Professional Documents
Culture Documents
Atmospheric Optics
Atmospheric Optics
Atmospheric Optics
Craig F. Bohren
Pennsylvania State University, Department of Meteorology, University Park,
Pennsylvania, USA
Phone: (814) 466-6264; Fax: (814) 865-3663; e-mail: bohren@ems.psu.edu
Abstract
Colors of the sky and colored displays in the sky are mostly a consequence of selective
scattering by molecules or particles, absorption usually being irrelevant. Molecular
scattering selective by wavelength – incident sunlight of some wavelengths being
scattered more than others – but the same in any direction at all wavelengths gives
rise to the blue of the sky and the red of sunsets and sunrises. Scattering by particles
selective by direction – different in different directions at a given wavelength – gives rise
to rainbows, coronas, iridescent clouds, the glory, sun dogs, halos, and other ice-crystal
displays. The size distribution of these particles and their shapes determine what is
observed, water droplets and ice crystals, for example, resulting in distinct displays.
To understand the variation and color and brightness of the sky as well as the
brightness of clouds requires coming to grips with multiple scattering: scatterers in
an ensemble are illuminated by incident sunlight and by the scattered light from
each other. The optical properties of an ensemble are not necessarily those of its
individual members.
Mirages are a consequence of the spatial variation of coherent scattering (refraction)
by air molecules, whereas the green flash owes its existence to both coherent scattering
by molecules and incoherent scattering by molecules and particles.
Keywords
sky colors; mirages; green flash; coronas; rainbows; the glory; sun dogs; halos; visibility.
1 Introduction 54
2 Color and Brightness of Molecular Atmosphere 55
2.1 A Brief History 55
54 Atmospheric Optics
blind alley, the inhabitants of which belong fell short of the mark. Yet his hypothesis
to the dead past.’’ that ‘‘the blueness we see in the atmo-
Despite this depreciatory statement, sphere is not intrinsic color, but is caused
hoary with age, correct and complete by warm vapor evaporated in minute and
explanations of the color of the sky still insensible atoms on which the solar rays
are hard to find. Indeed, all the faulty fall, rendering them luminous against the
explanations lead active lives: the blue infinite darkness of the fiery sphere which
sky is the reflection of the blue sea; lies beyond and includes it’’ would, with
it is caused by water, either vapor or minor changes, stand critical scrutiny to-
droplets or both; it is caused by dust. day. If we set aside Leonardo as sui generis,
The true cause of the blue sky is not scientific attempts to unravel the origins of
difficult to understand, requiring only a the blue sky may be said to have begun with
bit of critical thought stimulated by belief Newton, that towering pioneer of optics,
in the inherent fascination of all natural who, in time-honored fashion, reduced it
phenomena, even those made familiar by to what he already had considered: inter-
everyday occurrence. ference colors in thin films. Almost two
Our contemplative prehistoric ancestors centuries elapsed before more pieces in
no doubt speculated on the origin of the the puzzle were contributed by the exper-
blue sky, their musings having vanished imental investigations of von Brücke and
into it. Yet it is curious that Aristotle, the Tyndall on light scattering by suspensions
most prolific speculator of early recorded of particles. Around the same time Clau-
history, makes no mention of it in his sius added his bit in the form of a theory
Meteorologica even though he delivered that scattering by minute bubbles causes
pronouncements on rainbows, halos, and the blueness of the sky. A better theory
mock suns and realized that ‘‘the sun was not long in coming. It is associated
looks red when seen through mist or with a man known to the world as Lord
smoke.’’ Historical discussions of the blue Rayleigh even though he was born John
sky sometimes cite Leonardo as the first to William Strutt.
comment intelligently on the blue of the Rayleigh’s paper of 1871 marks the
sky, although this reflects a European bias. beginning of a satisfactory explanation
If history were to be written by a supremely of the blue sky. His scattering law, the
disinterested observer, Arab philosophers key to the blue sky, is perhaps the
would likely be given more credit for most famous result ever obtained by
having had profound insights into the dimensional analysis. Rayleigh argued that
workings of nature many centuries before the field Es scattered by a particle small
their European counterparts descended compared with the light illuminating it
from the trees. Indeed, Möller [1] begins is proportional to its volume V and
his brief history of the blue sky with to the incident field Ei . Radiant energy
Jakub Ibn Ishak Al Kindi (800–870), who conservation requires that the scattered
explained it as ‘‘a mixture of the darkness field diminish inversely as the distance
of the night with the light of the dust and r from the particle so that the scattered
haze particles in the air illuminated by power diminishes as the square of r. To
the sun.’’ make this proportionality dimensionally
Leonardo was a keen observer of light in homogeneous requires the inverse square
nature even if his explanations sometimes of a quantity with the dimensions of
Atmospheric Optics 57
a mixture) per molecule, approximately an ideal gas (to good approximation the
obeys Rayleigh’s law. The form of this atmosphere is an ideal gas), scattering by
expression betrays the incoherence of N molecules is N times scattering by one.
scattering by atmospheric molecules. The This is the only sense in which the blue sky
inverse of β is interpreted as the scattering can be attributed to scattering by fluctua-
mean free path, the average distance a tions. Perfectly homogeneous matter does
photon must travel before being scattered. not exist. As stated pithily by Planck, ‘‘a
To say that the sky is blue because chemically pure substance may be spo-
of Rayleigh scattering, as is sometimes ken of as a vacuum made turbid by the
done, is to confuse an agent with a presence of molecules.’’
law. Moreover, as Young [3] pointed out,
the term Rayleigh scattering has many 2.3
meanings. Particles small compared with Spectrum and Color of Skylight
the wavelength scatter according to the
same law as do molecules. Both can What is the spectrum of skylight? What is
be said to be Rayleigh scatterers, but its color? These are two different questions.
only molecules are necessary for the blue Answering the first answers the second
sky. Particles, even small ones, generally but not the reverse. Knowing the color of
diminish the vividness of the blue sky. skylight we cannot uniquely determine its
Fluctuations are sometimes trumpeted spectrum because of metamerism: A given
as the ‘‘real’’ cause of the blue sky. Pre- perceived color can in general be obtained
sumably, this stems from the fluctuation in an indefinite number of ways.
theory of light scattering by media in which Skylight is not blue (itself an impre-
the scatterers are separated by distances cise term) in an absolute sense. When
small compared with the wavelength. In the visible spectrum of sunlight outside
this theory, which is associated with Ein- the earth’s atmosphere is modulated by
stein and Smoluchowski, matter is taken Rayleigh’s scattering law, the result is a
to be continuous but characterized by a spectrum of scattered light that is nei-
refractive index that is a random function ther solely blue nor even peaked in the
of position. Einstein [4] stated that ‘‘it is blue (Fig. 1). Although blue does not pre-
remarkable that our theory does not make dominate spectrally, it does predominate
direct use of the assumption of a discrete perceptually. We perceive the sky to be
distribution of matter.’’ That is, he cir- blue even though skylight contains light of
cumvented a difficulty but realized it could all wavelengths.
have been met head on, as Zimm [5] did Any source of light may be looked upon
years later. as a mixture of white light and light of
The blue sky is really caused by scat- a single wavelength called the dominant
tering by molecules – to be more precise, wavelength. The purity of the source is
scattering by bound electrons: free elec- the relative amount of the monochromatic
trons do not scatter selectively. Because air component in the mixture. The dominant
molecules are separated by distances small wavelength of sunlight scattered according
compared with the wavelengths of visible to Rayleigh’s law is about 475 nm, which
light, it is not obvious that the power scat- lies solidly in the blue if we take this
tered by such molecules can be added. Yet to mean light with wavelengths between
if they are completely uncorrelated, as in 450 and 490 nm. The purity of this
Atmospheric Optics 59
2.4
Variation of Sky Color and Brightness
2πckT hc
ebλ ≈ , kT, (3)
λ4 λ
where h is Planck’s constant, k is Boltz-
mann’s constant, c is the speed of light
in vacuo, and T is absolute temperature.
Thus, the emission spectrum of a black-
Fig. 2 Even at an altitude of 10 km, well above
body with an infinite temperature has the most particles, the sky brightness increases
same functional form as Rayleigh’s scat- markedly from the zenith to the
tering law. astronomical horizon
60 Atmospheric Optics
with clean water, then add a few drops of path, the optical thickness is about 35 times
milk. The resulting dilute suspension il- greater (Fig. 4), which leads to several
luminated by sunlight has a bluish cast. observable phenomena.
But when more milk is added, the suspen- Even an intrinsically black object is
sion turns white. Yet the properties of the luminous to an observer because of
scatterers (fat globules) have not changed, airlight, light scattered by all the molecules
only their optical thickness: the blue suspen- and particles along the line of sight
sion being optically thin, the white being from observer to object. Provided that
optically thick. this is uniformly illuminated by sunlight
Optical thickness is physical thickness and that ground reflection is negligi-
in units of scattering mean free path, and ble, the airlight radiance L is approxi-
hence is dimensionless. The optical thick- mately
ness τ between any two points connected L = GL0 (1 − e−τ ), (5)
by an arbitrary path in a medium populated
by (incoherent) scatterers is an integral where L0 is the radiance of incident
over the path: sunlight along the line of sight with optical
2 thickness τ . The term G accounts for
geometric reduction of radiance because
τ= β ds. (4)
1 of scattering of nearly monodirectional
sunlight in all directions. If the line of
The normal optical thickness τn of the sight is uniform in composition, τ = βd,
atmosphere is that along a radial path where β is the scattering coefficient and d
extending from the surface of the Earth is the physical distance to the black object.
to infinity. Figure 3 shows τn over the If τ is small (1), L ≈ GL0 τ . In a
visible spectrum for a purely molecular purely molecular atmosphere, τ varies
atmosphere. Because τn is generally small with wavelength according to Rayleigh’s
compared with unity, a photon from law; hence the distant black object in
the sun traversing a radial path in the such an atmosphere is perceived to be
atmosphere is unlikely to be scattered
more than once. But along a tangential
bluish. As τ increases so does L but not transmitted to the observer without be-
proportionally. Its limit is GL0 : The airlight ing scattered out of the line of sight.
radiance spectrum is that of the source of For a long optical path, these two pro-
illumination. Only in the limit d = 0 is cesses compensate, resulting in a hori-
L = 0 and the black object truly black. zon radiance spectrum which is that of
Variation of the brightness and color of the source.
dark objects with distance was called aerial Selective scattering by molecules is not
perspective by Leonardo. By means of it we sufficient for a blue sky. The atmosphere
estimate distances to objects of unknown also must be optically thin, at least for
size such as mountains. most zenith angles (Fig. 4) (the black-
Aerial perspective belongs to the same ness of space as a backdrop is taken
family as the variation of color and for granted but also is necessary, as
brightness of the sky with zenith angle. Leonardo recognized). A corollary of this
Although the optical thickness along a path is that the blue sky is not inevitable: an
tangent to the Earth is not infinite, it is atmosphere composed entirely of nonab-
sufficiently large (Figs. 3 and 4) that GL0 sorbing, selectively scattering molecules
is a good approximation for the radiance of overlying a nonselectively reflecting earth
the horizon sky. For isotropic scattering (a need not be blue. Figure 5 shows calcu-
condition almost satisfied by molecules), lated spectra of the zenith sky over black
ground for a molecular atmosphere with
G is around 10−5 , the ratio of the solid
the present normal optical thickness as
angle subtended by the sun to the solid
well as for hypothetical atmospheres 10
angle of all directions (4π). Thus, the
and 40 times thicker. What we take to
horizon sky is not nearly so bright as
be inevitable is accidental: If our atmo-
direct sunlight.
sphere were much thicker, but identical
Unlike in the milk experiment, what
in composition, the color of the sky
is observed when looking at the hori-
would be quite different from what it
zon sky is not multiply scattered light.
is now.
Both have their origins in multiple scat-
tering but manifested in different ways.
Milk is white because it is weakly ab-
sorbing and optically thick, and hence all
components of incident white light are
multiply scattered to the observer even
though the blue component traverses a
shorter average path in the suspension
than the red component. White horizon
light has escaped being multiply scat-
tered, although multiple scattering is why
this light is white (strictly, has the spec-
trum of the source). More light at the
short-wavelength end of the spectrum is
Fig. 5 Spectrum of overhead skylight for the
scattered toward the observer than at the present molecular atmosphere (solid curve), as
long-wavelength end. But long-wavelength well as for hypothetical atmospheres 10 (dashes)
light has the greater likelihood of being and 40 (dots) times thicker
62 Atmospheric Optics
two orthogonal components into which observed brightness indicates some degree
its electric (or magnetic) field is resolved. of partial polarization.
Completely polarized light corresponds to In the analysis of any scattering prob-
complete correlation; completely unpolar- lem, a plane of reference is required. This
ized light corresponds to no correlation; is usually the scattering plane, determined
partially polarized light corresponds to par- by the directions of the incident and scat-
tial correlation. tered waves, the angle between them being
If an electromagnetic wave is completely the scattering angle. Light polarized perpen-
polarized, the tip of its oscillating electric dicular (parallel) to the scattering plane is
field traces out a definite elliptical curve, sometimes said to be vertically (horizon-
the vibration ellipse. Lines and circles are tally) polarized. Vertical and horizontal in
special ellipses, the light being said to this context, however, are arbitrary terms
be linearly or circularly polarized, respec- indicating orthogonality and bear no rela-
tively. The general state of polarization is tion, except by accident, to the direction
elliptical. of gravity.
Any beam of light can be consid- The degree of polarization P of light
ered an incoherent superposition of two scattered by a tiny sphere illuminated by
collinear beams, one unpolarized, the unpolarized light is (Fig. 6)
other completely polarized. The radiance
of the polarized component relative to 1 − cos2 θ
P= , (12)
the total is defined as the degree of po- 1 + cos2 θ
larization (often multiplied by 100 and where the scattering angle θ ranges from
expressed as a percentage). This can be 0◦ (forward direction) to 180◦ (backward
measured for a source of light (e.g., light direction); the scattered light is partially
from different sky directions) by rotat- linearly polarized perpendicular to the
ing a (linear) polarizing filter and noting scattering plane. Although this equation
the minimum and maximum radiances
transmitted by it. The degree of (linear)
polarization is defined as the difference
between these two radiances divided by
their sum.
3.2
Polarization by Molecular Scattering
airlight is bluish. But if it also is partially scattered field, which in turn is the sum
polarized, its radiance can be diminished of all the fields scattered by the individual
with a polarizing filter. Transmitted cloud- molecules. For an incoherent array we may
light, however, is unpolarized. Because the ignore the wave nature of light, whereas
radiance of airlight can be reduced more for a coherent array we must take it
than that of cloudlight, distant clouds may into account.
change from white to yellow to orange Water vapor is a good example to ponder
when viewed through a rotated polariz- because it is a constituent of air and
ing filter. can condense to form cloud droplets. The
difference between a sky containing water
vapor and the same sky with the same
4 amount of water but in the form of a cloud
Scattering by Particles of droplets is dramatic.
According to Rayleigh’s law, scattering
Up to this point we have considered only an by a particle small compared with the
atmosphere free of particles, an idealized wavelength increases as the sixth power
state rarely achieved in nature. Particles of its size (volume squared). A droplet of
still would inhabit the atmosphere even diameter 0.03 µm, for example, scatters
if the human race were to vanish from about 1012 times more light than does one
the Earth. They are not simply by- of its constituent molecules. Such a droplet
products of the ‘‘dark satanic mills’’ of contains about 107 molecules. Thus,
civilization. scattering per molecule as a consequence
All molecules of the same substance are of condensation of water vapor into
essentially identical. This is not true of a coherent water droplet increases by
particles: They vary in shape and size, about 105 .
and may be composed of one or more Cloud droplets are much larger than
homogeneous regions. 0.03 µm, a typical diameter being about
10 µm. Scattering per molecule in such
4.1 a droplet is much greater than scatter-
The Salient Differences between Particles ing by an isolated molecule, but not to
and Molecules: Magnitude of Scattering the extent given by Rayleigh’s law. Scat-
tering increases as the sixth power of
The distinction between scattering by droplet diameter only when the molecules
molecules when widely separated and scatter coherently in phase. If a droplet
when packed together into a droplet is is sufficiently small compared with the
that between scattering by incoherent and wavelength, each of its molecules is ex-
coherent arrays. Isolated molecules are cited by essentially the same field and
excited primarily by incident (external) all the waves scattered by them inter-
light, whereas the same molecules forming fere constructively. But when a droplet
a droplet are excited by incident light and is comparable to or larger than the wave-
by each other’s scattered fields. The total length, interference can be constructive,
power scattered by an incoherent array of destructive, and everything in between,
molecules is the sum of their scattered and hence scattering does not increase
powers. The total power scattered by a as rapidly with droplet size as predicted by
coherent array is the square of the total Rayleigh’s law.
Atmospheric Optics 67
The figure of merit for comparing different from the water vapor out of which
scatterers of different size is their scatter- it was born that the offspring bears no
ing cross section per unit volume, which, resemblance to its parents. We can see
except for a multiplicative factor, is the scat- through tens of kilometers of air laden
tering cross section per molecule. A scat- with water vapor, whereas a cloud a few
tering cross section may be looked upon tens of meters thick is enough to occult
as an effective area for removing radiant the sun. Yet a rainshaft born out of a
energy from a beam: the scattering cross cloud is considerably more translucent
section times the beam irradiance is the than its parent.
radiant power scattered in all directions.
The scattering cross section per unit 4.2
volume for water droplets illuminated The Salient Differences between Particles
by visible light and varying in size and Molecules: Wavelength Dependence of
from molecules (10−4 µm) to raindrops Scattering
(103 µm) is shown in Fig. 7. Scattering by
a molecule that belongs to a cloud droplet is Regardless of their size and composi-
about 109 times greater than scattering by tion, particles scatter approximately as
an isolated molecule, a striking example the inverse fourth power of wavelength
of the virtue of cooperation. Yet in if they are small compared with the wave-
molecular as in human societies there are length and absorption is negligible, two
limits beyond which cooperation becomes important caveats. Failure to recognize
dysfunctional: Scattering by a molecule them has led to errors, such as that yel-
that belongs to a raindrop is about 100 low light penetrates fog better because
times less than scattering by a molecule it is not scattered as much as light of
that belongs to a cloud droplet. This shorter wavelengths. Although there may
tremendous variation of scattering by be perfectly sound reasons for choosing
water molecules depending on their state yellow instead of blue or green as the
of aggregation has profound observational color of fog lights, greater transmission
consequences. A cloud is optically so much through fog is not one of them: Scat-
tering by fog droplets is essentially in-
dependent of wavelength over the visible
spectrum.
Small particles are selective scatterers;
large particles are not. Particles nei-
ther small nor large give the reverse of
what we have come to expect as nor-
mal. Figure 8 shows scattering of visible
light by oil droplets with diameters 0.1,
0.8, and 10 µm. The smaller droplets
scatter according to Rayleigh’s law; the
larger droplets (typical cloud droplet size)
are nonselective. Between these two ex-
Fig. 7 Scattering (per molecule) of visible light tremes are droplets (0.8 µm) that scatter
(arbitrary units) by water droplets varying in size long-wavelength light more than short-
from a single molecule to a raindrop wavelength. Sunlight or moonlight seen
68 Atmospheric Optics
4.3
The Salient Differences between Particles
and Molecules: Angular Dependence of
Scattering
4.4
The Salient Differences between Particles
and Molecules: Degree of Polarization of
Scattered Light
angle is
1 dT Re
= τn , (14)
T d H
where the scale height and normal opti-
cal thickness may be those for molecules
or particles. Not only do particles, be-
ing more concentrated near the surface,
give disproportionate attenuation of sun-
light on the horizon, but they magnify
the angular gradient of attenuation there.
A perceptible change in color across the
sun’s disk (which subtends about 0.5◦ )
on the horizon also requires the help
of particles.
Fig. 12 Because of scattering by molecules and
particles along the line of sight, each successive
ridge is brighter than the ones in front of it even
5 though all of them are covered with the same
Atmospheric Visibility dark vegetation
with a black backdrop, the contrast is sky assuming a contrast threshold of 0.02
and ignoring the curvature of the earth.
KGL0 exp(−βd) dλ We also observe contrast between ele-
C=− . (17) ments of the same scene, a hillside mottled
KGL0 dλ with stands of trees and forest clearings,
for example. The extent to which we can
resolve details in such a scene depends on
The ratio of integrals in this equation
sun angle as well as distance.
defines an average optical thickness:
The airlight radiance for a nonreflecting
C = − exp(−τ ). (18) object is Eq. (5) with G = p()s , where
p() is the probability (per unit solid angle)
This expression for contrast reduction that light is scattered in a direction making
with (optical) distance is mathematically, an angle with the incident sunlight and
but not physically, identical to Eq. (6), s is the solid angle subtended by the sun.
which perhaps has engendered the mis- When the sun is overhead, = 90◦ ; with
conception that atmospheric visibility is the sun at the observer’s back, = 180◦ ;
reduced because of attenuation. Yet as for an observer looking directly into the
there is no light from a black object to be sun, = 0◦ .
attenuated, its finite visual range cannot The radiance of an object with a finite re-
be a consequence of attenuation. flectance R and illuminated at an angle
The distance beyond which a dark is given by Eq. (11). Equations (5) and (11)
object cannot be distinguished from the can be combined to obtain the contrast be-
horizon sky is determined by the contrast tween reflecting and nonreflecting objects:
threshold: the smallest contrast detectable
Fe−τ
by the human observer. Although this C= ,
depends on the particular observer, the 1 + (F − 1)e−τ
angular size of the object observed, R cos
F= . (20)
the presence of nearby objects, and the πp()
absolute luminance, a contrast threshold
of 0.02 is often taken as an average. This All else being equal, therefore, contrast
value in Eq. (18) gives decreases as p() increases. As shown in
Fig. 9, p() is more sharply peaked in the
− ln |C| = 3.9 = τ = βd. (19) forward direction the larger the scatterer.
Thus, we expect the details of a distant
To convert an optical distance into a scene to be less distinct when looking
physical distance requires the scattering toward the sun than away from it if the
coefficient. Because K is peaked at around optical thickness of the line of sight has
550 nm, we can obtain an approximate an appreciable component contributed by
value of d from the scattering coefficient particles comparable to or larger than the
at this wavelength in Eq. (19). At sea wavelength.
level, the molecular scattering coefficient On humid, hazy days, visibility is
in the middle of the visible spectrum often depressingly poor. Haze, however,
corresponds to about 330 km for ‘‘forever’’: is not water vapor but rather water
the greatest distance at which a black that has ceased to be vapor. At high
object can be seen against the horizon relative humidities, but still well below
Atmospheric Optics 73
100%, small soluble particles in the gas and the scattering cross section σs of a
atmosphere accrete liquid water to become gas molecule:
solution droplets (haze). Although these
droplets are much smaller than cloud n = 1 + 12 αN, (21)
droplets, they markedly diminish visual k4 2
range because of the sharp increase in σs = |α| , (22)
6π
scattering with particle size (Fig. 7). The
same number of water molecules when where N is the number density (not mass
aggregated in haze scatter vastly more than density) of gas molecules, k = 2π/λ is the
when apart. wave number of the incident light, and α
is the polarizability of a molecule (induced
dipole moment per unit inducing electric
6 field). The appearance of the polarizabil-
Atmospheric Refraction ity in Eq. (21) but its square in Eq. (22) is
the clue that refraction is associated with
6.1 electric fields whereas lateral scattering
Physical Origins of Refraction is associated with electric fields squared
(powers). Scattering, without qualification,
Atmospheric refraction is a consequence often means incoherent scattering in all
of molecular scattering, which is rarely directions. Refraction, in a nutshell, is co-
stated given the historical accident that herent scattering in a particular direction.
before light and matter were well un- Readers whose appetites have been
derstood refraction and scattering were whetted by the preceding brief discussion
locked in separate compartments and sub- of the physical origins of refraction are
sequently have been sequestered more directed to a beautiful paper by Doyle [9]
rigidly than monks and nuns in neigh- in which he shows how the Fresnel
boring cloisters. equations can be dissected to reveal the
Consider a beam of light propagating in scattering origins of (specular) reflection
an optically homogeneous medium. Light and refraction.
is scattered (weakly but observably) later-
ally to this beam as well as in the direction 6.2
of the beam (the forward direction). The Terrestrial Mirages
observed beam is a coherent superposi-
tion of incident light and forward-scattered Mirages are not illusions, any more so
light, which was excited by the incident than are reflections in a pond. Reflections
light. Although refractive indices are of- of plants growing at its edge are not
ten defined by ratios of phase velocities, interpreted as plants growing into the
we may also look upon a refractive index water. If the water is ruffled by wind,
as a parameter that specifies the phase the reflected images may be so distorted
shift between an incident beam and the that they are no longer recognizable
forward-scattered beam that the incident as those of plants. Yet we still would
beam excites. The connection between not call such distorted images illusions.
(incoherent) scattering and refraction (co- And so is it with mirages. They are
herent scattering) can be divined from the images noticeably different from what they
expressions for the refractive index n of a would be in the absence of atmospheric
74 Atmospheric Optics
refraction, creations of the atmosphere, shallow surface layers, in which the pres-
not of the mind. sure is nearly constant, the temperature
Mirages are vastly more common than gradient determines the refractive index
is realized. Look and you shall see them. gradient. It is in such shallow layers that
Contrary to popular opinion, they are mirages, which are caused by refractive-
not unique to deserts. Mirages can be index gradients, are seen.
seen frequently even over ice-covered Cartoonists by their fertile imaginations
landscapes and highways flanked by deep unfettered by science, and textbook writers
snowbanks. Temperature per se is not by their carelessness, have engendered
what gives mirages but rather temperature the notion that atmospheric refraction can
gradients. work wonders, lifting images of ships, for
Because air is a mixture of gases, the example, from the sea high into the sky.
polarizability for air in Eq. (21) is an A back-of-the-envelope calculation dispels
average over all its molecular constituents, such notions. The refractive index of air at
although their individual polarizabilities sea level is about 1.0003 (Fig. 13). Light
are about the same (at visible wavelengths). from empty space incident at glancing
The vertical refractive index gradient can incidence onto a uniform slab with this
be written so as to show its dependence on refractive index is displaced in angular
pressure p and (absolute) temperature T: position from where it would have been
in the absence of refraction by
d 1 dp 1 dT
ln(n − 1) = − . (23) δ = 2(n − 1). (24)
dz p dz T dz
This yields an angular displacement of
Pressure decreases approximately ex-
about 1.4◦ , which as we shall see is a rough
ponentially with height, where the scale
upper limit.
height is around 8 km. Thus, the first term
Trajectories of light rays in nonuniform
on the right-hand side of Eq. (23) is around
media can be expressed in different ways.
0.1 km−1 . Temperature usually decreases
According to Fermat’s principle of least
with height in the atmosphere. An average
lapse rate of temperature (i.e., its decrease
with height) is around 6 ◦ C/km. The aver-
age temperature in the troposphere (within
about 15 km of the surface) is around
280 K. Thus, the magnitude of the second
term in Eq. (23) is around 0.02 km−1 . On
average, therefore, the refractive-index gra-
dient is dominated by the vertical pressure
gradient. But within a few meters of the
surface, conditions are far from average.
On a sun-baked highway your feet may
be touching asphalt at 50 ◦ C while your
nose is breathing air at 35 ◦ C, which cor-
Fig. 13 Sea-level refractive index versus
responds to a lapse rate a thousand times wavelength at −15 ◦ C (dashes) and 15 ◦ C (solid
the average. Moreover, near the surface, curve). Data from Penndorf, R. (1957), J. Opt.
temperature can increase with height. In Soc. Am. 47, 176–182 [2]
Atmospheric Optics 75
layer of air. Somewhat above the surface, originating from outside it had to have
however, energy is transported by air in been incident on it from an angle δ below
motion. As a consequence, the tempera- the horizon:
ture gradient steepens toward the ground
if the energy flux is constant. This vari- 2H 2H
δ= − − 2(n − 1), (29)
able gradient can lead to two observable R R
consequences: magnification and multi- where R is the radius of the Earth. Thus,
ple images. when the sun (or moon) is seen to be on
According to Eq. (28), all image points the horizon it is actually more than halfway
at a given horizontal distance are dis- below it, δ being about 0.36◦ , whereas the
placed downward by an amount propor- angular width of the sun (or moon) is
tional to the (constant) refractive index about 0.5◦ .
gradient. A corollary is that the closer Extraterrestrial bodies seen near the
an object point is to a surface, where horizon also are vertically compressed. The
the temperature gradient is greatest, the simplest way to estimate the amount of
greater the downward displacement of the compression is from the rate of change of
corresponding image point. Thus, non- angle of refraction θr with angle of inci-
linear vertical temperature profiles may dence θi for a uniform slab
magnify images.
Multiple images are seen frequently dθr cos θi
= , (30)
on highways. What often appears to dθi n2 − sin2 θi
be water on the highway ahead but
evaporates before it is reached is the where the angle of incidence is that for
inverted secondary image of either the a curved but uniform atmosphere such
horizon sky or horizon objects lighter than that the refracted ray is horizontal. The
dark asphalt. result is
dθr R
6.3 = 1 − (n − 1), (31)
dθi H
Extraterrestrial Mirages
according to which the sun near the
When we turn from mirages of terrestrial horizon is distorted into an ellipse with
objects to those of extraterrestrial bodies, aspect ratio about 0.87. We are unlikely
most notably the sun and moon, we to notice this distortion, however, be-
can no longer pretend that the Earth cause we expect the sun and moon to
is flat. But we can pretend that the be circular, and hence we see them
atmosphere is uniform and bounded. that way.
The total phase shift of a vertical ray The previous conclusions about the
from the surface to infinity is the same downward displacement and distortion of
in an atmosphere with an exponentially the sun were based on a refractive-index
decreasing molecular number density as in profile determined mostly by the verti-
a hypothetical atmosphere with a uniform cal pressure gradient. Near the ground,
number density equal to the surface value however, the temperature gradient is the
up to height H. prime determinant of the refractive-index
A ray refracted along a horizon path gradient, as a consequence of which the
by this hypothetical atmosphere and sun on the horizon can take on shapes
Atmospheric Optics 77
6.4
The Green Flash
of the sun to change. In particular, the theories. For example, coronas are said to
violet-bluish upper rim of the low sun can be caused by diffraction and rainbows by
be transformed to green. refraction. Yet both the corona and the
According to Eq. (29) and Fig. 13, the rainbow can be described quantitatively to
angular width of the green upper rim of high accuracy with a theory (the Mie the-
the low sun is about 0.01◦ , too narrow to ory for scattering by a sphere) in which
be resolved with the naked eye or even to diffraction and refraction do not explicitly
be seen against its bright backdrop. But appear. No fundamentally impenetrable
depending on the temperature profile, the barrier separates scattering from (specu-
atmosphere itself can magnify the upper lar) reflection, refraction, and diffraction.
rim and yield a second image of it, thereby Because these terms came into general
enabling it to be seen without the aid of a use and were entombed in textbooks be-
telescope or binoculars. Green rims, which fore the nature of light and matter was well
require artificial magnification, can be understood, we are stuck with them. But
seen more frequently than green flashes, if we insist that diffraction, for example, is
which require natural magnification. Yet somehow different from scattering, we do
both can be seen often by those who know so at the expense of shattering the unity
what to look for and are willing to look. of the seemingly disparate observable phe-
nomena that result when light interacts
with matter. What is observed depends
7 on the composition and disposition of the
Scattering by Single Water Droplets matter, not on which approximate theory
in a hierarchy is used for quantitative de-
All the colored atmospheric displays that scription.
result when water droplets (or ice crystals) Atmospheric optical phenomena are
are illuminated by sunlight have the same best classified by the direction in which
underlying cause: light is scattered in they are seen and by the agents respon-
different amounts in different directions sible for them. Accordingly, the following
by particles larger than the wavelength, sections are arranged in order of scattering
and the directions in which scattering is direction, from forward to backward.
greatest depends on wavelength. Thus, When a single water droplet is illumi-
when particles are illuminated by white nated by white light and the scattered
light, the result can be angular separation light projected onto a screen, the result
of colors even if scattering integrated over is a set of colored rings. But in the atmo-
all directions is independent of wavelength sphere we see a mosaic to which individual
(as it essentially is for cloud droplets and droplets contribute. The scattering pattern
ice crystals). This description, although of a single droplet is the same as the
correct, is too general to be completely mosaic provided that multiple scattering
satisfying. We need something more is negligible.
specific, more quantitative, which requires
theories of scattering. 7.1
Because superficially different theories Coronas and Iridescent Clouds
have been used to describe different op-
tical phenomena, the notion has become A cloud of droplets narrowly distributed in
widespread that they are caused by these size and thinly veiling the sun (or moon)
Atmospheric Optics 79
probably are caused by droplets. Many of the differential scattering cross section
clouds at temperatures well below freezing at which the conditions
contain subcooled water droplets. Only
dθ b
if a corona were seen in a cloud at a = 0, = 0 (36)
db sin θ
temperature lower than −40 ◦ C could one
assert with confidence that it must be an are satisfied. Missing from Eq. (35) are
ice-crystal corona. various reflection and transmission coeffi-
cients (Fresnel coefficients), which display
7.2 no singularities and hence do not deter-
Rainbows mine rainbow angles.
A rainbow is not associated with rays
In contrast with coronas, which are seen externally reflected or transmitted without
looking toward the sun, rainbows are internal reflection. The succession of
seen looking away from it, and are rainbow angles associated with one, two,
caused by water drops much larger than three . . . internal reflections are called
those that give coronas. To treat the primary, secondary, tertiary . . . rainbows.
rainbow quantitatively we may pretend Aristotle recognized that ‘‘Three or more
that light incident on a transparent sphere rainbows are never seen, because even
is composed of individual rays, each of the second is dimmer than the first, and
which suffers a different fate determined so the third reflection is altogether too
only by the laws of specular reflection and feeble to reach the sun (Aristotle’s view
refraction. Theoretical justification for this was that light streams outward from the
is provided by van de Hulst’s ([12], p. 208) eye)’’. Although he intuitively grasped that
localization principle, according to which each successive ray is associated with
terms in the exact solution for scattering by ever-diminishing energy, his statement
a transparent sphere correspond to more about the nonexistence of tertiary rainbows
or less localized rays. in nature is not quite true. Although
Each incident ray splinters into an infi- reliable reports of such rainbows are rare
nite number of scattered rays: externally (unreliable reports are as common as dirt),
reflected, transmitted without internal re- at least one observer who can be believed
flection, transmitted after one, two, and so has seen one [13].
on internal reflections. At any scattering An incident ray undergoes a total
angle θ, each splinter contributes to the angular deviation as a consequence of
scattered light. Accordingly, the differen- transmission into the drop, one or more
tial scattering cross section is an infinite internal reflections, and transmission out
series with terms of the form of the drop. Rainbow angles are angles of
b(θ) db minimum deviation.
. (35) For a rainbow of any order to exist,
sin θ dθ
The impact parameter b is a sin i , where n2 − 1
i is the angle between an incident cos i = (37)
p(p + 1)
ray and the normal to the sphere. Each
term in the series corresponds to one must lie between 0 and 1, where i is
of the splinters of an incident ray. A the angle of incidence of a ray that gives
rainbow angle is a singularity (or caustic) a rainbow after p internal reflections and
Atmospheric Optics 81
n is the refractive index of the drop. A to the pencil-and-paper variety, are nec-
primary bow therefore requires drops with essarily observed in an atmosphere, the
refractive index less than 2; a secondary molecules and particles of which scat-
bow requires drops with refractive index ter sunlight that adds to the light from
less than 3. If raindrops were composed the rainbow but subtracts from its purity
of titanium dioxide (n ≈ 3), a commonly of color.
used opacifier for paints, primary rainbows Although geometrical optics yields the
would be absent from the sky and we positions, widths, and color separation
would have to be content with only of rainbows, it yields little else. For
secondary bows. example, geometrical optics is blind to
If we take the refractive index of water to supernumerary bows, a series of narrow
be 1.33, the scattering angle for the primary bands sometimes seen below the primary
rainbow is about 138◦ . This is measured bow. These bows are a consequence
from the forward direction (solar point). of interference, and hence fall outside
Measured from the antisolar point (the the province of geometrical optics. Since
direction toward which one must look supernumerary bows are an interference
in order to see rainbows in nature), this phenomenon, they, unlike primary and
scattering angle corresponds to 42◦ , the secondary bows (according to geometrical
basis for a previous assertion that rainbows optics), depend on drop size. This poses
(strictly, primary rainbows) cannot be the question of how supernumerary bows
seen when the sun is above 42◦ . The can be seen in rain showers, the drops
secondary rainbow is seen at about 51◦ in which are widely distributed in size. In
from the antisolar point. Between these a nice piece of detective work, Fraser [16]
two rainbows is Alexander’s dark band, a answered this question.
region into which no light is scattered Raindrops falling in a vacuum are spher-
according to geometrical optics. ical. Those falling in air are distorted by
The colors of rainbows are a conse- aerodynamic forces, not, despite the de-
quence of sufficient dispersion of the pictions of countless artists, into teardrops
refractive index over the visible spectrum but rather into nearly oblate spheroids with
to give a spread of rainbow angles that their axes more or less vertical. Fraser ar-
appreciably exceeds the width of the sun. gued that supernumerary bows are caused
The width of the primary bow from violet by drops with a diameter of about 0.5 mm,
to red is about 1.7◦ ; that of the secondary at which diameter the angular position
bow is about 3.1◦ . of the first (and second) supernumerary
Because of its band of colors arcing bow has a minimum: interference causes
across the sky, the rainbow has become the position of the supernumerary bow
the paragon of color, the standard against to increase with decreasing size whereas
which all other colors are compared. Lee drop distortion causes it to increase with
and Fraser [14, 15], however, challenged increasing size. Supernumerary patterns
this status of the rainbow, pointing out contributed by drops on either side of the
that even the most vivid rainbows are minimum cancel, leaving only the contri-
colorimetrically far from pure. bution from drops at the minimum. This
Rainbows are almost invariably dis- cancellation occurs only near the tops of
cussed as if they occurred literally in a rainbows, where supernumerary bows are
vacuum. But real rainbows, as opposed seen. In the vertical parts of a rainbow, a
82 Atmospheric Optics
horizontal slice through a distorted drop at the end of our journey: the glory. Because
is more or less circular, and hence these it is most easily seen from airplanes it
drops do not exhibit a minimum supernu- sometimes is called the pilot’s bow. Another
merary angle. name is anticorona, which signals that it
According to geometrical optics, all is a corona around the antisolar point. Al-
spherical drops, regardless of size, yield though glories and coronas share some
the same rainbow. But it is not necessary common characteristics, there are differ-
for a drop to be spherical for it to yield ences between them other than direction
rainbows independent of its size. This of observation. Unlike coronas, which may
merely requires that the plane defined by be caused by nonspherical ice crystals, glo-
the incident and scattered rays intersect ries require spherical cloud droplets. And
the drop in a circle. Even distorted a greater number of colored rings may be
drops satisfy this condition in the vertical seen in glories than in coronas because
part of a bow. As a consequence, the the decrease in luminance away from the
absence of supernumerary bows there is backward direction is not as steep as that
compensated for by more vivid colors away from the forward direction. To see
of the primary and secondary bows [17]. a glory from an airplane, look for colored
Smaller drops are more likely to be rings around its shadow cast on clouds be-
spherical, but the smaller a drop, the low. This shadow is not an essential part
less light it scatters. Thus, the dominant of the glory, it merely directs you to the
contribution to the luminance of rainbows antisolar point.
is from the larger drops. At the top of a Like the rainbow, the glory may be
bow, the plane defined by the incident and looked upon as a singularity in the dif-
scattered rays intersects the large, distorted ferential scattering cross section Eq. (35).
drops in an ellipse, yielding a range of Equation (36) gives one set of conditions
rainbow angles varying with the amount of for a singularity; the second set is
distortion, and hence a pastel rainbow. To
the knowledgeable observer, rainbows are sin θ = 0, b(θ) = 0. (38)
no more uniform in color and brightness That is, the differential scattering cross
than is the sky. section is infinite for nonzero impact
Although geometrical optics predicts parameters (corresponding to incident
that all rainbows are equal (neglecting rays that do not intersect the center of the
background light), real rainbows do not sphere) that give forward (0◦ ) or backward
slavishly follow the dictates of this approx- (180◦ ) scattering. The forward direction
imate theory. Rainbows in nature range is excluded because this is the direction
from nearly colorless fog bows (or cloud of intense scattering accounted for by the
bows) to the vividly colorful vertical por- Fraunhofer theory.
tions of rainbows likely to have inspired For one internal reflection, Eq. (38) leads
myths about pots of gold. to the condition
n
7.3 sin i = 4 − n2 , (39)
The Glory 2
which is satisfied only for refractive indices
Continuing our sweep of scattering direc- between 1.414 and 2, the lower refractive
tions, from forward to backward, we arrive index corresponding to a grazing-incidence
Atmospheric Optics 83
When the sun is high in the sky, days a year over a 16-year period, with
hexagonal plates cannot give a sharp halo extremes of 29 and 152 halos a year. Al-
but hexagonal columns – another possible though the 22◦ halo was by far the most
form of atmospheric ice particles – can. frequently seen display, ice-crystal displays
The stable position of a falling column is of all kinds were seen, on average, more
with its long axis horizontal. When the often than once every four days at a loca-
sun is directly overhead, such columns tion not especially blessed with clear skies.
can give a uniform halo even if they all lie Although thin clouds are necessary for ice-
in the horizontal plane. When the sun is crystal displays, clouds thick enough to
not overhead but well above the horizon, obscure the sun are their bane.
columns also can give halos.
A corollary of Fraser’s analysis is that
halos are caused by crystals with a range of 9
sizes between about 12 and 40 µm. Larger Clouds
crystals are oriented; smaller particles
are too small to yield distinct scatter- Although scattering by isolated particles
ing patterns. can be studied in the laboratory, parti-
More or less uniformly bright halos with cles in the atmosphere occur in crowds
the sun neither high nor low in the sky (sometimes called clouds). Implicit in the
could be caused by mixtures of hexagonal previous two sections is the assumption
plates and columns or by clusters of bullets that each particle is illuminated solely
(rosettes). Fraser opines that the latter is by incident sunlight; the particles do not
more likely. illuminate each other to an appreciable de-
One of the by-products of his analysis is gree. That is, clouds of water droplets or
an understanding of the relative rarity of ice grains were assumed to be optically
the 46◦ halo. As we have seen, the angle of thin, and hence multiple scattering was
minimum deviation depends on the prism negligible. Yet the term cloud evokes fluffy
angle. Light can be incident on a hexagonal white objects in the sky, or perhaps an
column such that the prism angle is 60◦ overcast sky on a gloomy day. For such
for rays incident on its side or 90◦ for clouds, multiple scattering is not negligi-
rays incident on its end. For n = 1.31, ble, it is the major determinant of their
Eq. (41) yields a minimum deviation angle appearance. And the quantity that deter-
of about 46◦ for = 90◦ . Yet, although 46◦ mines the degree of multiple scattering is
halos are possible, they are seen much less optical thickness (see Sec. 2.4).
frequently than 22◦ halos. Plates cannot
give distinct 46◦ halos although columns 9.1
can. Yet they must be solid and most Cloud Optical Thickness
columns have hollow ends. Moreover, the
range of sun elevations is restricted. Despite their sometimes solid appearance,
Like the green flash, ice-crystal phenom- clouds are so flimsy as to be almost
ena are not intrinsically rare. Halos and nonexistent – except optically. The fraction
sun dogs can be seen frequently – once of the total cloud volume occupied by
you know what to look for. Neuberger [22] water substance (liquid or solid) is about
reports that halos were observed in State 10−6 or less. Yet although the mass
College, Pennsylvania, an average of 74 density of clouds is that of air to within
Atmospheric Optics 87
a small fraction of a percent, their optical usually translucent, not transparent, yet
thickness (per unit physical thickness) is not completely opaque.
much greater. The number density of air The scattering coefficient of cloud
molecules is vastly greater than that of droplets, in contrast with that of air
water droplets in clouds, but scattering per molecules, is more or less independent
molecule of a cloud droplet is also much of wavelength. This is often invoked as
greater than scattering per air molecule the cause of the colorlessness of clouds.
(see Fig. 7). Yet wavelength independence of scatter-
Because a typical cloud droplet is much ing by a single particle is only sufficient,
larger than the wavelengths of visible light, not necessary, for wavelength indepen-
its scattering cross section is to good dence of scattering by a cloud of particles
approximation proportional to the square (see Sec. 2.4). Any cloud that is optically
of its diameter. As a consequence, the thick and composed of particles for which
scattering coefficient [see Eq. (2)] of a cloud absorption is negligible is white upon
having a volume fraction f of droplets is illumination by white light. Although ab-
approximately sorption by water (liquid and solid) is not
identically zero at visible wavelengths, and
d2 selective absorption by water can lead to
β = 3f , (42)
d3 observable consequences (e.g., colors of
the sea and glaciers), the appearance of all
where the brackets indicate an average but the thickest clouds is not determined
over the distribution of droplet diameters by this selective absorption.
d. Unlike molecules, cloud droplets are Equation (42) is the key to the vastly
distributed in size. Although cloud parti- different optical characteristics of clouds
cles can be ice particles as well as water and of the rain for which they are the
droplets, none of the results in this and the progenitors. For a fixed amount of water
following section hinge on the assumption (as specified by the quantity fh), optical
of spherical particles. thickness is inversely proportional to mean
The optical thickness along a cloud diameter. Rain drops are about 100 times
path of physical thickness h is βh for larger on average than cloud droplets, and
a cloud with uniform properties. The hence optical thicknesses of rain shafts are
ratio d3 /d2 defines a mean droplet correspondingly smaller. We often can see
diameter, a typical value for which is through many kilometers of intense rain
10 µm. For this diameter and f = 10−6 , the whereas a small patch of fog on a well-
optical thickness per unit meter of physical traveled highway can result in carnage.
thickness is about the same as the normal
optical thickness of the atmosphere in the 9.2
middle of the visible spectrum (see Fig. 3). Givers and Takers of Light
Thus, a cloud only 1 m thick is equivalent
optically to the entire gaseous atmosphere. Scattering of visible light by a single
A cloud with (normal) optical thickness water droplet is vastly greater in the
about 10 (i.e., a physical thickness of about forward (θ < 90◦ ) hemisphere than in the
100 m) is sufficient to obscure the disk of backward (θ > 90◦ ) hemisphere (Fig. 9).
the sun. But even the thickest cloud does But water droplets in a thick cloud
not transform day into night. Clouds are illuminated by sunlight collectively scatter
88 Atmospheric Optics
seen against the faint twilight sky may spectral response of the human ob-
be relatively so dark as to seem like server. Also sometimes called photometric
ink blotches. brightness.
Because dark objects of our everyday .5.5
lives usually owe their darkness to absorp-
tion, nonsense about dark clouds is rife: Mirage: An image appreciably different
they are caused by pollution or soot. Yet of from what it would be in the absence of
all the reasons that clouds are sometimes atmospheric refraction.
seen to be dark or even black, absorption
is not among them. Neutral Point: A direction in the sky for
which the light is unpolarized.
Contrast Threshold: The minimum rela- Scattering Angle: Angle between incident
tive luminance difference that can be and scattered waves.
perceived by the human observer.
Scattering Coefficient: The product of scat-
Inferior Mirage: A mirage in which images tering cross section and number density of
are displaced downward. scatterers.
Irradiance: Radiant power crossing unit Scattering Cross Section: Effective area of a
area in a hemisphere of directions. scatterer for removal of light from a beam
by scattering.
Lapse Rate: The rate at which a physical
property of the atmosphere (usually tem- Scattering Plane: Plane determined by
perature) decreases with height. incident and scattered waves.
Luminance: Radiance integrated over the Solar Point: The direction toward the
visible spectrum and weighted by the sun.
90 Atmospheric Optics
Superior Mirage: A mirage in which im- [15] Lee, R. (1991), Appl. Opt. 30, 3401–3407.
ages are displaced upward. [16]∗ Fraser, A. B. (1983), J. Opt. Soc. Am. 73,
1626–1628.
Tangential Optical Thickness: Optical [17] Fraser, A. B. (1972), J. Atmos. Sci. 29, 211,
212.
thickness through the atmosphere along [18]∗ van de Hulst, H. C. (1947), J. Opt. Soc.
a horizon path. Am. 37, 16–22.
[19]∗ Bryant, H. C., Cox, A. J. (1966), J. Opt. Soc.
Am. 56, 1529–1532.
References [20]∗ Nussenzveig, H. M. (1979), J. Opt. Soc.
Am. 69, 1068–1079.
[21]∗ Fraser, A. B. (1979), J. Opt. Soc. Am. 69,
Many of the seminal papers in atmospheric 1112–1118.
optics, including those by Lord Rayleigh, are [22] Neuberger, H. (1951), Introduction to Phys-
bound together in Bohren, C. F. (Ed.) (1989), ical Meteorology. University Park, PA: Col-
Selected Papers on Scattering in the Atmosphere, lege of Mineral Industries, Pennsylvania
Bellingham, WA: SPIE Optical Engineering State University.
Press. Papers marked with an asterisk are in
this collection.
Several monographs on light scattering by a few relevant chapters. Two popular science
particles are relevant to and contain exam- books on simple experiments in atmospheric
ples drawn from atmospheric optics: van physics are heavily weighted toward atmo-
de Hulst, H. C. (1957), Light Scattering by spheric optics: Bohren, C. F. (1987), Clouds in
Small Particles. New York: Wiley-Interscience; a Glass of Beer. New York: Wiley; Bohren, C. F.
reprint (1981), New York: Dover Publica- (1991), What Light Through Yonder Window
tions; Deirmendjian, D. (1969), Electromag- Breaks? New York: Wiley.
netic Scattering on Polydispersions. New York: For an expository article on colors of the sky
Elsevier; Kerker, M. (1969), The Scattering see Bohren, C. F., Fraser, A. B. (1985), Phys.
of Light and Other Electromagnetic Radiation. Teacher 23, 267–272.
New York: Academic Press; Bohren, C. F., An elementary treatment of the coherence
Huffman, D. R. (1983), Light Scattering by properties of light waves was given by For-
Small Particles. New York: Wiley-Interscience; rester, A. T. (1956), Am. J. Phys. 24, 192–196.
Nussenzveig, H. M. (1992), Diffraction Effects This journal also published an expository arti-
in Semiclassical Scattering. Cambridge, UK: cle on the observable consequences of multiple
Cambridge University Press. scattering of light: Bohren, C. F. (1987), Am. J.
The following books are devoted to a wide Phys. 55, 524–533.
range of topics in atmospheric optics: Although a book devoted exclusively to atmo-
Tricker, R. A. R. (1970), Introduction to Mete- spheric refraction has yet to be published,
orological Optics. New York: Elsevier; McCart- an elementary yet thorough treatment of mi-
ney, E. J. (1976), Optics of the Atmosphere. New rages was given by Fraser, A. B., Mach, W. H.
York: Wiley; Greenler, R. (1980), Rainbows, (1976), Sci. Am. 234(1), 102–111.
Halos, and Glories. Cambridge, UK: Cambridge Colorimetry, the often (and unjustly) neglected
University Press. Monographs of more limited component of atmospheric optics, is treated
scope are those by Middleton, W. E. K. (1952), in, for example, Optical Society of Amer-
Vision Through the Atmosphere. Toronto: Uni- ica Committee on Colorimetry (1963), The
versity of Toronto Press; O’Connell, D. J. K. Science of Color. Washington, DC: Optical
(1958), The Green Flash and Other Low Society of America. Billmeyer, F. W., Saltz-
Sun Phenomena. Amsterdam: North Holland; man, M. (1981), Principles of Color Technol-
Rozenberg, G. V. (1966), Twilight: A Study in ogy, (2nd ed.), New York: Wiley-Interscience.
Atmospheric Optics. New York: Plenum; Hen- MacAdam, D. L. (1985), Color Measurement,
derson, S. T. (1977), Daylight and its Spectrum, (2nd ed.), Berlin: Springer.
(2nd ed.), New York: Wiley; Tricker, R. A. R. Understanding atmospheric optical phenom-
(1979), Ice Crystal Haloes. Washington, DC: ena is not possible without acquiring at
Optical Society of America; Können, G. P. least some knowledge of the properties
(1985), Polarized Light in Nature. Cambridge, of the particles responsible for them. To
UK: Cambridge University Press; Tape, W. this end, the following are recommended:
(1994), Atmosphere Halos. Washington, DC: Pruppacher, H. R., Klett, J. D. (1980), Micro-
American Geophysical Union. physics of Clouds and Precipitation. Dor-
Although not devoted exclusively to atmospheric drecht, Holland: D. Reidel. Twomey, S. A.
optics, Humphreys, W. J. (1964), Physics of the (1977), Atmospheric Aerosols. New York:
Air. New York: Dover Publications, contains Elsevier.