ E: again θ(A) > θ(E) but no work by the buoyancy force can
be generated since the displacement is horizontal (i.e., at right angle to
the force). This situation will thus not occur spontaneously.
Figure 3.6: Sloping convection in the latitude (y) - height (z) plane. Kinetic
energy can be gained in certain swapping of parcels while they conserve their
potential temperature θ.
This suggests that if the displacement of fluid parcels follows a sloping path
shallower than the isentropic slope, kinetic energy can be gained in the pro-
cess: an instability occurs.
NB: the paths shown in Fig. 3.6 are imposed, and all we do is compute
the work done by buoyancy forces for a given path. You can easily convince
yourself that the work done must be path dependent(e.g., it is zero along A-C
but > 0 along A-E-C with same start and end points).
The gain in kinetic energy after a time t has elapsed since the swap can
be estimated from the parcel’s equation, having in mind θp = θ(A),
Z t Z t Z zD
1 2 dwp (θp − θe ) (θp − θe )
∆ wp = wp dt = g wp dt = g dz (3.27)
2 0 dt 0 θe zA θe
where the integral is along the slanted path A ↔ D. Define ∆z = zD − zA
and rewrite the integral as,
1 2 (θp − θe )
∆ wp = g∆z < > (3.28)
2 θe
where the bracket <> denotes an average along the path. Approximate the
latter using a Taylor expansion, as,
(θp − θe ) 1 ∂θ ∂θ
< >≈ − ( ∆z + ∆y)/θe (3.29)
θe 2 ∂z ∂y
3.6. SLOPING CONVECTION 73
in which ∆y = y(D) − y(A) > 0 is the gain in latitude of the parcel and
we have introduced the vertical (∂θ/∂z > 0) and meridional (∂θ/∂y < 0)
temperature gradients. Note that the implicit assumption used here is that
the parcel conserves its θ from A to D (θp − θe = 0 at A while θe − θp =
∆z∂θ/∂z + ∆y∂θ/∂y at D, hence the factor 1/2 and the negative sign).
Inserting this result, and after introducing the slope of the displacement
µ = ∆z/∆y and the isentropic slope µθ = −(∂θ/∂y)/(∂θ/∂z) > 0, we have,
1 2 N 2 (∆y)2
∆ w ≈ µ(µθ − µ) (3.30)
2 p 2
For a fixed displacement ∆y, we recover the discussion at the beginning of
this section:
• A ↔ B: µ > µθ and the kinetic energy gained is negative!
• A ↔ C: µ = µθ and the kinetic energy gained is zero.
• A ↔ E: µ = 0 and the kinetic energy gained is zero.
• A ↔ D: 0 < µ < µθ and the kinetic energy gained is positive. The

maximum possible kinetic energy gained is when µ = µθ /2,
2
N 2 (∆y)2 2 g(∆y)2

∂θ ∂θ
KEmax = µθ = / (3.31)
8 8θe ∂y ∂z
Not all air parcels can go on average upward and poleward –otherwise
mass would be continuously lost at low latitudes and low levels. So we should
think of a swap: parcel A replacing parcel D and parcel D replacing parcel A.
What would be the kinetic energy gained in the displacement D → A? This
can be computed from the previous result by simply making the substitution
∆y → −∆y and ∆z → −∆z. As you can check, it yields the exact same
result. This is really what happens in midlatitude weather systems, a swap
of low θ, low level, low latitude air with high θ, high latitude, high level air.
This takes the form of a wave, as schematized in Fig. 3.7. Interestingly, eq.
(3.31) indicates that the larger the meridional scale of the swap (∆y), the
larger the kinetic energy released. So it is not a surprise that midlatitude
storms are so big and span such a large range of latitudes (typically 20◦ −60◦ ).
The bottom line is that we can understand the extra-tropical storms in

Fig. 1.5 (the “wavy bits”) in the same way as we understand their tropical
counterparts (the “spotty bits”). They result from a convective instability,
“sloped”, as opposed to vertical. The distortion of θ surfaces brought about
Figure 3.7: Schematic of an Eady wave, after the Imperial College physicist
Eric Eady who came up with the simplest theory for weather systems in the
1950s. The swap of two air parcels in a plane making an angle δ with the
horizontal is shown. Eady reasoned that the wave could grow if δ ≈ µθ /2.
3.7. SUMMARY 75
Figure 3.8: Schematic of how the swap of air parcels in sloping convection
leads to an increase in the Brunt-Vaisala frequency N in eq. (3.13) in mid-
latitudes (column of air above A).
by the swapping of air parcels at a shallow angle (µ ≤ µθ ) lead to an en-

hancement of the local Brunt-Vaisala frequency in the extra-tropics, as seen
in Fig. 3.5. This effect is illustrated in Fig. 3.8.
3.7 Summary
Figure 3.9 offers a graphical summary of the various ideas discussed in this
chapter:
• Above the tropopause, the atmosphere can be thought to be in radiative

equilibrium: the sum of shortwave and longwave irradiances (grey and
black arrows, respectively), does not converge or diverge in the vertical.
Radiative heating in the shortwave balances exactly radiative cooling
in the longwave: Qrad,SW + Qrad,LW = 0.
• Below the tropopause, radiative effects alone lead to a strong lapse-rate.

Displacements of air parcels either purely upward (section 3.2) or in a
slanted way (section 3.6) are unstable and convective motions develop.
The troposphere as a whole (Tropics, midlatitudes and high-latitudes)
can be thought of being in a state of convection.
• Convection carries heat upwards (blue arrows, converging upward at

a rate Qconv from the surface to the tropopause). This transfer cools
the surface and heat the upper levels, leading to a weaker (but still
positive) lapse-rate.
• Convective heating is opposed by a net radiative cooling, seen in the

vertical divergence of the net longwave irradiances (black arrows): Qrad,SW +
Qrad,LW + Qconv = 0.
3.8 Problems
Q1. In this question we treat the Earth’s atmosphere as dry and in hydro-
static equilibrium.
(i) Show that the adiabatic lapse rate is simply Γ = Γd = g/cp,d where
cp,d = 1005Jkg −1 K −1 is the specific heat capacity of dry air at constant
pressure and g = 9.81ms−2 is gravity. Compute its numerical value in
K/km.
T ∂θ
(ii) Show that the actual lapse rate is Γ = Γd − θ ∂z
.
(iii) In light of the result in (ii) discuss whether a dry atmosphere can have
a lapse rate greater (or lower) than Γd .
Q2. Air flows over the ocean (so it is well-supplied with moisture), across
the coast, over a mountain which is 4 km high and down to a plateau on
the far side at 1 km above sea level. Estimate the difference in temperature
between sea level on the windward side and at the surface of the plateau.
[Take Γd = 10K/km and Γm = 6.5K/km].
Q3. Relative humidity (RH) is defined as the ratio of the vapour pressure
found in a sample of air at temperature T to that found when vapour and
liquid water are in thermodynamic equilibrium at the same temperature:
RH = e/eeq (T ). Discuss whether the situations below correspond to a state
of thermodynamic equilibrium: (i) Room temperature, RH = 0.7 (ii) pre-
cipitation falling into a dry air mass, RH = 0.5 (as can happen on the edge
of a cloud).
Q4. We study a simple model of radiative equilibrium (Fig. 3.10) by treating

the atmosphere as a two-layer system. Each of these layers is assumed to
3.8. PROBLEMS 77
Figure 3.9: A summary schematic of the radiative - convective equilibrum

view of the atmosphere. Black upwards arrows denote net (i.e., upward
minus downward) longwave irradiances while grey downward arrows denote
net (downward minus upward) shortwave irradiances (see Chapter 2). Blue
upwards arrows indicate the vertical heat flux driven by convective motions.
Figure 3.10: Simple radiative equilibrium model. The surface receives solar
radiation (σTe4 ≡ So (1 − αP )/4) and emits infrared radiation upward. The
atmosphere consists of two layers of longwave emissivity unity and shortwave
absorptivity zero.
radiate like a blackbody in the infrared. The Earth’s surface is also treated
like a blackbody. The atmosphere is assumed to be completely transparent to
solar radiation, so that at each level, the solar flux is So (1 − αP )/4. Compute
the radiative equilibrium temperatures Ts (surface), T1 (lower atmosphere)
and T2 (upper atmosphere).
Q5. The temperature profile predicted in the previous question is not stable
to vertical displacements of air parcels. The latter lead to convective motions
and a new temperature distribution close to an adiabat. To represent this
simply, we fix the temperature difference ∆T = T1 − T2 = Ts − T1 in the
previous model, and also acknowledge the presence of additional heating
terms: a surface heat exchange Fs and a convective heat flux Fc (Fig. 3.11).
(i) Taking the two atmospheric layers to be at a pressure of 700hP a and

400hP a respectively, and using your answers in Q3, check that indeed
the radiative equilibrium temperature profile is unstable. You may take
the surface pressure to be 1000hP a.
(ii) Before doing any calculation, do you expect the surface temperature
to decrease or increase when the fluxes Fs and Fc are included in the
model?
3.8. PROBLEMS 79
Figure 3.11: Addition of convective fluxes, Fs and Fc to the simple radiative

equilibrium model.
(iii) Solve for the temperature Ts , T1 and T2 as a function of Te and ∆T .
(iv) Solve for the convective flux Fc and the surface heat exchange Fs . Ex-
press your answer as a function of Te and the non dimensional param-
eter x = ∆T /Te .
(v) Find a plausible value for x and discuss the range of values of this
parameter for which the model makes sense. Also check that for this
range of values the model agrees with your answer in (ii).
Q6. Using Figs. 1.5 and 3.2, estimate the kinetic energy gain in sloping
convection –eq. (3.31). How does this compare with your guess for the
kinetic energy in storms?
Q7. A parcel of air has the same potential temperature than the environment
at t = 0 and z = 0, and an upward velocity wo = 10cm/s. Find its height
after 1mn if (i) N 2 = 10−4 s−2 (ii) N 2 = −10−4 s−2 . (iii) N 2 = 0. Neglect the
effects of moisture on the motion.
Q8. The temperature structure in Fig. 1.3 in the mesosphere is very strange.
In the Summer Hemisphere, the mesopause has T ≤ −100◦ C while it is about
70◦ C warmer in the Winter Hemisphere!
(i) Estimate the mesospheric lapse rate (i.e., from stratopause to mesopause)
in the summer and winter hemispheres
(ii) Sketch how θ(z) looks like in the summer and winter hemispheres
(iii) Show that the profiles you have plotted in (ii) are consistent with the
Brewer-Dobson circulation sketched in the ppt slides for Chapter 1.
Chapter 4
Atmospheric motions
Key concepts: Eulerian and Lagrangian descriptions of motions, mate-

rial derivative, Coriolis force, Rossby number, geostrophic balance, “thermal
wind”, rotational and divergent flows, vorticity, Rossby waves.
The vertical instability associated with N 2 < 0 in Chapter 3 is at the

heart of convective cells and cloud clusters with horizontal scales ranging
from a few km up to 100km. Updrafts and downdrafts in such systems
are however not the sole motions driven by radiative processes. Indeed, the
net radiative loss at high latitudes and the net radiative heat gain at the
equator set up a whole range of motions. On scales of ≈ 1000km, we find
the traveling weather systems familiar at our latitudes, and their embedded
cold and warm fronts (≈ 100km). On even larger scale (10, 000km) we find
planetary cells organized in east - west (Walker cell) and north-south (Hadley
cell) directions –see Fig. 4.1 for a summary. In this chapter we study these
motions from first principles (Newton’s second law in a rotating frame of
reference). As we shall see, the Earth’s rotation has a profound impact on
the dynamics, giving a (counter-intuitive) rigid or solid - like behavior to
atmospheric motions on large scales.
4.1 Equations of motions

These are simply the three components of the “momentum” equation (New-
ton’s second law). Compared to the classical mechanics you have dealt with
so far, the difficulties and novelty lie in:
(i) the “fluid nature” of the motion, air parcels transporting their own
momentum (material derivative)
81
82 CHAPTER 4. ATMOSPHERIC MOTIONS
Figure 4.1: Space and time scales of atmospheric motions.
(ii) the Earth rotation, which introduces new forces as seen by an observer
at a fixed location on the planet. The effects of rotation on motions
are huge as is most spectacularly illustrated with the “Taylor columns”
(see Youtube video and the paper by Taylor (1923) on Blackboard).
(iii) the sphericity of the Earth, which introduces mathematical complexity

and a “channel-like” geometry to the study of atmospheric motions.
4.1.1 Forces acting on a parcel of air

Per unit mass, these are simply: the acceleration due to gravity (kept to
a constant g = 9.81ms−2 for practical purposes, owing to the thinness of
the atmosphere compared to the Earth’s radius), the pressure gradient force
we’ve seen in Chapter 3 (but now in 3D), and frictional (or viscous) forces,
a = g − α∇P + F f ric (4.1)
Note that in this equation a is the acceleration vector in an inertial frame

(say from an observer looking at the Earth from deep space), and α is the
volume per unit mass used extensively in Chapter 3.
4.1.2 Material derivative

Imagine a parcel of air undergoing an adiabatic ascent. Following the parcel,
there is no change in its entropy. At a fixed location, however, an observer
4.1. EQUATIONS OF MOTIONS 83
might see a change in entropy since the parcel might originate from a warmer
region than that in which the observer is sitting. To make it clear that the
change “following the parcel” is a special case of derivative, we will write it
as D/Dt. In the previous example, we would thus write,
Ds
=0 (4.2)
Dt
Mathematically, the change following the parcel (also called “Lagrangian
derivative”) includes the parcel’s displacement. If we write that entropy is
a function of space and time, i.e., s = s(x(t), y(t), z(t), t), a small change in
entropy δs in a small time interval δt is, after Taylor expansion,

∂s ∂s ∂s ∂s
δs = δt + δx + δy + δz (4.3)
∂t ∂x ∂y ∂z
Dividing by δt and taking the limit, we thus identify D/Dt as,
Ds δs ∂s ∂s ∂s ∂s
≡ lim = +u +v +w (4.4)
Dt δt→0 δt ∂t ∂x ∂y ∂z
Note that we have used the fact that the velocity components satisfy u =
limδt→0 δx/δt, v = limδt→0 δy/δt and w = limδt→0 δz/δt.
The local change with time can then be mathematically identified with,
∂s Ds
= − u.∇s (4.5)
∂t Dt
in which we have introduced the velocity vector u = (u, v, w). This equation
states that the local change in entropy results from the change following
the parcel minus that due to advection. The view encapsulated in (4.5) is
referred to as “Eulerian” (fixed location), as opposed to the “Lagrangian”
(following a parcel) view in (4.2).
We used entropy in the previous example, but the concept applies to
any variable, and even vector (applying the material derivative to each of its
component). For example, for the velocity vector,
Dr
u= (4.6)
Dt
(You can readily check that this expression provides the identity u = u∂x/∂x =
u, etc... since the coordinates x, y, z, t are independent).
Applying the notation to Newton’s second law, we obtain:
Du δu
a = g − α∇P + F f ric = = limδt→0 (4.7)
Dt δt
This is the Navier-Stokes equation for fluid flow in an inertial frame of ref-
erence. What is meant here is that the momentum of a fluid parcel changes
along its trajectory due to the gain / loss associated with gravitational, fric-
tional and pressure forces. Note that the Du/Dt is a vector with components
(Du/Dt, Dv/Dt, Dw/Dt). Each of those are non linear expressions, for ex-
ample, for the third component of momentum,
Dw ∂w ∂w ∂w ∂w
= +u +v +w (4.8)
Dt ∂t ∂x ∂y ∂z
showing that the upward motion carries its own upward momentum and that
the latter is also carried by the horizontal flow.
4.1.3 Rotating frame of reference

The choice of the Earth as a frame of reference is natural. However, this is
not an inertial frame because it is accelerating (rotating) with respect to a
coordinate system fixed in space. Newton’s law can be applied in the non-
inertial frame if the acceleration of the coordinates is taken into account,
introducing “apparent forces” in the equation of motion.
To introduce those, we need to relate the change of momentum of a parcel
in the inertial frame1 , denoted by (DuI /Dt)I to that measured in the rotating
frame of the Earth, denoted by (DuR /Dt)R . In this notation, uI and uR refer
to the velocity of a parcel in the inertial and rotating frames, respectively,
while (D./Dt)I and (D./Dt)R denote the change following a parcel in the
inertial and reference frames, respectively. Two important points need to be
emphasized:
• we are talking about the same parcel of air here and we just describe
differently its momentum “history” (see Fig. 4.2) by measuring changes
as (DuI /Dt)I or (DuR /Dt)R .
• such change in momentum history only arises because momentum is a

vector. In other words, in general

DA DA
6= (4.9)
Dt I Dt R
1
We will consider that a coordinate system with the Earth’s centre as its origin (orbiting
around the Sun) provides such inertial frame. This is the idea introduced in Year 1 (Classi-
cal mechanics) that a free falling object under the action of gravity behaves approximately
like an inertial frame.
Figure 4.2: A simple apparatus (left) to illustrate how the trajectory of

an object appears different in an inertial and a rotating frame (right) of
reference. In the inertial frame, a small ball thrown at point O on a parabolic
dish moves forward and backward (black trajectory on the right). Viewed in
the rotating frame of the dish, it describes a circle (blue trajectory). Friction
has been ignored. Picture taken from Wallace and Hobbs’ textbook.
for any vector A (simply think about a constant vector in a rotating

frame: it will have a zero value of D/Dt in the rotating frame, but a
non zero value in the inertial frame), but

DB DB
= (4.10)
Dt I Dt R
is always true for any scalar B since we follow the same parcel (think
about measuring temperature of the parcel in Fig. 4.2: it does not
matter whether we describe its trajectory as a circle or a straight line,
it is still the same temperature we measure since it is the same parcel).
We accept the result (proven in the Appendix) that, for any vector A,
we have:
DA DA
= +Ω×A (4.11)
Dt I Dt R
in which Ω is the Earth’s rotation vector. Applying (4.11) to the vector
position A = r (measured from the centre of the Earth), one recovers the
Galilean transformation of velocities between reference frames,
uI = uR + Ω × r (4.12)
Applying it to A = uI provides,

DuI DuI
= + Ω × uI (4.13)
Dt I Dt R
which, after using (4.12), reads,

DuI DuR
= + 2Ω × uR + Ω × (Ω × r) (4.14)
Dt I Dt R
(to obtain this result we acknowledge that (DΩ/Dt)R represents the change
in the vector Ω at the location of the parcel, as seen in the rotating frame.
At any given point along the trajectory, Ω is always the same since it is not
affected by the rotation effect encapsulated in (4.11). Thus (DΩ/Dt)R = 0.)
Rearranging, and using (4.7), we finally obtain,

DuR
= g − α∇P + F f ric − 2Ω × uR − Ω × (Ω × r) (4.15)
Dt R
4.1.4 Coriolis and centrifugal forces

Comparing (4.15) with (4.7), one sees two new forces on the r.h.s, namely
the Coriolis and centrifugal forces. The latter only depends on the position
of the parcel (like a conservative force), and in effect acts to reduce gravity,
g 0 ≡ g − Ω × (Ω × r) (4.16)
After noticing that −Ω × (Ω × r) = Ω2 r H = ∇(Ω2 rH 2

/2) where r H is the
component of r perpendicular to Ω, it is sometimes convenient to introduce
a slightly modified gravitational potential Φ such that,
g 0 = −∇(Φgravi − Ω2 rH
2
/2) ≡ −∇Φ (4.17)
We will in the following approximate surfaces of constant (net) gravitational

potential Φ as spheres, with the net gravity given by g 0 .
The Coriolis force (−2Ω×uR ) is required to explain why an object moving
in a straight line in an inertial frame appears to have a curved path in a
rotating frame. (As will be apparent throughout this chapter, there is more
to it than this.) For now, let’s just analyze its effect on the momentum
equation (4.15). Adopting the coordinate system in Fig. 4.3 (see also the
Figure 4.3: Local coordinate system (x, y, z). Longitude is denoted by λ,

latitude by φ.
sidenote below), uR = (u, v, w) and Ω = Ω(0, cos φ, sin φ) in which φ is

latitude and the three axes are i (West to East), j (South to North) and k
(anti-parallel with g 0 ). Thus the Coriolis force will have the components,
−2Ω × uR = −2Ω[(w cos φ − v sin φ)i + u sin φj − u cos φk] (4.18)
Thus a parcel of air with a purely eastward velocity (u > 0, v = w = 0)
will be accelerated southward in the Northern hemisphere (−u sin φ < 0)
and northward in the Southern Hemisphere. The general rule is accelerated
to the right of the motion in the Northern Hemisphere, and to the left in
the Southern Hemisphere. You will also notice an upward acceleration, but
this is pretty small in comparison to gravity. We will proceed to a more
systematic analysis of the magnitude of each term in (4.15) in section 4.2.
Technical sidenote: local coordinate system. The (x, y, z) coordinate system
in Fig. 4.3 is local and cartesian. As such, it has some advantages compared
to that of spherical geometry. Specifically,
uR = ui + vj + wk (4.19)
with u = r cos φ Dλ
Dt
, v = r Dφ
Dt
, w = Dr
Dt
. Here r = R + z in which R is the
(mean) Earth’s radius, and since z R for practical purposes, one can
further simplify and use,
Dλ Dφ Dz
u ≈ R cos φ ,v ≈ R ,w = (4.20)
Dt Dt Dt
This system of coordinate however has the complication that the i, j, k vectors
vary with location, and as a result, (DuR /Dt)R 6= i(Du/Dt)R +j(Dv/Dt)R +
k(Dw/Dt)R . Rather,

DuR Du Dv Dw Di Dj Dk
=i +j +k +u +v +w
Dt R Dt R Dt R Dt R Dt R Dt R Dt R
(4.21)
We will accept the result that,

Di u tan φ u
= j− k (4.22)
Dt R R R

Dj u tan φ v
=− i− k (4.23)
Dt R R R

Dk u v
= i+ j (4.24)
Dt R R R
As a result,
u2 tan φ vw u2 + v 2

DuR Du uv tan φ uw Dv Dw
=[ − + ]i+[ + + ]j+[ − ]k
Dt R Dt R R R Dt R R R Dt R R
(4.25)
As can be easily checked the extra-terms purely reflect the spherical geometry
of the Earth and disappear in the limit R → ∞.
4.1.5 Mass conservation

Ignoring the loss of mass when it rains, and its gain when evaporation occurs
above the Earth’s surface, the conservation of mass can be written as
∂ρ
+ ∇.(ρuR ) = 0 (4.26)
∂t
in which ρ = 1/α is the density of an air parcel. This equation is sometimes
called the “continuity equation”.
One interesting feature of this equation is that it is invariant to a change
of reference frame. Indeed, mass conservation in the inertial frame would
read exactly the same with uR being replaced by uI (you can check this
by yourself readily after noticing that solid body rotation is a non divergent
motion). This is very unlike momentum conservation which, as we have just
seen, requires a different formulation in the inertial and rotating frames. This
is our first hint that momentum might not be the best variable to consider in
rotating fluids. As we shall see later in this chapter, the curl of the velocity
field is much better suited (and its conservation equation is also invariant
under change of reference frame).
4.2. SCALE ANALYSIS OF THE EQUATION OF MOTIONS 89
4.2 Scale analysis of the equation of motions

The momentum equation (4.15) is quite complicated and we would not go
very far if we were not able to simplify it further. To do so we are going to
use a technique similar to dimensional analysis in which we are going to put
orders of magnitude on each terms in (4.15). We will ignore the effects of
friction and focus here on scales of motions typical of those seen on weather
charts, i.e. the sloping convection motions discussed at the end of Chapter
3:
Lengthscale (horizontal) L ' 106 m (a thousand kilometers)

Lengthscale (vertical) H ' 104 m (thickness of the troposphere)
Velocity (u, v) U ' 10ms−1
Velocity (w) W ' 10−2 ms−1
Time (t) T = L/U ' 105 s (about one day)
Horizontal Pressure gradient ∇L P ' 10hP a/1000km = 10−3 P am−1
Remember also that R = 6371km and 2Ω = 4π/1day ≈ 2 × 10−4 s−1 .
NB: The scales for L, H, U, T and ∇L P are relatively intuitive (say from your
own estimate of windspeed looking at cloud displacement and a casual in-
spection of weather charts). It is more difficult to estimate from observations
what the scale for the vertical velocity W should be. One way to proceed is
to use the continuity equation (4.26) and require that the horizontal and ver-
tical derivatives contribute as much to the total divergence. You can easily
check that this leads to W = HU/L ≈ 0.1ms−1 . This estimate is an order
of magnitude larger than the one used above. The reason is that, we shall
soon see, weather systems are in geostrophic balance with very little vertical
motion (nearly non divergent motions).
4.2.1 Vertical momentum equation

The vertical component of (4.15) is, using (4.25),
u2 + v 2

Dw ∂P
= + 2Ωu cos φ − α −g (4.27)
Dt R R ∂z
in which we have dropped g 0 for g and have ignored the friction force.
The first two terms on the r.h.s scale respectively as, U 2 /R and 2ΩU cos φ.
Their strength relative to the acceleration of gravity are U 2 /gR ' 10−6 and
2ΩU cos φ/g ≈ 2 × 10−4 cos φ, respectively. The l.h.s has several components
since,
Dw ∂w ∂w ∂w ∂w
= +u +v +w (4.28)
Dt R ∂t ∂x ∂y ∂z
The first three have the same scale (since we have chosen T = L/U ), namely
W/T , while the last term scales as W 2 /H. The ratio of these to gravity is
thus W/gT ≈ 10−8 and W 2 /gH = 10−9 , respectively. Note that because of
the values listed above, U/L ≈ 10W/H so from now on we will simply scale
all D/Dt terms by simply dividing by T (i.e., (Dw/Dt)R ≈ W/T ).
The bottom line is that all the terms considered are negligible compared
to the acceleration of gravity and so, to a very good approximation, we can
safely use,
∂P
0 ≈ −α −g (4.29)
∂z
This is the “hydrostatic equation” introduced earlier in Chapter 3, expressing
a near cancellation between the acceleration due to gravity and that due to
the vertical pressure gradient force.
To some extent the above scaling is misleading. It makes sense that grav-
ity opposes the vertical pressure gradient, but the motions we are considering
consist of pressure fluctuations superimposed on this state of pure hydro-
static equilibrium. (In other words, the pressure fluctuations used above,
i.e., 10hP a has nothing to do with the much larger pressure changes between
surface and tropopause which is ≈ 750hP a). What we should really check is
whether this perturbed state satisfies the hydrostatic approximation. To do
this, write
α = α + α0 and P = P + P 0 (4.30)
where the bar denotes the background state in hydrostatic balance and the
prime the fluctuations associated with weather systems. One can readily
obtain that,
∂P ∂P 0 α0
α +g ≈α −g (4.31)
∂z ∂z α
Putting orders of magnitude on each term, we find,
∂P 0 10hP a
α ' (1kgm−3 ) = 10−1 ms−2 (4.32)
∂z 10km
and
α0 θ0 3
−g ≈ −g ' 9.81 = 10−1 ms−2 (4.33)
α θ 300
These terms are still much larger than all the others in (4.27) and so the
hydrostatic approximation also applies to dynamic, as opposed to purely
static situations,
∂P 0 α0
0≈α −g (4.34)
∂z α
Most climate models use this approximation. (Among other things it allows a
change of vertical coordinate z → P which simplifies considerably the model
numerics).
NB? : You might rightly notice that the momentum balance (4.34) is different
than the one used in Chapter 3 for sloping convection. What was missing
there is the fact that the parcels ascending and descending are embedded in
a wave, with associated pressure variations. Their vertical gradients are the
primary mechanism opposing the buoyancy force, not vertical acceleration of
the parcels. Returning to Fig. 3.7, as parcel A goes up, it gains kinetic energy
through the work of buoyancy force (and as we know from section 3.6 this is
maximized if it moves at an angle equal to half the slope of the θ surfaces).
From eq. (4.34), we now know that it looses kinetic energy because of adverse
vertical pressure gradients. The later act as a brake on the wave whose energy
source is the work computed in section (3.6).
4.2.2 Horizontal momentum equation

The two components of the horizontal momentum equations take the form,

Du uw uv tan φ ∂P
=− + + 2Ωv sin φ − 2Ωw cos φ − α (4.35)
Dt R R R ∂x
vw u2 tan φ

Dv ∂P
=− − − 2Ωu sin φ − α (4.36)
Dt R R R ∂y
Scaling these terms, we obtain (from left to right): U/T ≈ 10−4 , U W/R ≈
10−8 , U 2 /R ≈ 10−5 , ΩU sin φ ' 10−3 sin φ, ΩW cos φ ≈ 10−6 cos φ and
α∇L P ' 10−3 (using the surface value for α ' 1m3 /kg) for the zonal
momentum equation. Likewise, for the meridional momentum equation:
U/T ≈ 10−4 , U W/R ≈ 10−8 , U 2 /R ≈ 10−5 , ΩU sin φ ≈ 10−3 sin φ, and
α∇L P ≈ 10−3 .
Looking at these numbers it becomes clear that, as long as we are not too
close to the equator where sin φ = 0, the approximate form of the horizontal
momentun equation is,
∂P
0 ' +2Ωv sin φ − α (4.37)
∂x
∂P
0 ' −2Ωu sin φ − α (4.38)
∂y
This approximation is called the geostrophic approximation, expressing a

near cancellation between the acceleration due to the horizontal pressure
gradient and Coriolis forces. One way to “visualize” this balance is that air
parcels tend to flow along lines of constant pressure (anticlockwise around a
low pressure system in the extra-tropics in the Northern Hemisphere).
After these two terms the next larger one is (Du/Dt)R for the zonal
momentum equation and (Dv/Dt)R for the meridional momentum equation.
Both terms scale as U/T = U 2 /L. How small is this term compared to the
Coriolis force is measured by a non dimensional number called the Rossby
number (Ro ),
U
Ro ≡ (4.39)
2ΩL
In other words, the smaller the Rossby number, the closer we are to geostrophic
balance. For the scales considered at the beginning of section 4.2, Ro ≈ 0.1.
NB: Not all atmospheric motions have small Rossby numbers. For example,
hurricanes have typical windspeed U = 30ms−1 and horizontal lengthscale
L = 100km, yielding Ro ≈ 30/(2.10−4 × 105 ) = 1.5. Thus the geostrophic
approximation is not valid for hurricanes and advection of momentum by the
horizontal flow plays a key role in their dynamics.
4.2.3 The thermal wind relation

The smallness of the Rossby number is the main reason why the Earth rota-
tion has so much influence on weather systems and many other motions. We
have everything at hand to see straight away one of the powerful constraints
resulting from Ro 1, namely a constraint on how motions vary with height.
All we need are the approximate momentum equations (4.29), (4.37) and
(4.38). Take the vertical derivative of (4.37), yielding,
∂v ∂ ∂P
f ≈ [α ] (4.40)
∂z ∂z ∂x
in which we have introduced the Coriolis parameter,
f ≡ 2Ω sin φ (4.41)
Using (4.29), this can be rewritten as,
∂v ∂α ∂P ∂α ∂P
f ≈ − (4.42)
∂z ∂z ∂x ∂x ∂z
Likewise,
∂u ∂α ∂P ∂α ∂P
f ≈− + (4.43)
∂z ∂z ∂y ∂y ∂z
or, in vector form,

∂V
f ≈ (∇α × ∇P )H (4.44)
∂z
in which V = ui + vj is the horizontal velocity vector and the subscript H
indicates we only consider the horizontal component.
The terms on the r.h.s of (4.44) vanish if α = α(P ). That is, in such
condition, the horizontal velocity field cannot vary with height! –see the
note below on “Taylor columns”. More realistically, in the atmosphere α =
α(T, P ) and the terms on the r.h.s do not vanish. Rather they set a constraint
on how much the horizontal wind can vary with height.
A geometrical derivation brings more insight into these relations. Using
the triple product rule for mixed derivative and the hydrostatic approxima-
tion, one has,

∂P ∂z ∂P ∂z
α = −α =g (4.45)
∂x z ∂x P ∂z x ∂x P
which relates the pressure gradient to the slope of a pressure surface. As a

result, the difference in meridional wind (v) between two heights (z1 > z2 )
is, using the geostrophic balance,

g ∂z g ∂z
v2 − v1 = − (4.46)
f ∂x P2 f ∂x P1
Likewise, for the zonal wind,

g ∂z g ∂z
u2 − u1 = − (4.47)
f ∂y P1 f ∂y P2
These relations show that variations with height of the geostrophic wind
reflect horizontal gradients in the thickness between two pressure surfaces
(Fig. 4.4). The thickness is itself directly related to temperature since the
hydrostatic equation can be rewritten, using the ideal gas law (and neglecting
moisture), as,
∂ ln P g
=− (4.48)
∂z Rd T
This indeed shows that for two given pressure surfaces (P1 , P2 ), the thickness
between them (= z1 − z2 ) increases with the temperature of the layer of air
sandwiched between these two surfaces,
Z z1 Z P2
Rd T
z1 − z2 = dz = d ln P (4.49)
z2 P1 g
Figure 4.4: The variations of the geostrophic wind with height reflect horizon-
tal temperature variations: in the Northern Hemisphere (shown in picture),
winds increasing with height have cold air to their left (if x is interpreted as
the east - west distance, the figure indicates southward wind v1 < 0 at upper
level and northward wind v2 > 0 at low level). The distance between two
pressure surfaces (the thickness) increases with temperature. In this diagram
two pressure surfaces (P = P1 and P = P2 ) are shown in black, being at a
height z = z1 (x) and z = z2 (x), respectively.
4.3. THE VORTICITY VIEW 95
This is why the wind variations in (4.46)-(4.47) are called the “thermal wind”
relations.
A vivid illustratioon of the thermal wind relation is also found in Fig.
1.3 in Chapter 1. If we look for example at the winter hemisphere at 30◦ ,
one observes that the contours of velocity close on themselves at a height
of about 10km (i.e., there is a wind maximum here, the Jet Stream). Thus
below the maximum the wind increases with height but above the maximum
they decrease with height. The thermal wind relation shows that this must
be associated with a reversal of the equator-to-pole temperature gradient.
Indeed the Tropics are warmer than the Poles below 10km but are colder (!)
above this level in the lower stratosphere.
NB? : The thermal wind relation offers insight into the “Taylor’s column”
effect discussed at the beginning of this chapter. In a fluid of constant den-
sity, it predicts ∂u/∂z = ∂v/∂z = 0. So the horizontal motion has to be
independent of height and we cannot expect the motion above the obstacle to
be solely limited to its vicinity. It happens that for small Rossby number and
constant density, one can also predict ∂w/∂z = 0 which implies that there
can be no motion above the obstacle at all (this would require having w non
zero just above the obstacle but zero away from it).
4.3 The vorticity view

4.3.1 The geostrophic flow, vorticity and divergence
The approximations derived in section 4.2 are very useful and allow to un-
derstand the basic structure of atmospheric motions. However, you might
have noticed that they do not include time derivatives. In other words, they
are just diagnostic relationships. Were you to forecast the weather based on
them, you would go nowhere!
The breakthrough to achieve this task came in the 1950s when it was
realized that the geostrophic flow is rotational: it goes around pressure cen-
ters (cyclones or anticyclones) but does not cross much the pressure lines.
And so it became clear that what needs to be forecast is not momentum but
vorticity ζ, the rotational (or curl) of the velocity field,
ζ ≡ ∇ × uR (4.50)
To see the importance of vorticity, simply consider its vertical component

(ζ) in the local coordinate system,

∂v ∂u
ζ ≡ ζ.k = − (4.51)
∂x ∂y
On the scales considered in section 4.2, one can, for practical purposes further
simplify the geostrophic balance as,
αo ∂P αo ∂P
v≈ , and u ≈ − (4.52)
fo ∂x fo ∂y
in which αo and fo are typical values of the specific volume and the Coriolis
parameter over the region of interest. From this,
αo ∂ 2 P ∂ 2P

ζ≈ + (4.53)
fo ∂x2 ∂y 2
while, the horizontal divergence of the (relative) flow δ is,
∂u ∂v
δ≡ + ≈0 (4.54)
∂x ∂y
Thus indeed, the geostrophic flow is close to being purely rotational, i.e.,
containing vorticity but no divergence. For such flows a streamfunction ψ
can be introduced from the definition,
∂ψ ∂ψ
u≡− and v = (4.55)
∂y ∂x
Comparison of (4.52) with (4.55) shows that ψ ≈ αo P/fo : the horizontal
flow follows approximately lines of constant pressure.
The concept of vorticity comes from Fluid Mechanics and is thus much
more general than the special case of geostrophic flows. The video shown in
the lecture illustrates its presence or absence in simple flows (e.g., solid body
rotation, source/sink flows in a kitchen sink –link provided on Blackboard).
Technical note: An important result is that a flow in solid body rotation has
a vorticity equal to twice the rate of rotation. A simple derivation can be
obtained as follows. Consider a tank of water rotating around a vertical axis
at a rate Ω. One can compute the vorticity of any fluid parcel embedded in this
flow by considering an anti-clockwise closed contour made of two lines r = r1
and r = r2 (with r1 = r2 + δr, i.e., we are considering a very small contour)
joined
H by RR
two radial lines making
RR a small angle δθ. From Stokes’ theorem,
u.dl = (∇×u).dS = ζdS, which in polar cylindrical geometry, reads
u1 r1 δθ − u2 r2 δθ = ζδrδθ(r1 + r2 )/2. In solid body rotation u1 = Ωr1 and u2 =
Ωr2 , hence, Ω(r12 −r22 )δθ = ζδrδθ(r1 +r2 )/2. Since r12 −r22 = (r1 −r2 )(r2 +r1 ),
one obtains ζ = 2Ω.
4.3.2 Predicting the vorticity of the flow: the vorticity

equation
We are going to restrict ourselves to horizontal flows on the sphere, i.e.,
flows which have exactly zero vertical motion (w = 0). In so doing we
are admittedly taking an extreme limit (for example all types of motions
discussed in Chapter 3 are excluded from this category), but it helps to
see more clearly how vorticity can be predicted. As it happens, horizontal
motions of planetary scales dominate the variability of atmospheric flows on
timescales greater than a week or so, so this is not an academic problem
but rather, one of great importance for seasonal and climate prediction (see
section 4.3.3).
We thus consider the motion of an “air shell” sandwiched between the
Earth and a concentric outer shell, and let us denote the (constant) distance
between the two spheres, i.e., the height of the air column, by d (Fig. 4.5).
We are going to follow in a Lagrangian sense an initial volume of air and
we will neglect any effect of compressibility. Thus we take α = cst = αo .
As a result, the volume of this air sample remains constant. Since d is also
constant, this implies that the surface area S of the sample is constant during
its motion (it can deform and change shape but its total area has to remain
the same). Focusing on flow of sufficiently large scale and small velocity
amplitude, the Rossby number is very small. The thermal wind relation thus
applies and without temperature variations (α = cst) this implies that the
horizontal velocity vector is constant with height. We can thus think of the
flow as simply representing the horizontal motion of a rigid column of fluid
(w = 0 by construction since the flow is bounded by two spherical shells
–Fig. 4.5).
Let’s describe what happens to our column from the point of view of an
observer in an inertial frame (say viewed from Space). Its velocity is,
uI = uR + Ω × r (4.56)
We are going to use the result, known as Kelvin’s identity (proven in Ap-
pendix), that if we follow a circuit made of fluid elements (hereafter called a
“material contour”), i.e., any closed circuit within the column, we have,
I I
D DuI
uI .ds ≡ .ds, (4.57)
Dt I Dt I
in which ds denotes an infinitesimal segment of the circuit. Using the equa-
tion of motion in an inertial frame (4.7), this can be rewritten as,
I I
D
uI .ds = (−g − α∇P + F f ric ) .ds, (4.58)
Dt I
Figure 4.5: Horizontal motion of a column of fluid (green shading, height d,

horizontal surface area S) sandwiched between the Earth and a concentric
outer shell.
H H
Since g = −∇Φgravi , we have g.ds = dΦgravi = 0 (same starting and
end
H point in a closed
H contour). Likewise, because we take α to be constant,
α∇P.ds = αo dP = 0. If, in addition, we neglect friction, we come to
the conclusion that: I
D
uI .ds = 0. (4.59)
Dt I
From Stokes’theorem, this is also,
ZZ
D
(∇ × uI ).dS = 0. (4.60)
Dt I
where the integral is taken along any surface encircling the material contour.
Using (4.56) and (4.50), this can be rewritten as,
ZZ
D
[ζ + ∇ × (Ω × r)].dS = 0. (4.61)
Dt I
Coming back to the motion in Fig. 4.5, a convenient choice of surface is

clearly a sphere. With this choice, the normal dS is in the vertical direction
(k in Fig. 4.5), and eq. (4.61) can be rewritten as,
ZZ
D
(ζ + f )dS = 0. (4.62)
Dt I
Note that to arrive at this expression, we have used the result (shown in the
Appendix) that ∇ × (Ω × r) = 2Ω. We now take the limit in which the
horizontal cross section of the column of air in Fig. 4.5 is infinitesimally small.
Then S → dS is constant following the flow from the previous arguments,
but we now also know, from (4.62), that the product of (ζ + f )S must be
constant following the flow. Hence we obtain the deceptively simple “vorticity
equation”:
ζ + f = cst (following the flow) (4.63)
As we illustrate in the following section, this equation provides everything
we need to predict the motion in Fig. 4.5.
Although we derived this equation by describing a material circuit in an
inertial frame, the end result should be independent of that choice. Indeed,
using the fact that (DB/Dt)I = (DB/Dt)R for any scalar B, we can as well
write, ZZ
D
(ζ + f )dS = 0. (4.64)
Dt R
One could derive this equation by following the same procedure as above in
the rotating frame of the Earth, but this would involve the (more mathemat-
ically involved) calculation of a “torque” by the Coriolis force.
The variables ζ and f refer to the projection of the vorticity vectors ζ and
2Ω onto the local vertical. There is a more general equation for the absolute
vorticity vector ζ a = 2Ω + ζ but we will not discuss it in this course (you can
see this equation in the picture on the “Welcome to Atmospheric Physics!”
page at the beginning of these notes. If you’re interested I have included a
derivation in the Appendix).
NB1? : In the inertial frame, the local rate of rotation of a fluid parcel is twice
the quantity ζ + f , which is called the “absolute vorticity” (≡ ζa ). It is the
sum of the vorticity ζ (the “relative vorticity”) due to the relative motion of
the parcel with respect to the solid body rotation of the Earth, and that due to
the solid body rotation of the parcel around the Earth’s axis (f , the “planetary
vorticity”). As the experiment in the vorticity video showed (specifically, the
part showing the vorticity meter in a solid body rotation of a fluid in a tank),
in a solid body rotation the fluid locally rotates at the same angular velocity
as the tank. The associated vorticity is twice this rate (see technical note in
section 4.3.1). In the rotating frame of reference, the local rate of rotation
of a fluid parcel is just twice the relative vorticity (the vorticity meter in a
pure solid body rotation does not change for an observer rotating itself with
the tank).
NB2? : You could take the other limit in which the circuit in the above dis-
cussion shrinks (like a the flow in a kitchen sink), i.e., S decreases. The
prediction from the circulation theorem is that (ζ + f )S must remain con-
stant. Hence, ζ + f will have to increase. If the circuit shrinks over a limited
range of latitude, this requires that the relative vorticity increases. In other
words, a cyclone is intensifying. This is a decent model for the growth of
hurricanes. There are many other applications of the circulation theorem.
An example is given in Q5 (Gulf Stream) of the problem sheet.
4.3.3 Rossby waves

A large class of motions of the atmosphere are loosely defined as being of
the “planetary wave” type, to indicate that they are of sufficiently large
spatial scale that they are affected by the fact that the Earth rotation vector
Ω has a projection on the local vertical which varies with latitude. An
example is given in Fig. 4.6, showing the height of the 500hP a surface for
the Northern Hemisphere on 10 February 2014 –this approximately provides
a streamfunction for the flow at mid-tropospheric levels from the discussion
in section 4.3.1 and eq. (4.45). A wavenumber 2 structure is clearly apparent
in the east-west direction, spreading from the North Pole to 30◦ N .
Figure 4.6: The height (in km) of the 500hP a pressure surface on
10/02/2014. Black contours indicate its absolute values while the
colours (in m) denote anomalies compared to the long time mean. From
http://www.cpc.ncep.noaa.gov/products/precip/CWlink/MJO/block.shtml.
The Rossby number for these motions is particularly low Ro = U/2ΩL ≈

10/(2 × 7.2 × 10−5 × 14, 000km) = 0.005 (using a wavenumber 2 at 45◦ N )
so that they satisfy the geostrophic balance and are nearly purely rotational
(section 4.3.1). A lot of insight can be gained into these motions by using
(4.64), which we express mathematically as:

∂ ∂ ∂
+u +v (f + ζ) = 0 (4.65)
∂t ∂x ∂y
Equation (4.65) states that absolute vorticity “sticks” to air parcels: you
can tell the origin of a given air mass from looking at its absolute vortic-
ity (e.g., a low value of ζa typically originates from low latitudes)! A very
powerful tool.
The equation (4.65) supports linear waves called Rossby waves after the
Swedish metorologist Carl Gustav Rossby who highlighted their dynamics in
the 1950s. Take the simplest possible case, i.e., linear motion on a state of
rest. Then (4.65) can be rewritten as,
∂ζ 0
+ βv 0 = 0 (4.66)
∂t
in which primes indicate perturbations and
df
β≡ (4.67)
dy
captures the latitudinal variation of the projection of Ω onto the local verti-
cal. This equation can be solved by using the streamfunction ψ introduced
in (4.55) and looking for plane wave solution2 ,
ψ 0 ∝ ei(kx+ly−ωt) (4.68)
This provides the dispersion relation,

βk
ω=− (4.69)
k2 + l2
and the associated phase velocity v P ,
β
vP = − (1, k/l) (4.70)
k2 + l2
2
The reason a streamfunction can be introduced rigorously here is that we are looking
at horizontal motion and have assumed constant density. Using these approximations, the
continuity equation (4.26) predicts that the horizontal divergence of the flow has to be
zero, i.e., ∂u/∂x + ∂v/∂y = 0, hence u = −∂ψ/∂y and v = ∂ψ/∂x.
The frequency of Rossby waves is typically much lower than that of the
weather system we experience nearly daily (the latter were the subject of
section 3.6 and, contrary to Rossby waves, the presence of upward and down-
ward motion is crucial for their existence). For example, for the observations
in Fig. 4.6, I found ω ≈ 2π/43days−1 (using a wavenumber 2 at 45◦ N for k
and l = 2π/3000km). No wonder why these waves are studied tremendously
since, owing to their long timescales, they introduce predictability to our
weather.
Rossby waves display a myriad of other interesting features:
• their phase speed is always westward (ω/k < 0). This follows directly
from the conservation of vorticity by fluid parcels, as encapsulated in
(4.65) and illustrated in Fig. 4.7. The usual way to describe a wave in
terms of a restoring force is not useful for Rossby waves since, as em-
phasized earlier, they are in geostrophic balance and the momentum
equation does not say much about time evolution. This is why “vortic-
ity thinking”, as illustrated in Fig. 4.7, brings much more insight into
the mechanism of propagation.
• they are associated with a transverse motion of air parcels. This is

illustrated in Fig. 4.8, where the velocity vector (blue arrows) is aligned
along lines of constant pressure (by geostrophy) while the phase speed
(black arrow) is at right angle. This is very different from other types
of waves supported by fluids (acoustic, gravity waves, i.e., swell) which
involve longitudinal motions. Again, this is a behaviour closer to that
of solids (e.g., shearwaves in the solid Earth generated by earthquakes).
As you can see, this all has to do with β 6= 0.
• they carry east-west momentum across latitude circles. This property

is illustrated in Fig. 4.8. Consider the portion of a latitude circle
corresponding to a zonal wavelength (marked by the red a-b-c line in
the figure). At a, a fluid parcel has u0 > 0 and v 0 > 0, hence it carries
eastward momentum northward. At b, the parcel goes southward (v 0 <
0) but it has a negative eastward momentum (u0 < 0), hence again,
it carries eastward momentum northward (removing negative eastward
momentum from a region is like adding positive eastward momentum
to it). At c, we have the repeat of what occurs at a. So averaged over
a zonal wavelength, there is a net transport of eastward momentum
northward. Mathematically this is expressed as a systematic positive
correlation between u0 and v 0 : averaged over a zonal wavelength (east-
west direction), the product u0 v 0 has a sign proportional to −kl (simply
take ψ 0 ∝ sin(kx+ly−ωt) and check that u0 v 0 ∝ −klcos2 (kx+ly−ωt)).
The bottom line is that Rossby waves carry zonal momentum with
them: this is the mechanism linking trade winds and westerlies that
was mentioned in Chapter 1.
4.4 References
-Rossby and collaborators, 1939: Relation between variation in the intensity
of the zonal circulation of the atmosphere and the displacement of the semi-
permanent centers of action, J. Mar. Res., 38-55.
-Taylor, G. I., 1923: Experiments on the motion of solid bodies in rotating

fluids, J. Fluid Mech., 104, 213-218.
4.5 Problems
Q1. A thunderstorm moving in a sheared flow has the following scales: L =
10km, H = 10km, U = 10ms−1 , W = 1ms−1 , f = 10−4 s−1 , P 0 = 10hP a.
Determine whether the geostrophic and hydrostatic approximations apply to
this system.
Q2. An aeroplane is due to fly eastward over the ocean at 45◦ N. At some
moment it is at an altitude of 6000m and a pressure altimeter indicates a
pressure of 100hP a. The pilot maintains this pressure altitude but notices
that after 1 hour the radar altimeter indicates an absolute altitude of 5750m.
How far, and in what direction, has the plane drifted off course? Clearly
state any assumptions you make. [Hint: this question is more difficult than
it looks.]
Q3. The derivation of the equation of motion (4.15), and its decomposition
into components (4.27), (4.35) and (4.36), was quite mathematical. This
question aims at re-deriving those equations from a more intuitive perspec-
tive.
(i) Particle moving purely east-west, i.e. u = (u, 0, 0) in the local frame
of reference (i, j, k). By thinking about the particle’s circular motion
around the Earth at constant latitude φ, and its associated centrifu-
gal force, show that it must experience an acceleration (Ω2 R cos φ +
u2 /R cos φ + 2Ωu)(cos φk − sin φj). Identify the corresponding terms
in (4.27) and (4.36).
4.5. PROBLEMS 105
Figure 4.7: Schematic of the propagation mechanism for Rossby waves in

the Northern Hemisphere (x−axis from west to east, y−axis from south to
north). The horizontal line indicates lines of constant planetary vorticity
(f ), increasing towards the +y direction (f1 < f2 < f3 ). In presence of an
anticyclone, these lines are bend and, for the case shown in the figure, low
latitudes air parcels move northward to the west of the anticyclone, and high
latitudes air parcels move equatorward to the east (black arrows). Assuming
that these parcels did not possess significant relative motion before they were
disturbed by the anticyclone (i.e., their initial absolute vorticity is simply f ),
the conservation of their absolute vorticity ζ + f as they move requires that a
region of ζ < 0 develops to the west of the anticyclone (blue patch), and that
a region of ζ > 0 develops to the east of the anticyclone (red patch). As a
result the anticyclone appears to move westward (an anticyclone corresponds
to a region of ζ < 0), and creates a cyclone in its lee (a cyclone corresponds
to ζ > 0). You can easily see from there why a wave of anticylone/cyclones
would naturally propagate westward.
Figure 4.8: Schematic of the phase lines (blue) for a Rossby wave with
k > 0, l < 0 in the Northern Hemisphere (same axes as in previous fig-
ure). The flow (blue arrows) is clockwise around the high pressure (H) and
anti-clockwise around the low pressure (L). The phase velocity (black arrow)
is perpendicular to these motions. For the case considered here, the transport
of east-west momentum is to the North (green upward arrow).
4.5. PROBLEMS 107
(ii) Particle moving purely north-south, i.e. u = (0, v, 0).
(a) A particle of unit mass is stationary on the Earth’s surface at

latitude φ. Show that its angular momentum is ΩR2 cos2 φ.
(b) Suppose this particle is subject to an impulsive force which sets
it moving northward, staying on the surface. It has experienced
no torque so its angular momentum must be conserved; why does
this imply that it must develop an eastward velocity component?
(c) If in a time δt the particle has reached latitude φ + δφ and ac-
quired an eastward velocity component δu, show that its angular
momentum is now [Ω + δu/R cos(φ + δφ)]R2 cos2 (φ + δφ).
(d) Using conservation of angular momentum, expanding cos(φ + δφ)
and neglecting 2nd order terms in small quantities, show that
δu = 2ΩRδφ sin φ and hence that δu/δt = 2Ωv sin φ. Identify this
term in (4.35).
(iii) Particle moving purely upward, u = (0, 0, w). Repeat part (ii) but
consider a particle impelled vertically upwards with speed w.
Q4. The schematic below (Fig. 4.9) represents a front between two air
masses at different temperatures. We suppose that the temperature differ-
ence is 6 K at all levels, that the front extends over a horizontal distance of
300 km and from the surface (P2 = 1000hP a) to a level of P1 = 200hP a. We
wish to estimate the wind at 200hP a at the center of the front.
(i) By using eqs. (4.47) and (4.49), show that the changes in zonal (east-
west) wind with height can be approximated as,

Rd P2 ∂T
u2 − u1 ≈ ln (4.71)
f P1 ∂y
where T is the averaged temperature over the 1000 − 200hP a layer.
(ii) Assuming that the mean latitude of the front is 45◦ N , estimate the
wind at its center at 200hP a. State any assumptions made.
Q5? . In this question, we apply the concept of vorticity to ocean currents

making up the great “subtropical gyres” (Fig. 4.10, red loop). We accept
that the impact of the wind on the ocean, within the 10◦ − 50◦ N band of
latitude where the gyres are found, is to drive an equatorward flow in its
Figure 4.9: Schematics of temperature variations at a front.
upper (0 − 1000m) layers. As water parcels flow equatorward, their absolute

vorticity ζ + f is accordingly changed. Note that throughout this question
absolute vorticity is not conserved, either because of the effect of the winds,
or because of the effect of frictional forces near the continental shelves.
(i) Show that since the equatorward current is on the order of a few cm/s
and its horizontal scale is that of the North Atlantic basin, the absolute
vorticity of water parcel in the equatorward flow is approximately f .
Determine the resulting loss of absolute vorticity along the equatorward
flow.
(ii) To close the mass budget, a “return” current must occur in the pole-
ward direction. In the North Atlantic, this current is called the Gulf
Stream. Considering that the latter has a magnitude of a few m/s and
an horizontal extent of a few tens of km, do you think the approxima-
tion ζa ≈ f is still valid in the Gulf Stream?
(iii) Based on your answer to the previous question, and thinking of the be-
haviour of absolute vorticity within a closed loop around the subtropi-
cal gyre, determine whether water parcels must gain or loose absolute
vorticity in the Gulf Stream.
(iv) By thinking about the effect of horizontal friction on vorticity explain

why the Gulf Stream is found on the east coast of the US and not on
the west coast of Portugal.
4.5. PROBLEMS 109
Figure 4.10: Schematics of the ocean circulation in the North Atlantic. The
subtropical gyre is highlighted as the red “loop”. Image courtesy of the SEOS
project.
Q6. One sometimes reads or hears that, by analogy with atmospheric low
pressure systems, the flow in a bath tube which is being emptied circulates in
opposite direction in the Northern and Southern Hemisphere. Give at least
two reasons why this is bonkers.
Q7. Past Exam Question (2003 no.4).
(i) What is vorticity and why is it useful for understanding the weather?
(ii) Consider a long-wave pattern at the level of non-divergence in a zonal

current of uniform constant velocity, U . From the principle of conser-
vation of absolute vorticity and making the assumption that the total
velocities are independent of latitude show the Rossby wave equation
can be written as:
∂ 2v0
(U − c) 2 + βv 0 = 0 (4.72)
∂x
where c is the wave velocity, v 0 the meridional velocity perturbation
and β the meridional rate of change of the Coriolis parameter. Find an
expression for c in terms of the wavelength and β.
(iii) Calculate the wavelength of the Rossby wave if it appears stationary at

50◦ N when the zonal wind is 50ms−1 . For the same zonal wind, what
would happen to a wave of the same wavelength at lower latitudes?
(The radius of the Earth is 6371 km).
Chapter 5
Climate change
We defined climate in Chapter 1 as the state of the atmosphere averaged over

a sufficiently long time. What that time averaging period needs to be is not
clear for several reasons. First, the atmosphere on its own is capable of fluc-
tuations of its statistics over timescales of decades. The MIT meteorologist
Ed Lorenz (the one who is now associated with the “Butterfly effect”) talked
about the atmosphere being an “almost intransitive system”, i.e., a system
whose statistics can vary significantly over time (Lorenz, 1968). Second, the
atmosphere interacts with other systems and this also leads to changes on a
very wide range of timescales:
• seasonal to interannual, due to changes in sea surface temperature
• multidecadal to multi-centennial due to changes in the state of the

oceans (temperature and circulation).
• centennial to millenial due to changes in the state of the ice sheets
• millions of years due to the motion of continental plates (tectonics)

In addition, the natural variability of solar activity, as well as the geometry
of the Earth’s orbit around the Sun, and of its own axis of rotation, also
introduces changes in the amount of solar energy reaching the top-of-the-
atmosphere. These occur on timescales on the order of decades for the so-
called “11-yr” solar cycle, and several tens of thousands of years to several
hundred of thousand years for the Milankovitch cycles (see the fascinating
account of their history in Imbrie and Imbrie, 1979).
We focus in this chapter on a very special type of climate change: the
one caused by perturbations of the composition of the atmosphere as a result
of human activities, with a particular emphasis on the burning of fossil fuel
(increase in carbon dioxide concentration). Only the physical aspects are
111
112 CHAPTER 5. CLIMATE CHANGE
discussed, with a focus on the ocean-atmosphere system. For a more holistic

view of the problem, I highly recommend the short book by Prof D. Archer
(2009).
5.1 Anthropogenic radiative forcing

As discussed in Chapter 2, carbon dioxide is a greenhouse gas owing to its
tri-atomic structure. The major absorption band of the CO2 molecule is
centered near 15µm, which happens to be in the spectral region where the
atmosphere emits most intensely infrared radiation (Fig. 1.6). Thus it is
no surprise that the accumulation of carbon dioxide in the atmosphere is
perturbing the heat balance of the Earth.
Figure 5.1 quantifies the magnitude of the change in downward radiation
at the tropopause with respect to pre-industrial times. Carbon dioxide is
seen to dominate the anthropogenic forcing, with an anomalous downward
energy flux (infrared) on the order of 1.5W m−2 . Methane comes second with
a forcing on the order of 1W m−2 (as you can see in the plot this number
involves not only the direct effect of methane on radiation but its knock-on
effect on other greenhouse gases through chemical reactions). Note that some
of the anthropogenic forcings act to cool the planet, for exemple aerosols,
because they increase the reflection of solar radiation. This effect is both
direct (aerosols reflecting directly the sunlight back to space) and indirect
(by promoting the formation of clouds with smaller water drops and higher
albedo). Aerosols also have a warming effect by absorbing and re-emitting
infrared radiation which cancel partly the previous two. The net is estimated
to be a cooling on the order of ≈ 0.25 + 0.55 = 0.8W m−2 in Fig. 5.1.
5.2 Response of the atmosphere to a sudden

doubling of CO2
We focus here on the CO2 impact only (infrared), and consider the “thought
experiment” of what would happen to the atmosphere if we were to suddenly
multiply its CO2 concentration by a factor of two. Even when restricting
attention to the atmosphere-ocean system only, as we’ll do, there are several
timescales involved in this process:
• The first (fast, on the order of a few days to weeks) is that in which the
temperature field simply reflects the local change in radiative heating
induced by doubling CO2 concentrations (i.e., increases where there is
5.2. RESPONSE OF THE ATMOSPHERE TO A SUDDEN DOUBLING OF CO2 113
Figure 5.1: Anomalous (i.e., 2011 vs 1750) downward radiative flux at the
tropopause in W m−2 . The radiative forcing of various constituents is listed.
Figure taken from the 5th assessment report of the IPCC.
more heating, decreases where there is less heating or more radiative

cooling). This effect must also account for the interaction of these
changes with convection, as discussed in Chapter 3.
• The second timescale is on the order of a few years and corresponds to

a warming of the upper ocean. This warming feedbacks on atmospheric
temperature profiles through radiation and convection.
• Finally, on timescales of years to decades and longer, the ocean circu-

lation moves the excess heat both laterally and vertically until a final
equilibrium is reached.
5.2.1 Fast response: atmospheric processes only

Figure 5.2 shows the change in upward (blue) and downward (red) longwave
fluxes (i.e., integrated over wavelengths greater than 4µm) for a doubling
of CO2 concentrations from 368ppm to 736ppm. This calculation is the
same than the one used in Chapter 2 (Fig. 2.16) for the reference case of
CO2 concentrations at 368ppm. Note that in this calculation, neither the
temperature or the moisture profiles of the atmosphere are changed, just the
amount of CO2 .
The 1st thing to notice is that there is a decrease in the upward flux (the
blue curve is negative) and an increase in the downward flux (the red curve
is positive). The decrease in upward flux is readily understood as an increase
in the optical thickness of the atmosphere in the infrared when CO2 is dou-
bled. Likewise, the increase in downward flux reflects the greater emissivity
(Kirchoff’s law) of the atmosphere as a result of the CO2 doubling. One
can actually obtain a qualitative explanation for the “bell shape” structure
seen in this figure using the simple calculation in Chapter 2 (section 2.4.3,
isothermal atmosphere). For the case of the downward irradiance for exam-
ple (red curve in Fig. 5.2), the absence of change at the TOA simply reflects
the boundary condition of no downward irradiance there. The fact that the
change decreases close to the Earth’s surface reflects that, irrespective of the
amount of CO2 , the downward irradiance ultimately converges to the “sat-
urated value” (Bo in section 2.4.3). The maximum change seen in between
reflects the greater optical depth of the atmosphere in the longwave when
CO2 is added and so the fact that “saturation” is reached faster. Similar
arguments can be made for the change in upward irradiance, as discussed in
the Lecture.
To estimate the effect of these fluxes on the temperature, we compute the
associated change in heating rate ∆Qrad = Qrad (736ppm) − Qrad (368ppm),
Figure 5.2: Changes in upward (blue) and downward (red) infrared fluxes
(in W m−2 ) as a result of a doubling of atmospheric concentrations of carbon
dioxide. The changes are shown as a function of pressure.
with
d
∆Qrad = (∆F ↓ − ∆F ↑ ) (5.1)
dz
using the definitions and notations in Chapter 2. The calculation in Fig. 5.2
was carried out in pressure coordinates rather than height z, so we must first
re-express the previous equation in pressure coordinates. To do so we use
the hydrostatic equation, yielding,
d
∆Qrad = −ρg (∆F ↓ − ∆F ↑ ) (5.2)
dP
The unit of Qrad is W m−3 which is not particularly helpful. It is more

intuitive to convert this heating rate in units of K/day and this can be done
by simply dividing by ρcp . (Note that cp and not cv is used because we focus
on an infinitesimal layer between P and P + dP , i.e., a layer of constant
pressure). The result is shown in Fig. 5.3. One observes anomalous heating
of the troposphere, on the order of 0.1K/day and a larger cooling of the
upper stratosphere and mesosphere (this increase in the magnitude mostly
reflects the little amount of mass at these high altitudes).
One can qualitatively understand these contrasting responses. The tropo-
spheric warming results both from more absorption of the radiation emitted
by the ground (same T but greater optical depth), and more infrared radia-
tion emitted by the stratosphere. The stratospheric cooling is perhaps more
surprising but can be understood in the following way. First, the stratosphere
emits more infrared both downward and upward, which tends to cool it. In
addition, it also receives less upward infrared radiation from the troposphere,
which also cools it. This effect is partly opposed by the increased opacity of
the stratosphere with more CO2 , i.e., more absorption of infrared radiation
in the stratosphere, but the net is overwhelmingly a cooling.
The size of the anomalous heating is such that it would take about 10days
to reach a typical change of 1K in the troposphere. In a globally averaged
sense, the atmosphere sees mostly ocean as its lower boundary. Because of
the latter’s large thermal inertia, the surface temperature would change by
much less in 10days. Thus the troposhere warms up but its lower layers
remain “climatological”. The atmospheric lapse-rate then decreases which
makes the atmosphere more stable to displacements of air parcels from low
levels (Chapter 3) –note that the vertical structure in the anomalous heating
in Fig. 5.3 adds further to this “boundary effect”. This is reflected in weaker
updrafts/downdrafts and a general weakening of the atmospheric circulation.
You’ll be pleased to learn that this effect was only emphasized recently in
the journal Nature Geosciences (Bony et al., 2013).
Figure 5.3: Same as Fig.5.2 but for the heating rate, expressed in K/day.
The same calculations can be carried out latitude by latitude, the re-
sult being that the tropospheric heating is more pronounced at the equator
(Fig. 5.4). This can be understood from Beer’s law, the reduction in upward
infrared radiation at the tropopause being proportional to the amount of
radiation emitted by the air column, which is greater in the Tropics. Inter-
estingly, this tropical enhancement is opposed by a larger downward radia-
tive flux from the stratosphere at high latitudes (this simply reflects that the
tropopause is lower down at the poles than at the equator, i.e., that there is
more “stratospheric mass” at the poles than at the equator). As can be seen
in Fig. 5.4 though, the tropospheric effect dominates and the Tropics gain
more energy than the poles when doubling CO2 concentrations.
The tropical amplification of the heating mimics the situation in the time
mean, whereby the Earth experiences an energy deficit at the poles and an
energy gain at the equator. Interestingly, this excess heating is truly realized
in the atmosphere (infrared heating), rather than mostly absorbed at the sea
surface as is the case for the time mean (see Chapter 1). In other words
there is more available potential energy within the atmosphere to power
the storms (Chapter 3). Assuming that conversion rates of this energy into
kinetic energy of storms is unchanged, the sudden doubling of CO2 will lead
to a change in the distribution of storms: either the same number but with
higher winds, or less weak storms, and / or more strong storms.
5.2.2 Slow response: the atmosphere interacts with

the upper ocean
Figure 5.2 indicates that there is a net anomalous downward radiative flux
at the Earth’s surface, with a magnitude ' 2W m−2 . The upper ocean is
well mixed over a layer of depth h ≈ 100m, and its heat capacity per unit
area is ρo co (density ρo = 1025kgm−3 and heat capacity of seawater co =
4000Jkg −1 K −1 , respectively) so it would take a time ∆t to generate a change
of 1K, where
1K
∆t = ρo co h = 6.5years (5.3)
2W m−2
This rough estimate shows that the interaction with the ocean is important
in setting the response of the atmosphere to a CO2 change on timescales
beyond a few years.
This upper ocean warming perturbs the radiative - convective equilibrium
at fixed surface temperature response discussed in the previous subsection.
A new type of equilibrium must be reached, in which a warmer ocean leads
to more destabilization of the atmosphere. An illustration of this effect is
Figure 5.4: Gain in energy (in W m−2 ) of the troposphere as a result of

doubling CO2 concentrations. The contribution from the troposphere (con-
tinuous curve) and the stratosphere (curves with circles and crosses) to this
gain (=the sum of the two) are shown. The two curves for the latter re-
flect different assumptions made regarding the adjustment of stratospheric
temperatures. The troposphere contribution equals the reduction in the up-
ward radiative flux at the tropopause. The stratospheric contribution equals
the increase in the downward radiative flux from the stratosphere. From
Ramanathan et al. (1979).
Figure 5.5: Equilibrium temperature profiles as a function of height for dif-

ferent carbon dioxide concentrations (and fixed relative humidity). This cal-
culation does not include changes in ocean circulation, solely the thermal
interaction of the ocean and atmosphere. From Manabe and Whetherald
(1967).
provided in Fig. 5.5, based on the radiative-convective equilibrium calcula-

tions discussed in Chapter 3. Qualitatively, the response is similar to that
expected from Fig.5.3, i.e., a warming of the troposphere and a cooling of the
stratosphere. The main difference is that now the whole lower troposphere
is warmer, as a result of the warmer sea surface temperature.
Two further important effects are introduced by the increase in sea surface
temperature:
• First, because the atmosphere warms as a whole, its equilibrium vapour
pressure increases. When condensation will occur, there will thus be
more water falling as rain. This is the argument sometimes made re-
garding increased chances of flooding and rain related damages as a
result of anthropogenic forcing.
• Second, since water vapour is the main greenhouse gas, its increas-
ing concentration in the atmosphere triggers a positive feedback loop:

more water vapour leads to greater surface warming, which leads to
greater surface evaporation and (assuming rainfall does not entirely
compensate), more water vapour in the atmosphere.
5.2.3 Very slow response: the atmosphere interacts

with the ocean circulation
Both the changes in atmospheric heating rates and sea surface temperature
discussed in the previous two subsections will lead to changes in atmospheric
circulation patterns (either shift or intensification, or both). These in turn
will alter ocean currents and the capacity of the ocean to transport heat
laterally and vertically. The associated timescales can be as fast as a decade
and as long as a millenium.
Figure 5.6 (right panel) illustrates the change in ocean heat content
through time in response to a sudden doubling of atmospheric CO2 in the
Environmental Physics climate model (EPcm), a simple climate model coded
in MATLAB (linked on Blackboard). One observes a rapid adjustment of
tropical and high latitude upper ocean temperatures (red and blue curves),
but it takes much longer to adjust the subsurface ocean temperatures (tropi-
cal in magenta, high-latitudes in cyan). This longer adjustment corresponds
to the time it takes to circulate waters through the different layers in the
model. If the ocean circulates water between upper and deeper layers with
an intensity Ψ (in m3 s−1 ) and if the deep layer volume is V , then it will take
a time,
1000m × (2.5 × 1014 m2 )
∆t = V /Ψ ≈ ≈ 1, 000years (5.4)
10 × 106 m3 s−1
to do so. To obtain this number, a thickness of 1000m was used for the deep
layer, a surface area equal to that of one hemisphere (EPcm covers only one
such hemisphere) and a value of 107 m3 s−1 was used for Ψ (this is about ten
times the volume flux of all the rivers of the world).
The timescale in (5.4) is clearly reflected in the top-of-the-atmosphere
net radiative flux (left panel) which, after a sudden increase to ' 4W m−2
associated with the initial CO2 doubling, only very slowly goes back to zero
(the new equilibrium)1 .
1
EPcm also displays the fast (weeks) and slow (years) response discussed in the previous
subsections. You can very easily convince yourself of this by running the model over one
month, over one year, etc, to see the rich behaviour of atmospheric responses to a doubling
of CO2 .
Figure 5.6: Changes in (right) heat content (in J) and (left) top-of-the-
atmosphere net radiative flux (in W m−2 ) in response to a sudden doubling
of atmospheric CO2 concentrations in EPcm. The different colors in the right
panel correspond to different oceanic layers (red=upper tropics, blue=upper
high latitudes, magenta=deep tropics, cyan=deep high latitudes).
5.3. CLIMATE CHANGE IN REALISTIC MODELS AND OBSERVATIONS123
The link between TOA radiative imbalance and ocean heat storage simply
reflects conservation of energy (Chapter 1):
ZZ ZZZ
∂
∆FT OA dxdy = ρo co ∆To dxdydz + ... (5.5)
∂t
in which ∆FT OA is the net radiative imbalance at the TOA and ∆To is the
change in ocean temperature caused by doubling atmospheric CO2 concen-
trations. Terms not explicitly written in this equation are the melting of the
cryosphere, and the weak storage of heat in the atmosphere and land.
5.3 Climate change in realistic models and

observations
This section very briefly summarises the slides discussed in the last lecture.
Regarding climate models:
• Climate models are numerical machines solving a large set of coupled,
non linear, ordinary differential equations. Their main issue is the dif-
ficult parameterization of processes occurring on a spatial scale smaller
than their grid size (clouds, convection, ocean turbulence, etc).
• In response to an increase in atmospheric CO2 concentration (and all
other forcings discussed in Fig. 1), the models show an increase in
global averaged surface temperature through time. The models all
capture the “troposphere warms / stratosphere cools” signature. Ocean
warming is sluggish and probably more diffusive in the models than in
the real world.
Concerning the observations:
• Observations do show the expected opposite temperature trends in the
troposphere and the stratosphere. The global averaged surface tem-
perature has also increased since the beginning of the instruthe mental
record (early 20th century), with pronounced decadal timescale fluctu-
ations.
• Top-of-the-atmosphere net radiative fluxes show an imbalance on the
order of +0.5W m−2 since 2001 (net heating). This is, as expected,
smaller than accounted for by anthropogenic forcing (Fig. 5.1) since
this must account for the extra emission of infrared radiation by the
ocean+atmosphere (“the response” to anthropogenic forcing). Obser-
vations of global ocean heat content agree with the TOA imbalance
qualitatively and quantitatively.
If you are interested to read more about the role of ocean heat uptake in
this problem, a recent and non technical summary is provided in Whitmarsh
et al. (2015).
Predictions of the 21st century climate rely entirely on climate models and
as such can be debated and criticized at will because of the many imperfec-
tions of the models. Likewise, all attempts to “prove” climate change through
analysis of existing timeseries of surface temperature (and other relevant cli-
matic timeseries) are endlessly subject to criticisms on statistical grounds.
My personal take is that we can say something from first principles about
the timescales response to, and the magnitude of the forcing associated with
anthropogenic accumulation of atmospheric carbon dioxide. The timescales
are mesmerizing: as we’ve seen in section 5.2, the interaction between the
atmosphere and the ocean leads, at fixed atmospheric CO2 concentration, to
adjustment times of several centuries; Earth system models with interactive
CO2 suggest adjustment times of several tens of thousands of years (Archer,
2009). Thus it is not simply the climate we or our children live in which we
are affecting. Overall, we should keep alive the “back-of-the-envelope” spirit:
it shows that the current rate of accumulation of CO2 in the atmosphere is
a major forcing in terms of rising sea level, melting of the cryosphere, etc
(Czaja, 2012).
5.4 References
-Archer, D., 2009: “The Long Thaw: how humans are changing the next
100,000 years of Earth’s climate”, Princeton University Press, 180pp.
-Bony et al., 2013: Robust direct effect of carbon dioxide on tropical circu-
lation and regional precipitation, Nature Geoscience, 6, 447-451.
-Czaja, 2012: A consideration of the forcings of climate change using simple

physics, Discussion paper 2 of the Grantham Institute for Climate Change.
-Imbrie and Imbrie, 1979: “Ice ages, solving the mystery”, Harvard Univer-
sity Press, 224 p.
-IPCC Fifth Assessment Report, 2013: please see Blackboard for a web link
to the various sections. If you have the stamina, this is an excellent source
of information / figures for issues related to climate change.
-Lorenz, E. N., 1968: Climatic Determinism, AMS monograph, 8, 30, 1-3.

5.5. PROBLEMS 125
-Whitmarsh, et al., 2015: Ocean heat uptake and the global temperature
record, Grantham Institute briefing paper No 14.
5.5 Problems
Q1. Sea ice melting. Satellite observations suggest that the Earth has
been gaining 0.5W m−2 of energy at the top-of-the-atmosphere since the early
2000s. Estimate how long it would take to (i) melt Arctic sea ice (ii) increase
the temperature of the global ocean by 0.1K, were all this energy available
to do so. [Data: latent heat of fusion lf = 3 × 105 Jkg −1 , ice density ρi =
916kgm−3 , ice thickness hi = 2m, ocean heat capacity co = 4000Jkg −1 K −1 ,
ocean density ρo = 1025kgm−3 , average depth of the oceans ho = 3800m.]
Q2. Emission level. One way to think about the response of tropospheric
temperature to a sudden increase in CO2 concentration is illustrated in Fig.
1. Before the increase, the atmosphere effectively emits at a height ze , defined
so that the total outgoing longwave radiation by the Earth (OLR) equals
σT (ze )4 in which σ is Stefan-Boltzmann constant.
(i) Compute the emission level of the Earth, making a reasonable assump-
tion regarding surface temperature and lapse-rate.
(ii) Explain why, as the CO2 concentration is suddenly doubled, the emis-
sion level must increase.
(iii) Estimate this increase if, at equilibrium, the troposphere and the Earth’s
surface warm by 3K, assuming no change in lapse-rate.
(iv) The lapse-rate is empirically found to decrease with surface warming

in climate models. Would this tend to amplify the surface temperature
change or to reduce it?
Q3. Clausius-Clapeyron scaling. From Thermodynamics (Year 2), the

pressure of water vapour at equilibrium (denoted here by the symbol eeq )
with liquid water satisfies, at a given temperature T ,
deeq sv − sl
= (5.6)
dT vv − vl
Figure 5.7: Schematic of temperature as a function of height, indicating the

emission temperature and emission level. The outgoing longwave radiation
(blue arrows) can be thought of originating from this level.
5.5. PROBLEMS 127
in which s is the specific entropy, v the specific volume and the subscripts l
and v refer to liquid and gas phases, respectively.
(i) Show that this can be rewritten as
deeq lv
' (5.7)
dT T vv
stating any assumptions made.
(ii) Show that after using the ideal gas law this can be further rewritten
as,
deeq eeq lv
' (5.8)
dT Rv T 2
in which Rv = kB /mv , mv being the mass of a water molecule.
(iii) Use this expression to prove that equilibrium vapour pressure increases
by 7 % for every degree kelvin increase in temperature. [Data: lv =
2.5 106 Jkg −1 , Rv = 461Jkg −1 K −1 ].
Q4. Using the formula in Q3 from Chapter 4, discuss whether you agree
with the following statement taken from a popular science article about the
very anomalous 2013-2014 winter weather in the Northern Hemisphere:
“The Jet Stream is driven in part by the temperature difference between cold
Arctic air and the warmer air of the middle latitudes. Because the Arctic is
warming more rapidly than the rest of the planet, that difference is shrinking.
This ought to produce a less potent Jet Stream.”
[Info: the polar warming amplification (≈ 1K) is restricted to a surface layer
extending roughly from 1000hP a to 600hP a.]
Q5. Ocean heat uptake. We consider a “box-model” of the climate sys-

tem to predict changes in top-of-the-atmosphere net radiative flux in re-
sponse to a sudden increase in atmospheric CO2 concentrations. The ocean
is decomposed into an upper (thickness h = 50m) and a deep layer (thick-
ness H = 2000m). The density and specific heat of seawater are denoted
by ρo = 1025kgm−3 and co = 4100Jkg −1 K −1 , respectively. At the top-of-
the-atmosphere, the perturbation in the net radiative flux (positive down-
wards) is written as N 0 = F 0 − αT 0 in which F 0 = 4W m−2 for t > 0 (and
F 0 = 0 for t < 0) is the anthropogenic forcing, T 0 is the global surface tem-
perature anomaly, and α = 1.6W m−2 K −1 is the climate feedback parameter.
Figure 5.8: Schematic of a box-model for ocean heat uptake O0 .
(i) Find the timescale tu beyond which one can neglect the heat capacity
of the upper ocean layer. Likewise, until which timescales td is the deep
ocean acting like a heat reservoir?
(ii) Discuss qualitatively and quantitatively the TOA radiative imbalance

on short timescales ( tu ).
(iii) Climate models suggest that on intermediate timescales ( tu but

td ), the ocean heat uptake can be reasonably well modeled as O0 ≈
κT 0 where κ = 0.5W m−2 K −1 is an “heat uptake efficiency”. Discuss
qualitatively and quantitatively the radiative imbalance at the TOA on
these timescales.
(iii) Discuss qualitatively and quantitatively the equilibrium state of the

model.
Chapter 6
Appendices
6.1 Radiative transfer

6.1.1 Radiation pencils vs. photon showers
You might have been wondering how to relate the concept of radiation in-
tensity with the quantum physics ideas you’ve been learning so far. To do
so, consider an horizontal plane and focus on computing the monochromatic
irradiance across a surface area A (Fig. 6.1). In Chapter 2, we have been
writing this as the integral of the radiation intensity Iλ over solid angle:
Z
Fλ = Iλ cos θdΩ (6.1)
A more intuitive view is to count photons with energy E such that hc/λ ≤
E ≤ hc/λ + dE crossing the horizontal surface A per unit time. In the
direction making an angle θ with the vertical, these occupy the volume of
the parallepiped of (slanted) side cdt. This volume is simply dV = Acdt cos θ.
If nE dE denotes the number density of photons with the required energy, the
number of photons contained in this volume is dN1 :
dN1 = (Acdt cos θ)nE dE (6.2)
Now, all the photons in this volume do not travel towards the surface in the
direction θ. If we were in 1D, we would simply introduce a factor 1/2 at this
stage, to represent photons travelling upward or downward. In 3D, we use the
solid angle and write that in that volume dV only a fraction χdΩ/4π travel
towards the surface along the direction θ (if the radiation were isotropic we
would have χ = 1; if the radiation were entirely coming from the direction
θ, we would have χdΩ/4π = δ(θ0 − θ) where the latter symbol reflects the
129
130 CHAPTER 6. APPENDICES
Figure 6.1: Calculation of the irradiance as a result of photons with energy

E = hc/λ travelling from a given direction at speed c towards the surface
A. This “photon shower” view contrasts with that of the “radiation pencil”
illustrated in Fig. 2.4 in Chapter 2.
Dirac symbol). The number of photons travelling towards the surface along
the direction θ and contained within the parallepiped is thus dN2 :
dN2 = dN1 χdΩ/4π = (Acdt cos θ)nE dEχdΩ/4π (6.3)
Since these all carry approximatively the energy E = hc/λ, the amount of
radiation energy through A per unit time, per unit energy interval is obtained
by summing over all parallepipeds. Let’s call this dFE :
X Z
dFE = dN2 E/(Adt) = cEnE dEχ cos θdΩ/4π (6.4)
Physically, dFE represents exactly the same quantity as Fλ dλ. Thus,

Z Z
cEnE χ cos θdΩ/4π dE = Iλ cos θdΩ dλ (6.5)
Using the fact that |dE| = hcdλ/λ2 , this suggests the following formula for
the intensity of radiation,
χ hc hc
Iλ = × (cnE 2 ) × (6.6)
4π λ λ
I’ve written it as the product of three terms, to reflect more clearly its mean-
ing. The first term is a geometric factor reflecting the geometry of the radia-
tion. The second is the expected “transport” term considering that photons
6.2. THERMODYNAMICS OF MOIST AIR 131
travel at speed c and accounting for the conversion “per unit energy interval”
to “per unit wavelength interval”. The last term is simply the energy of the
photons associated with the monochromatic radiation at wavelength λ.
6.1.2 Conservation of intensity

An interesting use of (6.6) is that the three terms identified on its right
hand side are conserved if photons do not interact with matter (scattering,
absorption). This is so because photons do not interact with each other so
that their statistics are unchanged in absence of scattering or absorption.
This provides a straightforward explanation why Iλ is conserved in absence
of scattering or absorption.
6.1.3 Application to blackbody radiation

For this special case, we can use χ = 1 (isotropic) and also the result from
Quantum Physics (Year 2):
8πE 2 1
nE = 3 3
× E/k T (6.7)
ch e B −1
Inserting this equation into (6.6), we recover eq. (2.1) in Chapter 2.
6.2 Thermodynamics of moist air

6.2.1 Entropy of cloudy air
We consider a sample of air in which vapour and liquid phases are assumed
to be in equilibrium, i.e., e = eeq (T ) (RH = 1, see Fig. ??). If m denotes
the total mass of this sample, its specific entropy s satisfies,
ms = md sd + mv sv + ml sl (6.8)
in which sd , sv and sl are the specific entropies for dry air, vapour and liquid
water, respectively. These are known functions of temperature and pressure:
T Pd
sd = sd,ref + cp,d ln − Rd ln (6.9)
Tref Pref
T e
sv = sv,ref + cp,v ln − Rv ln (6.10)
Tref Pref
and
T
sl = sl,ref + cl ln (6.11)
Tref
(The constants Tref and Pref are arbitrary values where we assume we know
the entropies).
Using the fact that on the vapour - liquid phase boundary sv − sl = lv /T
where lv is the latent heat of vaporization, we have,
lv
s = (1 − qt )sd + qv ( + sl ) + ql sl (6.12)
T
in which we have introduced the specific humidity qv , specific content of
liquid water ql and total specific content of water qt = qv + ql . Rearranging,
lv qv
s = (1 − qt )sd + + qt sl (6.13)
T
After using the formulas for sd and sl , this can be rewritten as,
T Pd lv qv
s = sref + [(1 − qt )cp,d + qt cl ] ln − (1 − qt )Rd ln + (6.14)
Tref Pref T
in which sref = (1 − qt )sd,ref + qt sl,ref . Considering that qt 1, this formula

can be approximated as,
T Pd lv qv
s ≈ sref + cp,d ln − Rd ln + (6.15)
Tref Pref T
which is the form used in Chapter 3.
6.2.2 A general formula for the Brunt-Vaisala frequency?

We will start from (3.5) and, to simply this expression further, we use the
fact that α is a state function, i.e., a mathematical function of any two
thermodynamic variables (this is only true because of our assumption of
thermodynamic equilibrium). Because of the additional assumption of isen-
tropic ascent, entropy s is a natural choice. We’ll take pressure P (the total
pressure, P = Pd + e) as the other variable,
α = α(P, s) (6.16)
As a result, and assuming small perturbations,

∂α
αe − αp ≈ (se − sp ) (6.17)
∂s P
6.3. DYNAMICS OF ROTATING FLUIDS 133
(the term involving pressure changes drops because of our assumption that
the environment and the parcel are at the same pressure). Using one of
Maxwell’s relations, this can be rewritten as,

∂T
αe − αp ≈ (se − sp ) (6.18)
∂P s
At this stage we haven’t actually used that sp = cst. Following the same
procedure used in the dry case when we considered how θ is conserved by a
parcel moving upwards or downwards, we write,
dse
se − sp ≈ (zp − zo ) (6.19)
dz
As a result, the parcel’s motion obeys,
d2 zp

g ∂T dse
2
=− (zp − zo ) (6.20)
dt αe ∂P s dz
showing that the Brunt-Vaisala frequency is,

2 g ∂T dse
N = (6.21)
αe ∂P s dz
Stepping back from all this Thermodynamic, we realize that we have
actually not explicitly introduced moisture here –the above derivation is en-
tirely general. For example, in the case of pure dry air (P = Pd ), we recover
(3.13) by using s = cp,d log θ, and (∂T /∂P )sd = αd /cp,d (since for dry air,
ds = cp,d dT /T − Rd dP/P and Pd αd = Rd T from the ideal gas law).
Acknowledging that at constant entropy, temperature always increases
with pressure ((∂T /∂P )s > 0), (6.21) provides a fairly simple rule to test
the stability of a temperature profile (for the case of either pure dry air, or
cloudy air in thermodynamic equilibrium): the profile is stable only if the
entropy increases with height.
6.3 Dynamics of rotating fluids

6.3.1 Formula for D/Dt in a change of frame of refer-
ence
Consider any vector A, which can be written in either an inertial frame with
orthonormal axes (iI , j I , kI ), or in a rotating frame with orthonormal axes
(iR , j R , kR ),
A = Ax,I iI + Ay,I j I + Az,I kI = Ax,R iR + Ay,R j R + Az,R kR (6.22)
Taking the (D/Dt)I of this equation, that is the rate of change of vector A
in the inertial frame, we have,

DA DAx,I DAy,I DAz,I
= iI + jI + kI (6.23)
Dt I Dt I Dt I Dt I
[NB: were we to take (DA/Dt)R we would obtain (DA/Dt)R = (DAx,R /Dt)R iR +

(DAy,R /Dt)R j R + (DAz,R /Dt)R kR since the axes (iR , j R , kR ) do not move
in the rotating frame.]
Equation (6.23) is also,

DA DAx,R iR DAy,R j R DAz,I kR
= + + (6.24)
Dt I Dt I Dt I Dt I
The r.h.s can be further expanded using the product rule, for example for
the 1st term:

DAx,R iR DiR DAx,R
= Ax,R + iR (6.25)
Dt I Dt I Dt I
Using (4.10), this is also,

DAx,R iR DiR DAx,R
= Ax,R + iR (6.26)
Dt I Dt I Dt R
hence:

DA DA DiR Dj R DkR
= + Ax,R + Ay,R + Az,R
Dt I Dt R Dt I Dt I Dt I
(6.27)
For the choice of a rotating frame centred at the Earth’s core, with kR = kI
(parallel to the axis of rotation of the Earth) and (iR , j R ) rotating at the
angular velocity of the Earth Ω = 2π/1day, we obtain,

DA DA
= +Ω×A (6.28)
Dt I Dt R
since (DiR /Dt)I = Ω × iR , (Dj R /Dt)I = Ω × j R and (DkR /Dt)I = 0 .
6.3.2 Kelvin’s identity

The rate of change of the contribution uI .ds of a small part of the circuit in
eq. (4.57) can be written as,
D Du Dv Dw
(uδx + vδy + wδz) = δx + δy + δz + uδu + vδv + wδw (6.29)
Dt Dt Dt Dt
6.3. DYNAMICS OF ROTATING FLUIDS 135
in which, in this section only, we denote by u, v, w the velocities in the inertial

frame and x, y, z the associated coordinates. It follows that for any segment
of the circuit,
Z Z
D DuI 1
uI .ds = .ds + [ u2I ], (6.30)
Dt I Dt I 2
where the last term is the difference in the values between the end points.
For a closed circuit this term vanishes, leading to eq. (4.57).
6.3.3 The vorticity equation?

To obtain an equation for the vorticity vector ζ (all components, not only
its vertical component), “all you need” is to take the curl of (4.15), and use
vector identities. In practice though, it is pretty tedious and so, to do this
in an efficient way though we need to a little bit of preparatory work. First,
we rewrite (4.15) as,
∂uR
+ (uR .∇)uR = −∇Φ − α∇P − 2Ω × uR + F f ric (6.31)
∂t
Then we use the following vector identity,
∇(A.B) = (A.∇)B + (B.∇)A + A × (∇ × B) + B × (∇ × A) (6.32)
to obtain,
∂uR
+ (2Ω + ζ) × uR = −∇(Φ + u2R /2) − α∇P + F f ric (6.33)
∂t
The quantity 2Ω + ζ that appears in this equation requires a physical in-
terpretation. It is the sum of the vorticity of the relative flow (ζ) and the
vorticity of the solid body rotation at angular velocity Ω. We will call it from
now on the absolute vorticity,
ζ a ≡ 2Ω + ζ (6.34)
and measures the total local spin (planetary + relative) of a fluid parcel –
see the discussion in Section 1.3.2. Typically in a cyclone, the relative and
planetary vorticity add up, so the total spin can be quite large. They tend
to cancel each other in anticyclones.
To see where the 2Ω comes from, acknowledge that the velocity of the
Earth’s solid body rotation is Ω × r = Ωr cos φi, so that it is a function
of latitude and radius r = R + z. As discussed in the lecture, the gradient
(or shear) in velocity associated with the variations with latitude (φ) adds
up to the shear introduced by the variations with r to produce a planetary
vorticity vector always parallel to Ω. To be more quantitative one needs to
do the math properly. A simple derivation is to compute ∇ × (Ω × r) in a
Cartesian coordinate system centred at the Earth’s core (since the curl of a
vector must be independent of the coordinate system used to compute it, we
might as well use the simplest coordinate system –we’ll denote it by i0 , j 0 , k0 ,
the latter being parallel to Ω). Using the vector identity,
∇ × (A × B) = A(∇.B) − B(∇.A) + (B.∇)A − (A.∇)B (6.35)
this becomes ∇ × (Ω × r) = Ω(∇.r) − r(∇.Ω) + (r.∇)Ω − (Ω.∇)r. The

second and third terms are zero because Ω is a constant vector. Since ∇.r =
3 and (Ω.∇)r = Ω∂(zk0 )/∂z = Ω, we get ∇ × (Ω × r) = 3Ω − Ω = 2Ω.
We are now ready to take the curl of (6.33). Note first that the first term
on the r.h.s. of (6.33) will not contribute since ∇ × ∇ = 0. So we simply
have,
∂ζ a
+ ∇ × (ζ a × uR ) = ∇ × (−α∇P + F f ric ) (6.36)
∂t
where we have also used ∂ζ a /∂t = ∂ζ/∂t. This can be simplified further by
using (6.35) to produce,

Dζ a
+ ζ a (∇.uR ) − (ζ a .∇)uR = ∇ × (−α∇P + F f ric ) (6.37)
Dt R
At first sight this looks more complicated, but using the continuity equation
(4.26), the first two terms on the l.h.s combine to produce,

D ζa ζ
( ) − ( a .∇)uR = α∇ × (−α∇P + F f ric ) (6.38)
Dt ρ R ρ
The first term on the r.h.s can also be simplified by using the vector identity,
∇ × (A∇B) = ∇A × ∇B (6.39)
to produce,

D ζa ζ ∇α × ∇P ∇ × F f ric
( ) = ( a .∇)uR − + (6.40)
Dt ρ R ρ ρ ρ
or, after multiplication by ρ:

D
ρ (αζ a ) = (ζ a .∇)uR − ∇α × ∇P + ∇ × F f ric (6.41)
Dt R

Notes

Uploaded by

Document Informationclick to expand document information

Copyright:

Available Formats

Notes

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Notes

Uploaded by

Copyright:

Available Formats

Notes on Atmospheric Physics

Physics Department &

1 Contact details: Huxley Building, Room 726; email: a.czaja@imperial.ac.uk.

1 An overview of the atmosphere 1

2 Radiative heating and cooling of the atmosphere 21

2.6.2 Longwave cooling . . . . . . . . . . . . . . . . . . . . . 41

5 Climate change 111

Welcome to Atmospheric Physics!

• I recommend the excellent textbook by Wallace and Hobbs (‘’Atmo-

Each chapter contains a set of problems, whose solutions will be provided

• https://earth.nullschool.net This is a wonderful website depicting

• http://www.ecmwf.int/s/ERA-40 Atlas/docs/ An excellent source

• http://www.sp.ph.ic.ac.uk/aczaja/EP ClimateModel.html This is

Please do not hesitate to come to Office Hours (Thursdays, 11.30-12.30;

Aims and Objectives for the Course

Aims To provide students with an understanding of the physics behind the

An overview of the atmosphere

Before we dive into quantitative analysis of the atmosphere it is important

• The atmosphere is our common environment. It is the fluid we all

• Observations. In fifty or a hundred years, we will still need to check that

1.1 Atmospheric composition

as does water vapour,

This shows that atmospheric pressure can be thought of as a mass mea-

Technical sidenote: units of pressure. Pressure is usually expressed in hP a =

1.2.2 Measures of water in air

In terms of size, because there is so little amount of water in the air, qd  qv ,

Expressed in mm of water per unit area by dividing this quantity by the

Figure 1.1: Annual mean specific humidity (qv , in g/kg) as a function of

1.3 Main features of the atmosphere

• a rich temperature structure, with, in some regions, temperature de-

• the presence of smaller time mean velocities in the North-South direc-

• its convective nature (Fig. 1.5) on scales ranging from a few km to

• The atmosphere is only one component among many (oceans, cryosphere,

1.4 What drives atmospheric winds, weather

Figure 1.4: Global, annual mean atmospheric temperature as a function of

• In addition, solar photons are in comparison much less absorbed by the

• The radiative cooling to space is relatively uniform spatially but incom-

Figure 1.6: Planck function Bλ (see Chapter 2) for a body at T = 5780K

Figure 1.7: A schematic of the radiative exchanges for an atmospheric layer.

1.8). Laboratory experiments with rotating tanks cooled at the out-

L = R cos φ(u + ΩR cos φ) (1.19)

Atmospheric angular momentum has thus a contribution from the Earth’s

1.5 Further reading

Q2 Why are deserts more likely to be present in the sub-tropics than in

Q3 These two subquestions are independent of each other.

(i) One kg of air of specific humidity q1 = 10g/kg is mixed with one kg of

Q5 (i) Show that in an isothermal atmosphere (T = To ) the pressure decays

Q6 Show that the specific volume α of a sample of air can be approximated

Q7 By inspection of a surface pressure map (for example from the on-line

Q8 A simple view of the Tropics is that it is primarily independent of lon-

Radiative heating and cooling

Key concepts: solid angle, irradiance, radiation balance, greenhouse effect,

2.1 Concepts and definitions

Figure 2.1: The intensity of radiation at wavelength λ is denoted by Iλ and

2.1.2 Blackbody radiation

given wavelength λ is given by the Plank function,

which has units of W m−2 per wavelength per solid angle.

2.1.3 Shortwave and longwave radiation

In terms of size, because there is so little amount of water in the air, qd qv ,

λ = αλ (Kirchoff’s law) (2.7)

NB: For a blackbody, λ = αλ = 1 for all wavelengths. Note also that

Using Kirchoff’s law (2.7), λ = αλ which leads to 1 = Jλ /Bλ and so