Kidder S. Q. & Haar T. H. - Satellite Meteorology (1995)
Kidder S. Q. & Haar T. H. - Satellite Meteorology (1995)
Satellite
Meteorology
AN I N T R O D U C T I O N
Stanley Q. Kidder
T h e U n i v e r s i t y o f A l a b a m a in H u n t s v i l l e
Academic Press
San D i e g o N e w York Boston London Sydney Tokyo Toronto
This b o o k is printed o n acid-free paper.
Copyright © 1 9 9 5 by A C A D E M I C P R E S S
Academic Press
A D i v i s i o n o f Harcourt B r a c e & C o m p a n y
5 2 5 Β Street, Suite 1 9 0 0 , S a n D i e g o , California 9 2 1 0 1 - 4 4 9 5
Library o f C o n g r e s s C a t a l o g i n g - i n - P u b l i c a t i o n D a t a
Kidder, Stanley Q.
Satellite m e t e o r o l o g y : an introduction / b y Stanley Q. Kidder,
T h o m a s H. V o n d e r Haar,
p. cm.
Includes index.
ISBN 0-12-406430-2
1. Satellite m e t e o r o l o g y . I. V o n d e r Haar, T h o m a s Η. II. Title.
QC879.5.K53 1995
551.6'354-dc20 94-42859
CIP
PRINTED IN T H E U N I T E D STATES O F A M E R I C A
98 99 00 MM 9 8 7 6 5 4 3
Preface
During more than three decades since the launch of Vanguard 2, Explorer 7,
and TIROS 1, no textbook has appeared to guide students in the use of what has
become an indispensable tool for meteorology: the weather satellite. This book
is an attempt to remedy this problem at the introductory level. The book is
intended for upper level undergraduates or beginning graduate students who have
a background in calculus and physics and have been introduced to atmospheric
science (meteorology). It can be useful to students in any field which deals with
the atmosphere, in particular with atmospheric measurements.
This is not an "advanced" book. We have included the basics—what we think
every atmospheric scientist should know about satellite meteorology. Measure-
ment of basic meteorological quantities (temperature, wind, etc.) is stressed over
ways to analyze or forecast various features. Our philosophy is that with a
thorough understanding of how the basic quantities are measured and how accu-
rate they are, the reader can proceed to use the data in many types of investigations
and will be able to understand more advanced techniques. Unfortunately this
means that many useful and interesting applications of weather satellite data have
not been included in this book. Some of these advanced applications can be found
in the bibliography at the end of each chapter.
We place special emphasis on the physical understanding of measurements
from space because it is this understanding which will allow both the useful
application of current techniques and the development of future techniques. We
also tend to emphasize operational techniques over experimental techniques. We
do this in the belief that most readers will not do their own information or
ix
x Preface
parameter retrievals from raw satellite data; they will use parameters retrieved
by others. Most often these parameters will be retrieved using operational, near
real-time methods. A thorough exploration of the operational techniques is there-
fore important.
Extensive lists of original references are included at the ends of chapters. There
is no substitute for reading the technical literature, not even this book. One of
our major goals is to enable students to read and understand the literature: articles
on which this book is based, articles which we did not cover, and future articles
which will be written in this rapidly evolving area of atmospheric science.
We would like to acknowledge the research support of NOAA and Colorado
State University, especially the staff at the Cooperative Institute for Research in
the Atmosphere, throughout the period involved in writing this book. In addition,
other research sponsors have provided support necessary to explore the field and
to complete this book. Many people contributed generously to this effort. We
spoke to a substantial fraction of the authors of papers cited and were constantly
amazed and pleased at the time and effort which they were willing to donate.
Many errors and misconceptions were detected and eliminated by these colleagues,
for which we are grateful. In particular, we would like to acknowledge the assis-
tance of the men and women of the National Aeronautics and Space Administra-
tion (NASA), and of the National Oceanic and Atmospheric Administration
(NOAA), without whose help this book would have been impossible. For many
of these scientists and engineers, the daily operation and use of weather satellites
is their profession. We gratefully acknowledge the assistance of Judy Sorbie Dunn,
who expertly drafted the figures; Fredi Boston, our librarian, who found all those
references for us; and Loretta Wilson, our special assistant. We extend special
thanks to our colleague, Donald L. Reinke, for his substantial contributions to
the layout and content of Chapter 5, Image Interpretation. We thank Dr. William
W. Vaughan who carefully read and made valuable comments on the draft.
Finally, we thank and dedicate this book to our wives, Bobbie Poole and Dee
Vonder Haar—and to our children of all ages—whose support and encourage-
ment was matched only by their patience throughout the extended period of au-
thorship.
Stanley Q. Kidder
Thomas H. Vonder Haar
7
Introduction
Ν THE MORE than 30 years since the first meteorological satellites were
/ launched, they have become indispensable for study of the Earth's
atmosphere. Indeed, together with their land- and ocean-sensing cousins, meteoro-
logical satellites view the Earth from a global perspective which is unmatched
and unmatchable by any other observing system. In this book we explore what
has become a very broad field: satellite meteorology. As we explore, we attempt
to reveal the excitement that new observing capabilities bring to science.
To begin, we present an overview of the history and important milestones of
satellite meteorology. This, we hope, will give readers the background necessary
to understand today's satellites and techniques. In the second section of this
chapter, we offer a preview of the remainder of the book to give readers some
perspective on where we are taking them.
1
2 Chapter 1 Introduction
systems (Vaughan, 1982). By the late 1940s, rockets carrying cameras were being
launched into suborbital flights. The photographs that they returned gave rise by
the early 1950s to serious scientific discussion of the possibility of observing the
weather from space (e.g., Wexler, 1954). Several groups, notably the U.S. Army
Evans Signal Laboratory and the University of Wisconsin, pursued the idea of
launching a weather satellite. These efforts were intensified after the launch by
the Soviet Union on 4 October 1957 of the first successful Earth satellite, Sputnik
1. The first successful U.S. satellite, Explorer 1, was launched 123 days later on
31 January 1958. These early days are chronicled in proceedings volumes edited
by Vaughan (1982) and Vonder Haar et al (1982).
Of fundamental importance to space flight, in general, and satellite meteorol-
ogy, in particular, was the formation of the National Aeronautics and Space
Administration (NASA) on 1 October 1958. For more than 30 years, NASA has
lead the development of all types of scientific satellites used for civilian purposes.
Appendix A of this book attempts to list all satellites that have made atmospheric
measurements. A large fraction were developed by NASA. Involved from the first
in satellite meteorology were agencies that now are components of the U.S. Na-
tional Oceanic and Atmospheric Administration (NOAA), particularly the U.S.
Weather Bureau (now the National Weather Service). Today operational U.S.
meteorological satellites are controlled by NOAA and the U.S. Air Force.
The first satellite with a meteorological instrument was Vanguard 2, launched
17 February 1959. Developed by the U.S. Army's Evans Signal Laboratory, Van-
guard 2 had a pair of photocells behind lenses that, much like today's scanning
radiometers, were supposed to sweep out a visible Earth image as the satellite
orbited and spun. Unfortunately, the satellite wobbled on its axis, causing the
scan lines to crisscross, which rendered the data unusable.
Explorer 6, launched 7 August 1959, was the second satellite with meteorologi-
cal instruments. It carried an imaging system and a Suomi radiometer (see below).
It went into a highly elliptical orbit, however, and was essentially unusable,
although it did return the first Earth photo.
The first successful meteorological instrument on an orbiting satellite was the
Suomi radiometer, which flew on Explorer 7, launched 13 October 1959. Devel-
oped by Verner Suomi and colleagues at the University of Wisconsin, it consisted
of hemispheres, painted either black or white, backed by aluminum mirrors, and
mounted on the equator of a spinning satellite. The mirrors reflected the scene
back to the hemispheres, such that the hemispheres acted like spheres isolated in
space. Since the satellite spun, the spheres sampled solar radiation and terrestrial
radiation independent of the orientation of the satellite's spin axis. The tempera-
ture of each hemisphere was monitored, and its time rate of change was related
to the net gain or loss of radiative energy at the sensor. The black hemispheres
absorbed all radiation; the white hemisphere reflected solar radiation but absorbed
infrared radiation. The difference between the radiation balance of the hemi-
spheres indicated solar radiation. With these data, coarse maps of the solar radia-
tion reflected by the Earth and the infrared radiation emitted by the Earth were
made for the first time.
1.1 History of Satellite Meteorology 3
VANGUARD 2
EXPLORER 7
TIROS NIMBUS
FIGURE 1.1. Four early metsats.
4 Chapter 1 Introduction
Section 2.4.1). This regularity increased its utility in operational forecasting. The
sunsynchronous orbit has been used ever since for U.S. operational metsats in
near-polar orbits. Nimbus l's High Resolution Infrared Radiometer (HRIR), a
scanning radiometer quite similar to those in operational use today, provided
night and day coverage. Figure 1.5 shows an HRIR image of Hurricane Gladys.
An important accomplishment of satellite meteorology is that since sometime
in the mid-1960s when metsat coverage became continuous, there have been no
undetected tropical cyclones anywhere on Earth. These ocean-born storms, which
for centuries menaced seafarers and coastal and island dwellers, can no longer
surprise potential victims. Lives are still lost to tropical cyclones, but many are
now saved because of the warnings that metsats make possible.
In total, seven Nimbus satellites were launched. Some experiments on the last
one, Nimbus 7, launched 24 October 1978, were still operational as this book
was being written! The Nimbus series tested many new concepts that have lead
to the operational instruments in use today. These instruments will be discussed
elsewhere in this book.
The first 5 years of satellite meteorology are also documented by Hubert and
Lehr (1967), to which the reader is referred for interesting details to augment the
references noted at the beginning of this section.
By 1966 the United States was ready to initiate an operational (as opposed to
experimental) series of metsats. The Environmental Science Service Administration
(NOAA's predecessor) commissioned nine satellites, ESSA 1 through 9, which
were launched between 3 February 1966 and 26 February 1969. Each was essen-
tially like TIROS 9; each flew in the cartwheel configuration, but in sunsyn-
Similar in appearance to the Nimbus satellites, the Landsat series was designed
for land remote sensing. Its sensors have extremely high resolution, 80 m in the
first satellite and up to 30 m in the latest satellite (Landsat 5). Landsat 1, also
called the Earth Resources Technology Satellite (ERTS), was launched on 23 July
1972. Landsat data are used in meteorology primarily to study small clouds and
surface features that may influence weather.
The first generation of semioperational geostationary metsats began with the
launch of the Synchronous Meteorological Satellite 1 (SMS 1) on 17 May 1974.
SMS 2 was launched on 6 February 1975. These satellites carried the first Data
Collection Platform (DCP) repeater. Data from meteorological or other platforms
on the surface (Fig. 1.6) could be relayed by the satellite to a central receiving
site. Thus data from remote ground sites could be easily obtained for the first
time. The cloud cameras on the ATS satellites made images in the visible portion
of the spectrum only. SMS and the succeeding GOES have an infrared radiometer
as well. Since 27 June 1974, when SMS 1 became operational, we have had
continuous, uninterrupted, 24-hour-per-day monitoring of most of the Western
Hemisphere from space.
On 11 September 1976 the DMSP Block 5D series of the USAF began. The
primary Block 5D instrument is the Operational Linescan System (OLS). It is
still the highest-resolution (600 m) meteorological instrument in space. OLS is
interesting in another way, also. Its shortwave sensor has a broad passband, which
means that it can collect enough radiation to make some images by moonlight, and
it can sense city lights at night. (See Section 5.5.2.4.)
The second series of operational Soviet metsats began on 11 July 1975 with
the launch of Meteor-2 1. Eighteen satellites have been launched in the series.
The first truly operational geostationary metsat, the Geostationary Operational
Environmental Satellite 1 (GOES 1), was launched on 16 October 1975. GOES
2 and 3 were similar. Since the launch of SMS 2, the United States has generally
maintained two geostationary satellites in orbit, one at 75° west longitude, and
one at 135° west longitude.
In 1977 and 1978 two more geostationary metsats were launched: Japan's
Geostationary Meteorological Satellite 1 (GMS 1) was stationed it at 140° east
longitude, and the European Space Agency's Meteosat 1 was stationed at the
prime meridian. Meteosat was the first geostationary satellite to make images of
mid- to upper-troposphere water vapor at 6.7 μπι (Fig. 1.7) in addition to visible
and ΙΟ-12-μπι infrared.
The third generation of U.S. polar-orbiting metsats began on 13 October 1978
with the launch of TIROS N. This series, which continues today, is discussed in
Chapter 4.
India has been quite active in satellite meteorology. Two Indian polar orbiters
have been launched: Bhaskara 1 on 7 June 1979 and Bhaskara 2 on 20 November
1981. On 31 August 1983, the geostationary Insat IB was launched from the
Space Shuttle. Stationed at 74° east longitude, Insat IB completed geostationary
coverage of the tropics and midlatitudes around the Earth. (Although in 1978
and 1979, GOES 1 was temporarily stationed at 55° east longitude to provide
global coverage for the First GARP Global Experiment.) Insat was also the first
three-axis stabilized geostationary metsat.
On 9 September 1980 GOES 4, the first in the second generation of GOES
satellites, was launched. This series of satellites is discussed in detail in Chapter 4.
FIGURE 1.8. ERBS being launched from the Space Shuttle. [Courtesy of NASA.]
1.2 Scope of the Book 11
power necessary to turn raw counts received from the satellite into useful measure-
ments, satellite meteorology would still be in its infancy. Although the reader will
see in this book only glimpses of the communications and computer technology
that underlies all of satellite meteorology, it should not be forgotten.
In the remainder of this chapter, we offer the reader a preview of the contents
and philosophy of this book.
Within the context of related areas of study, Fig. 1.10 denotes the scope of
this basic, introductory book. As shown in the figure, the technical subareas that
constitute the core of satellite meteorology are (1) time and space sampling of
weather and climate features, (2) algorithms and interpretation methods, (3)
satellite senors, and (4) weather and climate products and applications.
Chapters 2 through 4 present basic material necessary to understand satellites
and how they can be used in meteorology. Chapter 2 discusses satellite orbits.
Satellites are not free to travel any path in space, they must follow those dictated
by the laws of physics. Knowledge of these laws and of possible orbits is essential
to understanding satellite meteorology. Chapter 3 discusses electromagnetic radia-
"12 Chapter 1 Introduction
tion, which is the only quantity that meteorological satellites directly measure.
The sources of this radiation and its interaction with the Earth's surface and
atmosphere are explored. Chapter 4 discusses the instruments that make the
measurements. Using instruments on current satellites as examples, the basic
operation and capabilities of meteorological instruments are detailed.
Chapter 5 presents the basics of weather satellite image analysis, that is, inter-
preting the pictures returned by metsats. This was the first, and for many is still
the only, application of meteorological satellite data. Because this subject can
1.2 Scope of the Book 13
(and does) fill volumes, we present only what we think everyone should know
about satellite images. References at the end of the chapter direct the reader to
more detailed discussions.
Chapters 6 through 10 discuss the fundamental parameters that can be retrieved
from meteorological satellite data. We concentrate on three issues: what meteoro-
logical parameters can be retrieved, how the parameters are retrieved from the
electromagnetic measurements, and how accurate the retrieved parameters are.
The parameters covered in Chapters 6 through 10, respectively, are temperature
and trace-gas concentration, winds, clouds and aerosols, precipitation, and radia-
tion budget. Many specialized applications of satellite data are not covered in
this book. We have confined ourselves to the basic parameters in the interest of
brevity, but also in the belief that the fundamental parameters are better able to
be combined with other measurements such as rawinsonde data or radar data to
further our knowledge of the atmosphere.
Finally, in Chapter 11 we look into the future as best we can to indicate what
the next decade may bring to satellite meteorology. The reader is warned that
the technological, political, and economic rates of change are such that making
forecasts of the future of satellite meteorology is a difficult task, yet pleasant. In
planning one's career, or indeed one's next project, however, it is necessary have
some idea of what the future holds.
Now, dear reader, welcome to the world of satellite meteorology. We hope
that you enjoy the following tour.
Bibliography
Hubert, L. F., and P. E. Lehr (1967). Weather Satellites. Blaisdell Publishing Co., Waltham, MA.,
120 pp.
Rao, P. K., S. J. Holmes, R. K. Anderson, J. S. Winston, and P. E. Lehr (eds.) (1990). Weather
Satellites: Systems, Data, and Environmental Applications. American Meteorological Society,
Boston.
Vaughan, W. W. (ed.) (1982). Meteorological Satellites—Fast, Present, and Future. N A S A Conference
Publication 2 2 2 7 , Washington, D C , 60 pp.
Vonder Haar, Τ. Η., W. W. Vaughan, Μ. H. Davis, and M. A. Cook (eds.) (1982). The Conception,
Growth, Accomplishments, and Future of Meteorological Satellites. N A S A Conference Publication
2 2 5 7 , Washington, D C , 101 pp.
Wexler, H. (1954). Observing the weather from a satellite vehicle. / . Br. Interplanetary Set., 1 3 ,
269-276.
2
Orbits
and
Navigation
Isaac Newton discovered the basic principles that govern the motions of
1
1
English physicist and mathematician, 1 6 4 2 - 1 7 2 7 .
15
16 Chapter 2 Orbits and Navigation
F = ma = ~j~> m
(2.1)
a distance r is
F=^ » h , ( 2 . 2 )
satellite. The force of gravity that supplies this centripital force is Gm rn/r , where e
2
m is the mass of the Earth (Appendix E) and m is the mass of the satellite.
e
r r L
SATELLITE
Division by m eliminates the mass of the satellite from the equation, which means
that the orbit of a satellite is independent of its mass. The period of the satellite
is the orbit circumference divided by the velocity: Τ = lirrlv. Substituting in Eq.
2.3 gives
4-7Γ 2
Gm e
The current NOAA satellites orbit at approximately 850 km above the Earth's
surface. Since the equatorial radius of the Earth is about 6378 km, the orbit
2
radius is about 7228 km. Substituting in Eq. 2.4 shows that the NOAA satellites
have a period of about 102 min.
As a second example, we calculate the radius required for a satellite in geosyn-
chronous orbit, that is an orbit in which the satellite has the same angular velocity
as the Earth. The angular velocity of a satellite is
2π
ξ =~ (2.5)
^ =ψ · (2.6)
Inserting the angular velocity of the Earth (Appendix E), the required radius for a
geosynchronous orbit is 42,164 km, or about 35,786 km above the Earth's surface.
have the advantage of Newton's work. Kepler formulated his laws by analyzing
a mass of data on the position of the planets. This task was complicated by the
rotation of the Earth and the motion of the Earth about the sun, which make
2
Specifications call for them to orbit at either 833 or 870 km; 8 5 0 km is a representative value.
3
The reader is referred to Escobal (1965) and Goldstein (1950) for two quite different, but equally
lengthy, derivations.
4
German astronomer, 1 5 7 1 - 1 6 3 0 .
Chapter 2 Orbits and Navigation
18
planetary motions seem very complex. In modern form, Kepler's laws may be
stated as follows:
Kepler's Laws
1. All planets travel in elliptical paths with the sun at one focus.
2. The radius vector from the sun to a planet sweeps out equal areas in
equal times.
3. The ratio of the square of the period of revolution of a planet to the
cube of its semimajor axis is the same for all planets revolving around
the sun.
The same laws apply if we substitute satellite for planet and Earth for sun.
Equation 2.4 is a statement of Kepler's third law for the special case of a circu-
lar orbit.
The parameters that are used to specify satellite orbits are based in part on
geometric terminology. Figure 2.2 illustrates the geometry of an elliptical orbit.
The point where the satellite most closely approaches the Earth is the perigee, or
more generally, the perifocus. The point where the satellite is furthest from the
Earth is called the apogee or apofocus. The distance from the center of the ellipse
to the perigee (or apogee) is the semimajor axis and will be denoted by the symbol
a. The distance from the center of the ellipse to one focus (to the center of the
Earth) divided by the semimajor axis is the eccentricity and will be denoted by
the symbol ε. For an ellipse, the eccentricity is a number between zero and one
(0 < ε < 1). A circle is an ellipse with zero eccentricity. The equation for the
ellipse, that is, the path that the satellite follows, is given in polar coordinates
SATELLITE
APOGEE PERIGEE
(APOFOCUS) (PERIFOCUS)
1 + ε cos0
The angle θ is the true anomaly and is always measured counterclockwise (the
direction of satellite motion) from the perigee.
Μ = n(t - t ) = e - ε sine,
p (2.8)
where Μ is the mean anomaly; Μ increases linearly in time at the rate n, called
the mean motion constant, given by
n=
Y=J¥- ·
(2 9)
(2.10a)
(2.10b)
FIGURE 2.3. The geometric relationship between true anomaly (0) and eccentric anomaly {e).
Chapter 2 Orbits and Navigation
with the Earth's spin axis. The χ axis is chosen such that it points from the center
of the Earth to the sun at the moment of the vernal equinox, when the sun is
crossing the equatorial plane from the Southern Hemisphere to the Northern
Hemisphere. The y axis is chosen to make it a right-handed coordinate system.
6
5
Because the origin of this coordinate system moves about the sun with the Earth, it is not truly
inertial. However, the sun's gravity causes the satellite to rotate around the sun as does the Earth.
Therefore, the satellite acts as if the right ascension-declination coordinate system were inertial.
6
This χ axis is also referred to as the First Point of Aries because it used to point at the constellation
Aries. Because of the influence of the sun and m o o n on the nonspherical Earth, the Earth's spin axis
precesses like a top with a period of 2 5 , 7 8 1 years. This causes the vernal equinox to change. Today,
the χ axis points to the constellation Pisces, but it is still referred to as the First Point of Aries.
2.2 Keplerian Orbits 21
Ζ
CO
X
FIGURE 2.4. The right ascension—declination coordinate system.
X
FIGURE 2.5. Coordinates used in the right ascension-declination coordinate system: right ascension
(Ω), declination (δ), and radius (r).
EARTH'S
NODE
FIGURE 2.6. Angles used to orient an orbit in space.
22 Chapter 2 Orbits and Navigation
final parameter, included in Table 2.1, is the time when these elements are observed
or are "valid." This time is called the epoch time (t ). Some of the orbital elements
Q
change with time, as we shall see below. A subscript " o " on an orbital element
indicates a value at the epoch time.
There is some variation in how the orbital elements are specified. ESA, for
example, substitutes true anomaly for mean anomaly. Also, in less formal descrip-
tions of satellite orbits, one frequently sees the height of the satellite above the
Earth's surface substituted for the semimajor axis. Since the Earth is not round,
the height of a satellite will vary according to its position in the orbit. Specifying
the semimajor axis is a much better way to describe a satellite orbit.
Orbits in which the classical orbital elements (except M) are constant are called
Keplerian orbits. Viewed from space, Keplerian orbits are simple. The satellite
moves in an elliptical path with the center of the Earth at one focus. The ellipse
maintains a constant size, shape, and orientation with respect to the stars (Fig
2.7a). Perhaps surprisingly, the only effect of the sun's gravity on the satellite is
to move the focus of the ellipse (the Earth) in an elliptical path around the sun
(the Earth's orbit).
Viewed from the Earth, Keplerian orbits appear complicated because the Earth
rotates on its axis as the satellite orbits the Earth (Fig. 2.8). The rotation of the
Earth beneath a fixed orbit results in two daily passes of the satellite near a point
on the Earth (assuming that the period is substantially less than a day and that
the inclination angle is greater than the latitude of the point). One pass occurs
during the ascending portion of the orbit; the other occurs during the descending
portion of the orbit. This usually means that one pass occurs during daylight and
one during darkness.
Although satellites travel in nearly Keplerian orbits, these orbits are perturbed
by a variety of forces (Table 2.2). Forces arising from the last five processes are
small and can be viewed as causing essentially random perturbations in the orbital
elements. Operationally they are dealt with simply by periodically (1) observing
N O A A orbital elements for the polar-orbiting satellites are broadcast in the form of "TBUS
7
Element Symbol
Semimajor axis a
Eccentricity ε
Inclination i
Argument of perigee ω 0
Mean anomaly M 0
Epoch time 0t
FIGURE 2.7. The change with season of (a) a Keplerian orbit and (b) a sunsynchronous orbit.
24 Chapter 2 Orbits and Navigation
FIGURE 2.8. The orbit of a representative satellite as viewed from a point rotating with the Earth.
the orbital elements and (2) adjusting the orbit with on-board thrusters. Forces
due to the nonspherical Earth cause secular (linear with time) changes in the
orbital elements. These forces can be predicted theoretically and indeed are useful.
The gravitational potential of Earth is a complicated function of the Earth's
shape, the distribution of land and ocean, and even the density of crustal material.
As a first-order correction to a spherical shape, we may treat the Earth as an
oblate spheroid of revolution. In cross section the Earth is approximately elliptical.
The distance from the center of the Earth to the equator is, on average, 6378.140
km, whereas the distance to the poles is 6356.755 km. One can think of the Earth
as a sphere with a 21-km-thick "belt" around the equator. The gravitational
potential of the Earth is approximately given by
( 1 - 3 sin 6) + - , (2.11)
17= "
2
where r is the equatorial radius of the Earth, δ is the declination angle, and
ee
Force Source
terms are more than two orders of magnitude smaller than the quadrupole term
and will not be considered here, although they are necessary for very accurate
calculations.
How does this belt of extra mass affect a satellite's orbit? One might expect
it to cause the satellite to orbit at a different speed, and indeed it does. The time
rate of change of the mean anomaly (dM/dt) is given by the mean motion constant
η in the unperturbed orbit and by the anomalistic mean motion constant, n, in
a perturbed orbit. Considering only the quadrupole term, Escobal (1965)
shows that
dM _
—-— = η — η (2.12)
dt
When the inclination angle is less than 54.7°, η is greater than n; the satellite
orbits faster than it would in an unperturbed orbit. However, for larger inclina-
tions, the satellite orbits more slowly than it otherwise would.
Because the belt exerts an equatorward force, one might also expect that it
would have an effect on the inclination angle. This force, however, affects the
right ascension of the ascending node rather than the inclination angle. Just as
the force of gravity causes a top to precess rather than to fall over, so the attraction
of the belt causes the orbit to precess about the ζ axis rather than to change its
inclination angle. Escobal (1965) gives the rate of change of the right ascension
of ascending node as
dCi
—τ- = -n f / ( ^ ) ( l - e ) - c o s / ] .
2 2 2
(2.13)
dt 2
The final effect of the belt is to cause the argument of perigee to rotate or
precess. Escobal (1965) gives
άω _ 3 (r, x2
/
T
^ j ( l - e ^ ( - f s m ^ ) ] . (2.14)
dt
2 2 2
The other three orbital elements, α, ε, and /, undergo small, oscillatory changes
that may be neglected.
If SI Units are used, Eqs. 2.9, 2.12, 2.13, and 2.14 result respectively in values
of n, n, d£lldt, and άω/dt whose units are radians per second.
The anomalistic period of a perturbed orbit is simply
— 27Γ
Τ
= ψ·
η (2.15)
However, because Μ is measured from perigee, the anomalistic period is the time
for the satellite to travel from perigee to moving perigee. Of more use is the
synodic or nodal period, T, which is the time for the satellite^ to travel from one
ascending node to the next ascending node. An exact value of Τ must be calculated
2£ Chapter 2 Orbits and Navigation
(2.16)
Nearly all present meteorological satellites are in one of two orbits, sunsynchro-
nous or geostationary, but other orbits are also useful.
tion of the satellite is correctly chosen, the right ascension of its ascending node
can be made to precess at this same rate. An orbit that is so synchronized with
the sun is called a sunsynchronous orbit. For a satellite with a semimajor axis of
7228 km and zero eccentricity, Eq. 2.13 requires an inclination of 98.8° to be
sunsynchronous. Figure 2.7b shows the change with season of a sunsynchro-
nous orbit.
Because the sun-Earth-ascending node angle is constant in a sunsynchronous
8
orbit, the satellite is often said to cross the equator at the same local time every
Apart from small changes due to the elliptical orbit of the Earth.
2.4 Meteorological Satellite Orbits 27
L T
"' U +
TF ( 2
· 1 7 )
of a particular point. Equator crossing time (ECT) is the local time when a
9
Y s u n = -15°(iu-12) (2.19)
9
The other use of local time refers to the time on one's watch, that is, the time in a particular
time zone. Time zones are defined as areas where time is agreed to be the local time (in our sense)
on a particular meridian. Eastern Standard Time, for example, is the local time on the 75° west
meridian (Ψ = - 7 5 ° ) .
28 Chapter 2 Orbits and Navigation
NOAA II
T h r e e Orbits on 2 2 March 1 9 9 0
Start time: 0 2 5 8 UTC End time: 0 8 0 4 UTC
a = 7 2 2 9 . 6 0 6 km Λ = 29.31059°
0
i = 98.97446° ω = 167.74754 °
0 .
€ = 0.00119958 Epoch time = 2 2 Mar 1 9 9 0 Γ I 5 5 2 . 3 5 3 UTC
M S
M = 192.28166°
0 Nodal Period = 1 0 2 . 0 7 6 4 m i n
FIGURE 2.9. The ground track of a typical sunsynchronous satellite.
GOES 6
(One Orbit on 6 March 1 9 9 0 )
3°N 1 1 • 1
0900
2° Ν
APOGEE 0 6 2 0
1200 (42185.12 km)
0600
ΓΝ
^ / A S C E N D I N G NODE
I50O
EQ 1 1 1
0300
l°S
/
DESCENDING NODE
1800«
loooo
PERIGEE 1 8 1 8 ^ ^
2°S (42158.47km) ;
)0
2|
3°S 1 1 1
a = 42171.798 km i l = 80.259°
0
i=l.973IO° ω = 223.891°
0
»1
M = 9ÄiiS!SP
0
=
147 0 2 0
E
P° , p
N o d a
= |
c h t i m e
e r i o d
6
s
M o r 1 9 9 0
4 3 6
4 h
8 2 0 UTC
297min
m s
FIGURE 2.10. The ground track of a geostationary satellite. N o t e that the satellite's orbit is not quite
geostationary; it drifts west slightly each day.
M = M + ^-(t-t ),
a 0 (2.21a)
Ω= Ω +^(ί-ί ),
0 0 (2.21b)
ω= ω +^(ί-ί ).
0 0 (2.21c)
Then the satellite is positioned by one of several methods. We find two methods
useful: the vector rotation method and the spherical geometry method.
15.11 The Vector Rotation Method
Figure 2.11 illustrates what we call the vector rotation method. It is discussed
in a somewhat different form by Escobal (1965) and others. In the first step, the
satellite is located in the plane of its orbit; that is, the true anomaly Θ and the
radius r are calculated. This is done by (1) solving for e using Eq. 2.8, (2) calculating
θ using Eq. 2.10a, and (3) calculating r using Eq. 2.7. (For a circular orbit, this
step is simplified because the mean anomaly, the eccentric anomaly, and the true
anomaly are identical, and r is constant.)
In the second step, a vector is formed that points from the center of the Earth
to the satellite in the right ascension-declination coordinate system. The Cartesian
Such bulletins are available from a variety of sources. Because these sources change rapidly, we
1 0
suggest that the interested reader contact the agencies listed in Section 4.4 to find a convenient source
of satellite bulletins.
2.5 Satellite Positioning, Tracking, and Navigation 31
a b
FIGURE2.il. Rotations used to position a satellite in its orbit: (a) the satellite in the plane of its orbit,
(b) rotation about the ζ axis through the argument of perigee (ω), (c) rotation about the χ axis through
the inclination angle (i), and (d) rotation about the ζ axis through the right ascension of ascending
node (Ω).
( \ ( ^\
X T COS
(2.22)
At this point, the orbital ellipse is assumed to lie in the x—y plane with the perigee
on the positive χ axis (Fig 2.11a).
In the final three steps, the vector is rotated so that the orbital plane is properly
oriented in space.
In the third step, the vector is rotated about the ζ axis through the argument
of perigee (Fig. 2.11b). This rotation is conveniently accomplished by multiplying
32 Chapter 2 Orbits and Navigation
(
cosft -sinn 0 \ /x"\ ix" cosü - y" sinilX
sinü cosft 0 J I y" \ = x" sinil + y" cosü ] . (2.25)
0 0 i / \ * 7 \ z
" /
The vector (*'", y " ^" ) is the location of the satellite in the right ascension-
,
5
,
declination coordinate system at time t. This vector may be converted into the
radius, declination, and right ascension of the satellite by
Vx"' + / " 2 2
+ *"' = r, 2
(2.26a)
δ - sin" [j-J
5
1
, (2.26b)
n ^ t a n " 1
^ - (2.26c)
After one has calculated the right ascension, declination, and radius of the
satellite, it is useful to calculate the latitude and longitude of the subsatellite point.
Assuming that the Earth is a sphere, the latitude (known as the geocentric latitude)
is simply equal to the declination. The longitude of the subsatellite point is the
difference between the right ascension of the satellite and the right ascension of
the prime meridian (0° longitude) which passes through Greenwich, England (Fig.
2.12). The right ascension of Greenwich can be calculated knowing its right
ascension at a given time and the rotation rate of the Earth. Since the rotation 11
rate changes very slightly, due to the actions of the wind and ocean currents, very
accurate knowledge of the right ascension of Greenwich requires observations.
Some satellite bulletins give the right ascension of Greenwich in addition to the
satellite orbital elements.
If nothing else is available, one can use the following: at 0 0 0 0 UTC on 1 January 1 9 9 0 the
1 1
right ascension of Greenwich was 100.38641°, and the rotation rate was 7 . 2 9 2 1 1 5 9 2 2 x 1 0 ~ radians 5
PRIME
MERIDIAN
VERNAL
EQUINOX
FIGURE 2.12. The relationship between Earth longitude and right ascension.
The inverse problem of finding when a satellite passes over (or close to) a
particular point is solved iteratively by (1) estimating the time, (2) calculating the
position of the satellite, and (3) correcting the time estimate. Steps 2 and 3 are
repeated until a satisfactory solution is found.
r = r,
s (2.28a)
Θ = 6 = sin (sinr sin/),
5 S
_1
(2.28b)
the epoch time. The quantity (diljdt — άΩ,/dt) is the relative Earth rotation
rate, that is, the rotation rate of the Earth relative to the orbital plane. For a
sunsynchronous satellite, it must be exactly 2π radians per day.
Normally the arctangent term would be written t a n ( t a n r c o s / ) . The form in Eq. 2.28c is used
1 2 _1
because it allows the quadrant of the angle to be determined unambiguously. In Fortran, for example,
ATAN2(sinrcos/,cosr) will result in the correct angle.
34 Chapter 2 Orbits and Navigation
For a circular orbit, or one which is so nearly circular that no significant error
occurs from neglecting its elliptical nature,
and the spherical geometry method is particularly easy to apply. Polar orbiters
can often be treated with this approximation. When they are so treated, the orbital
parameters may come in a different form. The supplied parameters may be (1)
the longitude of ascending node, (2) the nodal period, (3) the radius (or semimajor
axis), (4) the inclination, (5) the time of ascending node, and (6) the nodal longitude
increment (ALON), which is the difference in longitude between successive 13
ascending nodes:
i L 0
^(f-f) - f (2.30)
where Τ is the nodal period. The above equations still apply, but one must
remember that Γ = 0, [η + άω/dt) = 2π/Τ, and {düjdt - dill dt) = A L O N / f .
0
2.5.2 Tracking
A list of time versus position of a celestial body is called an ephemer is (plural:
ephemer ides). To track a satellite, one must be able to point one's antenna at it.
The elevation angle, measured from the local horizontal, and the azimuth angle,
measured clockwise from the north, can be calculated from the ephemeris data
as follows.
Suppose the subsatellite point is at latitude Θ and longitude Ψ and that the 5 85
satellite is at radius r from the center of the Earth. Suppose also that the antenna
s
(
χλ I r c o s 0 cosYA
s s
y $ = r cos@ sinY
s s s , (2.31)
zj \ r sin@
s s /
(
*e\ / t C O S 0 COS^PA
r
e
y e = r cos© sh^
e e e . (2.32)
ζJ \ r sin@
e e /
The difference vector (r = r — r ) points from the antenna to the satellite (Fig.
D % e
2.13). Assuming a spherical Earth, the vector r points to the local vertical (Fig. e
13
That is, the next ascending node occurs ALON west of the current ascending node.
2.5 Satellite Positioning, Tracking, and Navigation 35
2.14). The cosine of the satellite's zenith angle ζ (the complement of the elevation
angle) is given by
COS£ = £ T ^ - (2.33)
Finding the azimuth angle is a little more difficult. First, we need to find two
vectors in the tangent plane at the antenna. The first points north:
-sin@ cos^ N e e
-sin@ sinY | .
e e (2.34)
cos@ e
FIGURE 2.14. Definition of zenith angle (ζ) and azimuth angle (ψ).
36 Chapter 2 Orbits and Navigation
directions of f and r as
t D
(2.35a)
(2.35b)
cos ψ = ρ (2.37)
One must be careful when taking the inverse cosine. If the satellite is west of the
antenna, ψ will be greater than 180°. It also must be noted that these equations
assume a spherical Earth. Fortunately, most receiving antennas are insensitive to
the slight errors this assumption causes.
2.5.3 Navigation
In addition to knowing where a satellite is in its orbit, it is necessary to know
the Earth coordinates (latitude, longitude) of the particular scene it is viewing.
The problem of calculating the Earth coordinates is known as the navigation
problem; fundamentally, it is a complex geometry problem. It requires an accurate
knowledge of where the satellite is in its orbit, the orientation of the satellite (its
attitude), and the scanning geometry of the instrument involved.
In simplified form, we can proceed as follows. Suppose that at a particular
time a satellite is at position (# ,y ,£ ) with respect to the center of the Earth
s s s
(2.38)
Figure 2.15 shows that the ray from which the telescope receives radiation (that
is, the line in space through the satellite and in the direction of the telescope) is
given by
(2.39)
The location at which this ray strikes the spherical Earth is the solution of
the equation
(x + s xj)
s
2
+ (y + s y )
s T
2
+ (z + s z )
s T
2
= r\. (2.40)
This is a quadratic equation in s that has no real roots, if the ray does not intersect
the Earth; or two real roots, if it does. The smaller root is to be chosen; the larger
root represents the location from which the ray reemerges from the opposite side
of the Earth. When the ray is just tangent to the Earth, the two roots are equal. 14
After a solution for s has been found, Eq. 2.39 gives the Cartesian coordinates
in the right ascension-declination coordinate system of the point on the Earth's
surface being viewed. The latitude and longitude can then be found as in Sec-
tion 2.5.1.
Satellite images are usually the result of a scanning instrument. The data come
in the form of scan lines, each divided into elements or samples known as pixels
or scan spots. Because scanning is very carefully timed, each pixel has a unique
time associated with it. Therefore, calculating the latitude and longitude of a pixel
is accomplished using the equations of Sections 2.5.1 and 2.5.3 in the forward
direction; time yields satellite position and telescope pointing angles, which then
yield latitude and longitude. The opposite problem, finding the pixel which ob-
served a particular point on Earth (latitude and longitude), must be solved in an
iterative manner because the exact time when the point was observed is unknown.
In brief, the time of observation is estimated, the actual point being observed at
that time is calculated, and a correction is made in the time which moves the
point of observation closer to the desired point. This procedure is iterated until
satisfactory convergence is achieved.
The scheme outlined here for finding latitude and longitude is simple and very
general. It is applicable to a wide variety of satellite orbits and instruments. For
1 4
For a geostationary satellite, this occurs about 81° from the satellite subpoint and explains why
geostationary satellites never observe the poles.
33 Chapter 2 Orbits and Navigation
each instrument, the difficult part is to determine the telescope pointing angles
δ and Ω .
τ χ
ary satellites.
Unfortunately, the satellite's spin axis is not exactly parallel to the Earth's spin
axis. Furthermore, although the radiometer's telescope is rigidly oriented with
respect to the principal axis of the satellite, the spin axis deviates slightly from
the principal axis, which causes deviations similar to pitch, roll, and yaw in a
low Earth orbiter (see next section). Corrections for these effects and for the non-
spherical Earth can be made. The interested reader is referred to Hambrick and
Phillips (1980).
The parameters that describe the satellite orbit and attitude must be accurately
known to perform accurate navigation. These parameters can be determined by
the use of landmarks. Normally the orbit and attitude parameters are accurate,
as is navigation performed with them. However, for up to 18 h after the thrusters
are fired in an orbit- or attitude-correcting maneuver, navigation parameters are
poorly known, and pixels can be significantly misplaced. These errors can be
partially corrected by displaying the data as an image and shifting the image up
or down and right or left until a landmark is properly positioned. Rotation of
the image is sometimes necessary to achieve good navigation, especially if a large
(continent-size) area is being studied.
It is interesting to note that the GOES satellites can detect a few stars at the
edges of the image frame. These stars can be used to very accurately determine
the attitude of the satellite. Then landmarks can be used to determine the orbital
elements (Hambrick and Phillips, 1980).
2.5.3.2 Low Earth Orbit Geometry
Low Earth orbit satellite instruments have many scanning patterns. Navigation
of these data can be achieved using different approachs. We outline an approach,
based on the discussion above, which is general enough for use with many scanning
patterns. The basis of the technique is that if we can determine where a scan spot
is in relation to the satellite, then we can use nearly the same rotation matrices
with which we position the satellite to position the scan spot. First we must define
the angles and a coordinate system used to specify satellite attitude.
The instruments on many low Earth orbit satellites are mounted on the under-
side of the satellite and scan perpendicular to the velocity vector through the
subsatellite point (see Chapter 4). A convenient coordinate system (Fig. 2.16) is
one in which the ζ axis points from the satellite toward the center of the Earth,
2.5 Satellite Positioning, Tracking, and Navigation 39
the χ axis points in the direction of satellite motion, and the y axis is chosen to
complete a right-handed coordinate system. Three angles specify the orientation
of the satellite in this coordinate system. Rotation about the y axis is called pitch,
rotation about the χ axis is called roll, and rotation about the ζ axis is called yaw.
A combination of these angles can be used to specify nearly any scan geometry.
Instruments that scan through nadir perpendicular to the satellite motion vector
are described by changing the roll angle. Instruments that scan in a cone can be
described by a constant pitch plus a variable yaw. Instruments that scan through
the subpoint but at an oblique angle with respect to the satellite motion vector
can be described with a roll plus a constant yaw.
To calculate the position of a scan spot with respect to the satellite, we proceed
as follows. First, position the satellite at radius r , declination zero, and right
s
ascension zero, and let its velocity vector point east. Assume that the telescope
is pointing straight down, or δ = 0 and Ω = π. That is, the telescope pointing
15
χ τ
pitch, roll, and yaw angles that describe the position of the telescope at time i,
1 5
If the satellite is not pointing straight d o w n , its deviation is described by pitch, roll, and yaw
bias errors, which are usually small. For example, horizon sensors and a sun sensor on the current
N O A A satellites maintain pitch, roll and y a w bias errors to within ± 0 . 2 ° of zero. If the bias errors
are known, and if the desired precision of the calculation requires it, the initial telescope pointing
vector can be corrected for bias errors at this point.
40 Chapter 2 Orbits and Navigation
or for the scan position desired. At the assumed position and orientation of the
satellite, in the right ascension-declination coordinate system, the pitch rotation
matrix is
cosctp sinctp 0^
—sina P cosa P 0
0 0 1/
where a is the pitch angle; the roll rotation matrix is
?
'cosa: R 0 -sina^
0 1 0
v sina: R 0 cosa R
' 1 0 0
0 cosa Y — sina Y
^0 sina Y cosa Y
After the telescope pointing vector has been determined, the distance s from
the satellite to the scan spot is calculated using Eq. 2.40, and the position of the
spot relative to the satellite is calculated using Eq. 2.39.
Finally, the scan spot is moved along with the satellite to its actual position
by (1) rotating about the ζ axis through the argument of latitude, (2) rotating
about the χ axis through the inclination angle, and (3) rotating about the ζ axis
through the right ascension of the ascending node minus the right ascension
of Greenwich.
An advantage to this method is that if the orbit is sufficiently circular, the
vectors to the scan spots can be calculated in advance and simply rotated into
position at successive times.
Note that limb scanners, which scan the atmosphere above the Earth's horizon,
can be treated with this procedure except that Eq. 2.40 is not applicable because
the ray does not strike the Earth. Instead, the distance to the tangent point, that
is, the point where the ray most closely approaches the Earth, can be used for s.
If a is the angle between the initial telescope pointing vector (straight down) and
the final vector, then
s = r cosa. s (2.41)
Finally, we would like to outline a simple calculation that is frequently useful
in satellite meteorology: how to find the distance of a scan spot from the subsatellite
If the satellite is thought of as an airplane, a positive pitch angle is defined here as the nose
1 6
pointing up, a positive roll as the right wing pointing up, and a positive yaw as a counterclockwise
rotation of the plane as viewed from above.
2.6 Space-Time Sampling 41
FIGURE 2.17. Determining the distance of a scan spot from the subsatellite point.
point. Figure 2.17 shows the geometry of this calculation. If a is the scan angle,
then the law of sines gives the angle β as
(2.42)
The swath width is the width of the entire scan of the satellite instrument. If
the instrument scans equally on each side of the ground track, then the swath
width is 2 0 r , where a is the maximum scan angle.
e
17
90Ni
9 0 Si
90E 180
I 2 or more
FIGURE 2.18. One day's coverage by a hypothetical instrument on a sunsynchronous satellite. The
coverage at the equator is 5 0 % . The orbit is circular with a semimajor axis of 7228 km and an
inclination angle of 98.8°. N o t e that slightly different orbit parameters can result in quite different pat-
terns.
point in this area can be observed as frequently as their instruments will allow;
that is, it can be observed at any local time. However, since each point has a
fixed geometric relationship to the satellite, it is viewed at only one zenith and
one azimuth angle.
For a satellite in low Earth orbit, these questions depend on the satellite's
orbit and the scanning geometry of its instruments. Most meteorological satellite
instruments are designed such that the area viewed on one orbit touches or overlaps
the area viewed on previous and successive orbits. If the satellite's inclination angle
is large enough, the instrument views every point on Earth twice per day, at least.
The poles are observed on every orbit. Usually each point is viewed at a wide
range of zenith and azimuth angles. Many meteorological satellites are in sunsyn-
chronous orbits, which have constant equator crossing times. These satellites view
each point (except near the poles) only in a small range of local times [±iT)
centered on the two equator crossing times.
For instruments whose scans on successive orbits do not overlap, it is often
best to plot the coverage for a day and to determine visually which areas are
observed and which are not. Figure 2.18, for example, shows the one-day coverage
of a hypothetical instrument in a sunsynchronous orbit which has 50% coverage
at the equator. Some areas are not observed, some are observed once, and some
18
are observed twice or more. This pattern will be the same on succeeding days,
except that it will drift in longitude. The drift rate can be calculated as follows.
Divide the length of day by the nodal period and round to the nearest integer,
1 8
That is, the swath width divided by the sine of the inclination angle is 5 0 % of ALON.
2.6 Space-Time Sampling 43
nodal period of 103.27 min, which means that ALON was 25.82°. The distance
between ascending nodes at the equator was about 2874 km. The Multispectral
Scanner (MSS) scanned across the satellite track with a ground swath width of
only 185 km; only a small fraction of the equator was observed on any one day.
The daily longitude drift was -1.43° (-ALON/18), or about 160 km west of an
ascending node on the previous day. Since the swath width was greater than the
westward movement, the swaths on successive days overlapped. In 18 days the
satellites observed every point on the equator and began the cycle anew.
The French SPOT satellites (see SPOT User's Handbook) utilize the same type
of repeat cycle, except that they orbit 14 /i6 times per day. The swaths on
5
successive days do not overlap, but in 26 days the entire Earth is imaged. Landsats
4 and 5 have similar orbits.
Note that this repeat cycle is very sensitive to the semimajor axis. If the orbital
altitude of Landsats 1, 2, and 3 had been decreased by only 19 km, they would
have made exactly 14 orbits in one day. There would have been no westward
progression of the swaths. Some parts of the Earth would be observed every day;
the rest would never be observed.
For studies of diurnal variation, a point must be observed at local times through-
out the day. Since sunsynchronous satellites view a point at nearly the same two
local times every day, they are not useful for diurnal variation studies. A satellite
designed specifically to measure diurnal variation is the Earth Radiation Budget
Satellite (ERBS; see Chapter 10). It is in a 57°-inclination orbit at an altitude of
600 km. The right ascension of ascending node moves west by 3.955° per day,
while the mean sun moves east 0.986° per day (360° in one year). Thus the angle
between the sun and the ascending node changes 4.94° per day. Because the
satellite makes observations both as it ascends and as it descends, all local times
will be sampled when the sun-Earth-ascending node angle has changed by 180°.
The ERBS, then, samples all local times in about 36 days.
Many space-time sampling strategies are possible. The reader is encouraged
to use the equations presented above to investigate some of the possibilities.
44 Chapter 2 Orbits and Navigation
ary altitude.
"A IK
39.7 m
35.4 m
u
Delta 3914 Atlas/
Centaur
FIGURE2.19. Rockets used to launch recent U.S. meteorological satellites. [After Chen (1985). Reprinted
by permission of Academic Press.]
2.7 Launch Vehicles and Profiles 45
Bibliography
Barnes, J. C , and M. D . Smallwood (1982). TIROS-N Series Direct Readout Services Users Guide.
N O A A , Washington, D C .
Brouwer, D., and G. M. Clemence (1961). Methods of Celestial Mechanics. Academic Press, N e w York.
Chen, H. S. (1985). Space Remote Sensing Systems: An Introduction. Academic Press, Orlando.
46 Chapter 2 Orbits and Navigation
47
48 Chapter 3 Radiative Transfer
The fundamental unit of frequency is the hertz (Hz), or one cycle per second.
The frequency ν is related to the wavelength λ by
v = i> (3-D
where c is the speed with which electromagnetic radiation travels and is known
as the speed of light. In a vacuum the speed of light is 2.99792458 x 10 m s" . 8 1
In the atmosphere, it travels slightly more slowly, due to interaction with air
molecules.
The index of refraction, n, of a substance is the ratio of the speed of light in
vacuum to the speed with which electromagnetic radiation travels in that sub-
stance. At sea level, the index of refraction of air is approximately 1.0003. For
I0 9
-IO 5
io 8 X-RAYS -ID 4
SOFT
I0 7
-ID 3
IO 6
0.01
ULTRAVIOLET FAR
I0 5
0.1
—flFNt
io - 4 VISIBLE
__NEAR Μ Lü
I 0 - INFRARED
3
MIDDLE 10
I
ΙΟΟ FAR 100
10 1Λ 3
I MICROWAVES
an 10°
0.01 10 s
RADIO WAVES ΙΟ 7
10 Ho 8
Frequency V Hz
Wavelength λ m
Wavenumber κ m- 1
Radiant energy Q J
Radiant exposure Η Jm" 2
Radiant flux Φ W
Radiant flux density Μ, Ε Wm" 2
Irradiance Ε Wm" 2
Radiance L W m" s r 2 1
Emittance ε Unitless
Absorptance α Unitless
Reflectance Ρ Unitless
Transmittance τ Unitless
Absorption coefficient ν* nT 1
Scattering coefficient σ 8
m" 1
Extinction coefficient m- 1
Single-scatter albedo
Z e
ω Unitless
Absorption number α Unitless
Vertical optical depth δ Unitless
Slant-path optical depth Unitless
Scattering angle φ. rad
Scattering phase function sr" 1
a
We use primarily the radiation symbols recommended by the Radiation Commission
of the International Association of Meteorology and Atmospheric Sciences (IAMAS)
as described in Raschke (1978).
most purposes, therefore, the speed of light in a vacuum can be safely used even
in the atmosphere. However, strong vertical gradients of atmospheric density and
humidity result in strong vertical gradients of η (see Section 3.5.3). These cause
bending of electromagnetic rays and can cause slight mislocation of satellite
scan spots.
One also sees radiation specified by wavenumber, κ, which is the reciprocal
of the wavelength. Traditionally, wavenumber is expressed in inverse centimeters.
Radiation with a 15-μ,ηι wavelength has a 667-cm wavenumber, for example.
-1
1
Some units, such as the lumen and the candela, are based on h o w bright an object appears to
the human eye. These units are no longer used in satellite meteorology.
50 Chapter 3 Radiative Transfer
energy is the joule. Radiant flux is radiant energy per unit time, measured in watts
[W; joules per second (J s" )]. Radiant flux depends on area, which is often
1
to indicate which way the energy is traveling. Radiant exitance (M) is radiant
flux density emerging from an area, and irradiance (E) is radiant flux density
incident on an area.
In nature, radiation is a function of direction. The directional dependence is
taken into account by employing the solid angle. If one draws lines from the
center of the unit sphere to every point on the surface of an object, the area of
the projection on the unit sphere is the solid angle (Fig. 3.3). The solid angle of
an object that completely surrounds a point is 4π steradians (sr), the area of the
unit sphere. The solid angle subtended by an infinite plane is 2π sr. For an object
with cross-sectional area A at a distance r from a point (A « r ), the solid
c c
2
represents the zenith angle (the angle measured from the vertical or from the
normal to a surface), and φ represents the azimuth angle (Fig. 3.4), then a differen-
tial element of solid angle is mathematically given by
dCl - s'mQdQdfy = -άμάφ, (3.2)
where μ = cosö.
Radiant flux density per unit solid angle is known as radiance and is preferably
assigned the symbol L. Suppose that a small element of surface is emitting radiation
with radiance L. A question that arises is: What is the radiant exitance, that is,
what is the total amount of radiation leaving the surface? This question is answered
by integrating the radiance over the 2π sr above the surface. However, radiance
represents the radiation leaving (or incident on) an area perpendicular to the
beam. For other directions, we must weight the radiance by cos0. Therefore, the
Object
Projection of object
χ
FIGURE 3.3. Illustration of a solid angle.
3.1 Basic Quantities
51
FIGURE 3.4. Mathematical representation of a solid angle. [After Raschke (1978), with permission of
the International Association of Meteorology and Atmospheric Sciences.]
radiant exitance is
= Ζ,(μ,,φ) μ άμ άφ
h Jo
The above energy-based quantities may all be prefixed with the word mono-
chromatic or spectral to indicate the wavelength dependence of the radiation.
A subscript (λ, ν, or κ) is used to indicate whether wavelength, frequency, or
wavenumber is being considered. Because radiance is simply the integral over all
wavelengths (frequencies, wavenumbers) of monochromatic radiance, we must
have
L=fL dk=i
k L dv=\
v L dK
K (3.5)
Jo Jo Jo
or
(3.6)
52 Chapter 3 Radiative Transfer
In other words, radiation per unit wavelength interval is the same as radiation
per unit frequency or wavenumber interval, if we take into account the different
size of the intervals.
The most fundamental radiation unit for satellite meteorology is mono-
chromatic radiance, which is the energy per unit time per unit wavelength
(frequency, wavenumber) per unit solid angle crossing a unit area perpendicular
to the beam. The reason radiance is most fundamental is that the basic satel-
lite instrument, the radiometer (see Chapter 4), has a detector of a certain area
whose output is proportional to the energy per unit time striking it. Further, the
sensor is usually at the focus of a telescope which collects radiation from a certain
solid angle and which has filters that pass radiation of only a certain narrow
range of wavelengths. Normalizing the sensor output by area, solid angle, and
wavelength range results in a quantity that is most closely related to monochro-
matic radiance.
Radiance also has the useful property that it is independent of distance from
an object as long as the viewing angle and the amount of intervening matter are
not changed. Consider a satellite viewing a small object. The irradiance reaching
the satellite from the object will decrease inversely as the square of the distance
of the satellite. However, the solid angle of the object subtended at the satellite
will also decrease inversely as the square of the distance of the satellite. The
radiance of the object as viewed by the satellite, which is simply the irradiance
divided by the solid angle, is, therefore, independent of distance. Of course, if
the object were at the Earth's surface, its radiance measured at the Earth's surface
would be different from that measured at the satellite, due to the intervening atmo-
sphere.
the revolutionary assumption that an oscillating atom in the wall of a cavity can
exchange energy with the radiation field inside a cavity only in discrete bundles
called quanta given by Δ Ε = hv, where h is known as Planck's constant (see
Appendix E). With this assumption, he showed that the radiance being emitted
by a blackbody is given by
where k is Boltzmann's constant (see Appendix E), and Tis the absolute tempera-
3
ture. Equation 3.7 is known as the Planck function; it earned him the Nobel
Prize in 1918. The Planck function is more conveniently written as
B , m = . (3.8)
«ρ(*)-1
where c and c are the first and second radiation constants (see Appendix E).
1 2
Figure 3.5 shows Β (Τ) plotted versus wavelength and temperature. It is import-
λ
wavelength that may be determined by setting the partial derivative of B (T) with k
k T=
m 2897.9 /xmK, (3.9)
where X is the wavelength (expressed in micrometers) of maximum emission for
m
M BB = ΓττΒ (Τ)άλ
λ = £c 2 T
LC
4 4
= σΤ\ (3.10)
2
M a x Karl Ernst Ludwig Planck, German physicist, 1 8 5 8 - 1 9 4 7 .
3
Ludwig Eduard Boltzmann, Austrian physicist, 1 8 4 4 - 1 9 0 6 .
4
Wilhelm Wein, German physicist, 1 8 6 4 - 1 9 2 8 .
54 Chapter 3 Radiative Transfer
Ό.Ι "IfJ ίο 2
ίδ*~
WAVELENGTH (/xm)
FIGURE 3.5. The Planck function, (a) Planck radiance versus wavelength for the indicated temperature,
(b) temperature versus wavelength for the indicated Planck radiances (W m " s r " V m ) , (c) Planck
2 -1
radiance versus temperature for the indicated wavelengths. The dashed line is Wein's displacement
law, which gives the wavelength of maximum emission as a function of temperature.
3.2 Blackbody Radiation
55
where σ is called the Stefan-Boltzmann constant (see Appendix E), and Eq. 3.10
is called the Stefan-Boltzmann law. 5
B (T) =^λ~ Τ.
k
4
(3.11)
5
N a m e d after Josef Stefan, Austrian physicist, 1 8 3 5 - 1 8 9 3 , w h o discovered it by observing the
cooling rates of hot bodies; and Boltzmann, w h o demonstrated it using thermodynamics.
6
After John William Strutt, 3rd Baron Rayleigh, English physicist, 1 8 4 2 - 1 9 1 9 ; and Sir James
H o p w o o d Jeans, English mathematician and astronomer, 1 8 7 7 - 1 9 4 6 .
5£ Chapter 3 Radiative Transfer
3.2.2 Nonblackbodies
Since real material is not perfectly black, a way must be devised to quantify
how closely it approximates a blackbody. The emittance of a body is defined as
emitted radiation at λ ...
ε
^ ßjT) ( 3
· 1 2 )
reflected radiation at λ ^
ρ = reflectance = -—— τ.—: —-> 3.13b)
a x
incident radiation at λ
λ Λ
transmitted radiation at λ ν
τ = transmittance = ——— r.—: — (3.13 c)
incident radiation at λ
λ t
Because these three processes are the only possibilities for the incident radiation, 7
by energy conservation, each quantity must be between zero and one, and
«Α + Ρχ + τ = 1 .
λ (3.14)
Kirchhoff discovered that a body is exactly as good an absorber as it is an
8
α = ε.
λ λ (3.15)
This law applies only to material that is in local thermodynamic equilibrium,
which means that it can be characterized by a single thermodynamic temperature.
This is a good assumption below about 100 km in the Earth's atmosphere. Above
100 km, collisions between molecules are rare enough that different chemical
species can have different thermodynamic temperatures. For most satellite meteo-
rology applications, however, the Earth's atmosphere can be considered to be in
local thermodynamic equilibrium.
Since the emittance of a blackbody is by definition one, its absorptance must
also be one; that is, a blackbody, in addition to being a perfect emitter, is also a
perfect absorber. It therefore appears black, thus the name blackbody.
The emittance of real materials is enormously variable. Shown in Fig. 3.6 is
the emittance of two materials used in the Suomi radiometer discussed in Chapter
1. The black paint is supposed to approximate a blackbody by absorbing all
radiation incident on it. Anodized aluminum looks like flat white paint. It is
supposed to reflect solar radiation and absorb infrared radiation emitted by the
We will not consider Raman scattering or fluorescence, in which radiation is absorbed at one
7
9
Assuming the contrary can be shown to violate the second law of thermodynamics.
3.3 The Radiative Transfer Equation 57
WAVELENGTH (yum)
FIGURE 3.6. Emittance as a function of wavelength for two materials used in a satellite radiometer.
Normalized Planck curves representing solar radiation (5800 K) and terrestrial radiation (260 K) are
shown. (Do not be deceived by these normalized curves. The sun's radiance is larger than the Earth's
at every wavelength.) [Adapted from Smith, W. L. (1985). Satellites. In D . D . Houghton (ed.),
Handbook of Applied Meteorology. Copyright © 1 9 8 5 , John Wiley &c Sons, Inc. Reprinted by
permission of John Wiley & Sons, Inc.]
Earth and its atmosphere (terrestrial radiation). The difference between measure-
ments made with instruments coated with these materials is related to the amount
of incident solar radiation.
^f* = A + B + C + D. (3.16)
58 Chapter 3 Radiative Transfer
FIGURE 3.7. Differential volume element containing material which alters a beam of radiation passing
through it.
Terms A and C remove radiation from the beam and are known as depletion
terms. Terms Β and D add radiation to the beam and are known as source terms.
Beer's law states that the rate of decrease in the intensity of radiation as it
10
coefficient and is equal to p)3 (X), where ρ is the density of absorber and β (λ) is
a α
the mass absorption coefficient. If Beer's law is integrated over a finite depth
11
L = L exp(-|\(X)&),
x 0 (3.17)
where s is distance along the path and L is the radiance incident on the absorber. Q
Scattering of radiation out of the beam follows Beer's law. Term C takes the
form —a (k)L , where σ (λ) is the volume scattering coefficient.
s k 5
Finally, Term D describes the amount of radiation from other directions that
is scattered into the beam. This term is complicated by the fact that all directions
must be considered. If we are concerned with radiation traveling in a direction
specified by the angles θ and φ, then Term D is given by
After Wilhelm Beer, German astronomer, 1 7 7 9 - 1 8 5 0 . It is also known as Bouguer's law, after
1 0
units of m k g .
2 - 1
3.3 The Radiative Transfer Equation 59
where (θ',φ') represents the direction of incoming radiation, and i// is the scattering s
radiation from direction {θ',φ') is scattered into direction (0,φ). ρ(ψ ) has the δ
property that
(For an isotropic scatterer, ρ(ψ ) = 1.) Term D can be thought of as the product
8
of cr (X) and a directionally weighted average [ρ(ψ ) is the weight] of the radiance
s 8
(L' ) - j -
K Γ Ι (β',φ') ρ(ψ ) sinö' Jö'
λ 5 (3.23)
Combining all terms, the radiative transfer equation for nonpolarized radia-
tion becomes
12
absorptionally, and the beam passes through it unchanged. If σ (λ) is not zero, Ά
the temperature of the layer and the radiance itself determine the change in the
beam. The beam is augmented if B (T) is greater than Ζ. (0,φ); it is diminished
k λ
if B (T) is less than Ζ, (0,φ). The second term on the right represents the effects
k Λ
1 2
The radiative transfer equation for polarized radiation is not much more complicated. Polarized
radiation can be described by a vector whose elements are the four Stokes parameters. The scattering
phase function becomes a 4 x 4 matrix. The reader is referred to Liou (1980) for further details.
60 Chapter 3 Radiative Transfer
of scattering. If scattering particles are absent, σ (λ) is zero, and scattering has
5
no effect on the beam. If σ (λ) is nonzero, the beam is augmented if the directionally
8
weighted average radiance (L' ) is greater than the beam radiance L (0,</>); it is
x x
The radiative transfer equation can be formally solved. Since we will not use
the formal solution, we refer readers to Goody and Yung (1989) for further
discussion of it. Instead, we will concentrate on several simplifications that are
used in satellite meteorology.
The first two terms on the right-hand side of Eq. 3.24 can be combined to
form -σ (λ)Ι, (0,φ), where
6 λ
(3.27)
(3.28)
Since the Earth's atmosphere is thin in comparison with the radius of the Earth,
the two optical depths are related by
Ssi(*i,s ) = 2
(3.29)
fr
dz
FIGURE 3.8. Relationship of depth to slant path through an infinitesimal atmospheric slab.
3.3 The Radiative Transfer Equation β]
to s . We will use the symbol δ without arguments to mean the vertical optical
2 λ
depth between the surface and level z, that is, δ = δ (0,ζ), and therefore, λ λ
άδ = a dz = μσ^άε.
λ e
δ χ - ^ » (3.30)
σ,(λ)
and the single-scatter albedo, ώ , is defined as
λ
σ,(λ)
ώ χ - ^ · (3-31)
σ,(λ)
With these definitions, the radiative transfer equation becomes
Note that we have substituted μ for cos0 in this equation, as is commonly done.
Now we are ready to simplify and solve the radiative transfer equation. We
will explore two common simplifications: the no-scattering case and the no-
emission case.
μ ^ = - 1 ^ ) + 8»(Τ), (3.33)
The infrared radiance observed by a satellite (in the absence of scattering) can be
calculated by formally integrating Schwarzchild's equation. Since Schwarzchild's
equation is a first-order, linear, ordinary differential equation, its integration is
straightforward. A detailed derivation is justified, however, because the result is
arguably the most important equation in satellite meteorology. To start, we re-
arrange Eq. 3 . 3 3 and multiply by the integrating factor εχρ(δ /μ)/μ: λ
fe) t S + e x p
fe) * '
LM ]
i e x p
fe) ΒΛη (3J4)
1 3
After Karl Schwarzchild, German astronomer, 1 8 7 3 - 1 9 1 6 .
Chapter 3 Radiative Transfer
62
ι λ
8 +d8
x
/ i
θ / J
Using δ as the vertical coordinate (Fig. 3.9), Eq. 3.35 can be integrated from the
λ
Because the left-hand side of Eq. 3.36 is the integral of an exact differential, the
integral is simply
L, = L e x ( - ^ )
0 P + j\xp(-iVlöjW ( T ) A. (3>38)
\ μI J
o \ μ I μ
This equation forms the basis for sounding the atmosphere and for corrections
necessary for surface parameter estimation (Chapter 6).
The vertical transmittance (see Section 3.2.2) of the atmospheric layer between
optical depths h and δ isx 2
Τλ(δι,δ ) = βχρ(-|δ -δ |;
2 2 1 (3.39)
We will use the symbol τ without arguments to mean the vertical transmittance
λ
of the layer between level δ and the satellite (δ ), that is, τ = τ (δ ,δ ). The
λ 0
15
λ λ λ 0
Meteorological satellites orbit well above the effective top of the atmosphere; therefore, δ is
1 4
0
transmittance is that between the surface and level δ . Since the transmittance between δ and δ is
λ λ 0
transmittance between the surface and the satellite is τ = τ (0,δ ). With the aid 0 λ ο
along the slant path to the satellite. The product is that portion of the surface
radiance which reaches the satellite. The second term is the contribution of the
atmosphere. Since we are not considering scattering,
^ =σ β , (3.40)
μ μ
that is, άδ /μ is the emittance of the atmospheric layer between δ and δ + d8 .
κ λ λ k
that portion of the radiance emitted by the atmospheric layer between 8 and k
δ + d8 which reaches the satellite. The integral indicates that the contributions
λ k
L = Ly >»+fB (T)T^-V^-
x 0 K (3.41)
J
r 0 μ
L = L r + \ B {T)dr .
x 0 0
l
k k (3.42)
•'τ
μ^ξ= -L GM>)
X \ λ
Ι (μ\φ')ρ(^)άμ·
λ αφ . 1
(3.43)
Neither liquid water nor water vapor absorbs much radiation in the visible
portion of the spectrum. At visible wavelengths, it is often acceptable to assume
that absorption in clouds is zero, or equivalently that ώ = 1. This approximation λ
L direct = L s^un £ χ ρ / _ δ
ο ~ δ
λ J, (344)
V Msun /
where L{ is the sun's radiance at the top of the atmosphere, and μ is the
un
81ιη
cosine of the solar zenith angle. The diffuse radiation is described by Eq. 3.43,
but one correction must be made. When the second term of Eq. 3.43 is calculated,
both direct and diffuse radiation must be considered. Since the sun's radiation
comes from essentially only one direction (0 ,φ ), the direct-beam term can 81ιη 81ιη
where Ω is the solid angle subtended by the sun, and ψ is the scattering angle
81ιη 8υη
between the sun and the observed direction (θ,φ). The product L^ n is the un
sun
WAVELENGTH (yLun)
FIGURE 3.10. Absorption cross sections of atmospheric gases in the visible and ultraviolet portions of
the spectrum. N o t e that these curves are intended to indicate the relative significance of various
absorbers and should not be taken as the source of quantitative data. [After Liou (1980). Reprinted
by permission of Academic Press.]
Chapter 3 Radiative Transfer
66
0.1750 μηι, 0 dissociates in the much stronger Schuman-Runge continuum.
2
Virtually all radiation between 0.2 and 0.3 μηι is absorbed by dissociation of
ozone ( 0 ) in the Hartley bands. Ozone also dissociates in the weak Chappuis
3
column. Figure 3.11 shows the vertical transmittance of the Earth's atmosphere
in the visible, ultraviolet, and near-infrared portions of the spectrum.
In vibrational transitions, a molecule changes vibrational energy states. These
transitions occur mostly in the infrared portion of the spectrum and are extremely
important for satellite meteorology. They are discussed in the next section. At
temperatures found in the Earth's atmosphere, most molecules are in the ground
vibrational state. The spectrum of vibrational transitions, therefore, is caused
primarily by transitions between the ground state and the first vibrational ex-
cited state.
In rotational transitions, a molecule changes rotational energy states. These
occur in the far-infrared and microwave portion of the spectrum. However, rota-
tional transitions can occur at the same time as a vibrational transition, which
complicates the spectrum. Figure 3.12 shows the infrared spectrum of the radia-
tively most important atmospheric gases. The far-infrared spectrum is dominated
by rotational transitions of water vapor. Measurement of water vapor in the
microwave region is an important use of rotational transitions.
Finally, forbidden transitions are those transitions which are not caused by
the interaction between the electric field of the radiation and the electric dipole
moment of a molecule. Some forbidden transitions do, in fact, occur. A meteoro-
logically important forbidden transition is caused by the reorientation of unpaired
electron spins in the O molecule. This results in an absorption band in the 5-mm
z
region which is used for temperature sounding. Figure 3.13 shows the microwave
spectrum of the Earth's atmosphere.
Figure 3.14 shows the complete spectrum of the Earth's atmosphere under
very low spectral resolution. Although much of the spectrum is opaque because
of absorption by atmospheric gases, there are several important areas, called
windows, where the atmosphere is relatively (but not absolutely) transparent.
The most important of these are the visible window, the 3.7-μηι window, the
microwave windows (2-4 mm and > 6 mm), and the 8.5-12.5 μπι window. This
CO
1 ,1 1 1
Y
CH
Τ
4
^"V Y
IT -
* 0
Λ/....
S
ir COe
Tli
\1
HDO
JT \!
Λ ^ ^ Μ ^ EARTH'S
} / r»V ATMOSPHERE
7 8 9 10 II 12 13 14 (15m)
5 6
9 % A WAVENUMBER
WAVELENGTH
FIGURE 3.12. Infrared transmittance of several gases in the Earth's atmosphere and the combined
atmospheric transmittance. [After Valley (1965).]
£g Chapter 3 Radiative Transfer
FREQUENCY (GHz)
300 200 100 70 50 30 20 10
1
1 1I ι τ • f X H
2 ° / ^
o 2
-
1r
0 if \
/
-
I5°N Annuol Atmosphere
H 0/
Λ
2
/ , , , , ι
WAVELENGTH ( c m )
FIGURE 3.13. Transmittance of the Earth's atmosphere in the microwave portion of the spectrum.
[Calculated using the model of Liebe and Gimmestad (1978).]
example, has neither a static nor a dynamic electric dipole moment because the
symmetry of the molecule is maintained. If a molecule has no electric dipole
moment, the electric field of incident radiation cannot interact with the molecule.
The infrared spectrum of C 0 (Fig. 3.12) shows that there is no absorption at
2
7.46 μ,ιη. For the same reason, C 0 cannot have a pure rotation spectrum. The 2
lack of an electric dipole moment also explains why the two most abundant gases
in the Earth's atmosphere, N and 0 , are transparent in the infrared.
2 2
en, cm
o _»| frH^fycgJ o μ' mostly opoque due to H 0_»| o q
H
f°
Ö
ζ o.e[
l
,0 3 c?t s 2 ; z
Η 0.6
£ 02
I 0 4
2
ι- 0.1/xm 0.3 0.6 1.0/xm 3 6 \0μτη 30 60 100/xm 300 600 0.1cm 0.3 0.6 1.0cm 3
WAVELENGTH
FIGURE 3.14. Spectrum of the Earth's atmosphere. [Adapted from Goody and Yung (1989), with
permission from Oxford University Press.]
3.4 Gaseous Absorption (ft
t
*-0 — C - 0 - > SYMMETRIC /O
STRETCHING H '
• X
K
> = ' 3 4 0 001-' * , = 3 6 5 7 . 0 5 cm'
λ, = 7 4 6 ^ X,= 2 . 7 3 m M
t •
o - c — 0 0
I I / \
BENDING
* = 6 6 7 . 4 0 cm" 1
^ ^ -i
κ2 = 1594.78 cm
2 Ω
* 2 = , 4
- 9 8
^ m
λ = 6.27/xm
2
0 —
ASYMMETRIC / \
—(J — C ^ O STRETCHING ^ ^
* 3 =2349.16 cm" 1
' „ -I c c fto
* 3 = 3 7 5 5 . 9 2 cm
λ 3 =4.26^m χ 2.66 m3 s M
FIGURE 3.16. The Lorentz lineshape for near-surface and near-tropopause conditions; κ is the central 0
A (T,P) =
K AK {pj(^J\
0 (3.47)
strength S is a function of temperature and the energy of the lower level (£") of
the transition. Four numbers, then, specify each line: κ , S , Δκ , and E" (McClat- 0 0 0
chey et 1973). If two or more absorption lines overlap, which is usually the
case, the volume absorption coefficients are summed. Calculation of absorption
coefficients by this method (the line-by-line method) would not be too difficult
except that there are more than 100,000 absorption lines between 1 and 25 μπι!
How can there be so many lines with only a few absorbing gases? Close
examination of the 15-μιη band of C 0 gives the answer. Figure 3.17 shows 2
600-740 cm" (16.67-13.51 μπι). The bending mode vibrational transition oc-
1
seen. The less intense Ρ and R branches of these isotopes are masked by the
C 0 spectrum. Other, less abundant isotopes are also present.
1 2
2
6
1.0
2
P BRANCH
BRANCH
660 680
WAVENUMBER (cm ) 1
FIGURE 3.17. Vertical transmittance of the atmosphere above 4 0 km. The P, Q, and JR branches of the
bending vibrational mode of C 0 can be seen along with absorption due to isotopes of C 0 (see
2 2
transition of the symmetric stretching mode from the ground state to the first
excited state simultaneous with the transition of the bending mode from the first
excited state to the ground state. An overtone band is one in which a molecule
absorbs (or emits) a photon of sufficient energy to cause it to jump two or more
levels. Water vapor, for example, absorbs in a band centered at 3151 cm" , 1
which is the result of double excitation of its bending mode. When all of these
complications are taken into account, 100,000 is not an unreasonable number
of absorption lines, even though only a handful of atmospheric gases contribute
to the infrared spectrum.
Comparing Fig. 3.17 with Fig. 3.12, one might ask: Why are the rotational
transitions not evident in Fig. 3.12? The answer has two parts. First, Fig. 3.12
shows the transmittance of the entire atmosphere. Near the surface, the individual
rotation lines are considerably broadened (Fig. 3.16) so that they overlap each
other. Second, Fig. 3.12 is what is called a low-resolution spectrum. The spectrum
is averaged over a passband and therefore appears smooth. The result of these
two effects is that in Fig. 3.12, only the envelopes of the P, Q, and R branches
of absorption bands are discernable. Note also that not all molecules have a Q
branch. Ozone, for example, has a bimodal absorption band centered at 9.6 μ,ιη,
which consists of the envelope of its Ρ and R branches.
A complete compilation of vibration-rotation lines important for satellite
meteorology has been made by the U.S. Air Force Geophysics Laboratory (McClat-
chey et al., 1973) and an optimized computer codes to calculate atmospheric
transmittances have been written, for example, FASCOD (Clough et al., 1982).
Readers who need to do line-by-line calculations are advised to obtain one of
these codes.
Vibrational transitions are the basis of temperature sounding and measurement
of a wide variety of trace gases in the atmosphere. The 15-μηι and 4.3-μπι C 0 2
bands are used for temperature sounding. The 6.3-μ,ιη H 0 band is used for
2
water vapor measurements. The 9.6-μπι band of ozone is used to make total
ozone measurements. Several other important gases have been measured in the
stratosphere using vibrational transitions. Among them are NO, N 0 , N 0 , 2 2
cm" wide (see Chapter 4), which includes several lines. Calculation of the radiance
1
'measured (3.48)
where f(k) is the response function of the radiometer channel. Even with fast line-
by-line routines such as FASCOD, these calculations are too time consuming to
be of practical use in many applications. Instead, one turns either to band models
or to polynomial expansions.
In a band model the transmittance of an atmospheric layer in an entire band
is calculated at one time using an appropriate parameterization, which is usually
based on fitting line-by-line calculations to a function. Goody and Yung (1989,
Chapter 4) explain several parameterizations. The LOWTRAN computer code
(McClatchey et ai, 1972; Kneizys et al., 1983), in which transmittance averaged
over a 20-cirT band is calculated, is an example of this type of model.
1
In polynomial expansions, the atmosphere is divided into layers. For each layer
and for each channel of the radiometer, transmittances are calculated line-by-line
for a variety of temperatures and humidities. Polynomial functions of temperature
and humidity are then fitted to the transmittances. To calculate the transmittance
of the atmosphere from a particular level to a satellite, one uses an estimate of
atmospheric temperature and humidity and calculates the transmittance of each
layer. The total atmospheric transmittance is the product of the layer transmit-
tances. Polynomial expansions are used extensively in satellite sounding (see Mc-
Millin and Fleming, 1976).
3.5 SCATTERING
1 6
See van de Hülst (1957) for a thorough discussion of scattering by particles.
1 7
Gustav Mie, German physicist, 1 8 6 8 - 1 9 5 7 .
1 8
James Clerk Maxwell, Scottish mathematician and physicist, 1 8 3 1 - 1 8 7 9 .
3.5 Scattering 73
where r is the radius of the sphere. The size parameter can be used to divide
scattering into three regimes (Fig. 3.18).
For size parameters in the range 0.1-50, the wavelength of the radiation and
the circumference of the particle are comparable. Radiation strongly interacts
with the particle, and, therefore, the full Mie equations must be used. These
equations have been applied extensively to the detection of raindrops by radar.
The study of aerosols (smoke, dust, haze) using visible radiation falls in the
Mie regime. Also in the Mie regime is the interaction of cloud droplets with
infrared radiation.
A complete discussion of Mie scattering is outside the scope of this book
because the scattering equations are very complicated. Insight into the results,
however, can be gained as follows. If the volume absorption coefficient is divided
by the number of scattering particles per unit volume and by the cross-sectional
area of each scatterer, the result is the scattering efficiency (Q ) for a single s
to the incident radiation. Q is a function of the size parameter and of the index
s
io 4
RAIN0R0PS
io 3
DRIZZLE
100
CLOUD
DROPS 10
I
SMOKE, 0.1
DUST, HAZE
0.01
io 3
AIR _
£ I I I I L
4
MOLECULES |Q
0.1 I 10 100 IO 3
IO 4
IO 5
IO 6
ζ ζ
Q
t/)2 λ {μπ\)
<<
-JO
o<
s i Ii <
(Ε
FIGURE 3.18. Scattering regimes. [Adapted from Wallace and Hobbs (1977). Reprinted by permissipn
of Academic Press.]
74 Chapter 3 Radiative Transfer
1
' ι
WATER
ICE
.V
0.1 I 10 100 1000
WAVELENGTH (μπ\)
FIGURE 3.19. Imaginary part of the index of refraction of water and ice. [Plotted from data in Irvine
and Pollack (1968).]
«' as a function of wavelength for water and ice. In the visible portion of the
spectrum, n' is negligibly small, but in the infrared it becomes significant.
Figure 3.20 shows the scattering efficiency for water drops (n = 1.33) as a
function of size parameter for several values of n'. Scattering efficiency in the
Mie regime is quite clearly a complicated function. In clouds, there is usually a
distribution of drop sizes. Suppose that N(r)dr is the number of drops per unit
volume in the radius range r to r + dr. If the scatterers are sufficiently far
apart (many wavelengths) that they act independently, the scattering coefficient
is given by
r 00
Τ Τ I 1 Ja! I 1 I
I ι ι ι11 11
-
sζ 3
- ι n'--.0/\l 1/ -
UJ
ο
- / n
' r /
\t //
n= l.33(Water)_
-
1 1 1 1 1 1 1 11 1 1 1 1 1 1 1 1
1 5 10 50 IC
SIZE PARAMETER
FIGURE 3.20. Scattering efficiency (Q ) of water spheres (n = 1.33) as a function of size parameter (χ)
s
for several values of n'. [Adapted from Liou (1980) and Hansen and Travis (1974). Reprinted by
permission of Academic Press, Inc. and Kluwer Academic Publishers.]
3.5 Scattering 75
Integration over the size distribution smooths the scattering efficiency. In any
case, it must be noted that scattering is a much smoother function of wavelength
than is gaseous absorption.
Also of interest is the scattering phase function, which determines in which
direction the radiation is scattered. Figure 3.21 shows the scattering phase function
for water drops for several size parameters. As the size parameter increases, the
phase function becomes strongly peaked in the forward direction; relatively little
radiation is backscattered toward the source of the radiation. Finally, we note
that in general scattering polarizes radiation; in some applications polarization
must be taken into account.
For χ greater than about 50, the sphere is large in comparison with the wave-
length of radiation. This is the realm of geometric optics, where rays, which are
reflected and refracted at the surface of a scatterer, can be traced. Ray tracing
can be used with scatterers that are nonspherical. As shown in Fig. 3.18, the
interaction of solar radiation with virtually all types of hydrometeors falls in this
regime. A wide variety of optical phenomena such as rainbows and halos can
be explained with geometric optics. The interaction of infrared radiation with
precipitation-size particles also falls within this regime.
FIGURE 3.21. Polar plots (note the logarithmic scales) of the scattering phase function of water drops
for several size parameters. [Plotted from data supplied by Steven A. Ackerman, Cooperative Institute
for Meteorological Satellite Studies, University of Wisconsin-Madison.]
Chapter 3 Radiative Transfer
For nonabsorbing spheres, Fig. 3.20 shows that the scattering efficiency asymp-
totically approaches 2 as the size parameter increases. Not only is the radiation
that directly strikes the drop scattered, but an equal amount of radiation that
comes near the drop is refracted around it. Since the refracted radiation changes
direction, it has been scattered.
For χ less than about 0.1, only the first term in the Mie equations must be
considered. The scattering efficiency becomes
2
8 " ' "
X
m (3.52)
m +22
Air molecules act as Rayleigh scatterers for visible and ultraviolet radiation.
The scattering coefficient for air molecules should be simply
σ = nr Q N,
8
2
s (3.53)
where Ν is the number of air molecules per unit volume. However, this formula
can be made more useful and more accurate as follows. First, we multiply by a
factor f = 1.061 to correct for the anisotropic behavior of air molecules. Second,
we note that the index of refraction of air is very close to one, and if only scattering
is to be considered, only the real part of the index of refraction need be used.
Therefore, m + 2 ~ 3, and m - 1 = (η + l)(n - 1) ~ 2(n - 1). Substituting
2 2
Third, the volume of each scatterer is 47rr /3. Since air fills any volume that it
3
occupies, this volume must be equal to the reciprocal of the number of scatterers
per unit volume (N). Therefore,
<Wf^(«-D . 2
(3.55)
N=(|^)p, (3.56)
3.5 Scattering 77
of the gas (28.966 kg kmol" for dry air). The index of refraction of air is given by
1
η — 1= (3.57)
where p and n are the sea-level values. The Rayleigh scattering coefficient for
Q 0
Κ - 1) x 10* = 64.328 + ^ l 1
! + (3.59)
where λ is in micrometers. Equation 3.58 states that for Rayleigh scattering, σ (λ) 8
Figure 3.11 shows the vertical transmittance of the Earth's atmosphere due to
Rayleigh (molecular) scattering. At visible and ultraviolet wavelengths, Rayleigh
scattering by air molecules must be taken into account. Aerosol optical depth can
be measured over the ocean // the observed radiance is corrected for Rayleigh
scattering by air molecules (see Section 8.5). Rayleigh scattering of air molecules
must also be taken into account when making ultraviolet measurements of ozone
(see Section 6.5).
The Rayleigh scattering phase function for unpolarized incident radiation is
given by
P K M W s ) = ^ + ™^X (3.60)
where t// is the angle between the incoming and the scattered radiation (Fig. 3.21).
s
(ψ = 180°).
5
Rayleigh particles, for example, cloud droplets, can also absorb radiation. The
Rayleigh absorption efficiency is given by
Q> = ^(StI)'
4
·
(3 61)
where I indicates the imaginary part. Because Q varies with χ while Q varies
m a s
with χ , absorption can quickly dominate scattering for Rayleigh particles that
4
have even the smallest amount of absorption. The absorption coefficient due to
a number of particles is (similar to Eq. 3.51)
σ , ( λ ) = \\r QMr)dr.
2
(3.62)
total volume of scatterer per unit volume of atmosphere. In a cloud, for example,
absorption due to Rayleigh-size cloud droplets is proportional to the liquid water
content. This is important in the microwave region where cloud droplets absorb,
but do not scatter. Using microwave measurements, the vertically integrated liquid
water content of clouds can be estimated (see Section 8.3).
3.5.4 Clouds
Because clouds play a large role in satellite meteorology, it is important to
understand in a general way the interaction of clouds with radiation. Clouds
consist of water drops or ice crystals with radii on the order of 10 μ,ιη. Drops
with radii of 100 μτη have significant fall speed and constitute drizzle. Drops
with radii of 1000 μ,ιιι (1 mm) are raindrops. Clouds have drop concentrations
on the order of 10 m~ ; that is, the drops are on the order of 1 mm apart.
8 3
In the visible portion of the spectrum (λ — 0.5 μτη), cloud drops are geometric
scatterers; therefore, the scattering efficiency is approximately 2. The scattering
coefficient, then, is —0.1 m" . A photon's mean free path (the average distance
1
Scattered (%)
Wavelength
(μτη) Absorbed (%) Out top Out bottom
a
From Table 2.12a of Welch et al. (1980).
3.6 Surface Reflection 7^
equally well. Clouds therefore appear white. Solar radiation extends into the near
infrared region of the spectrum. In the near infrared, absorption due to water
vapor increases as does the absorption of liquid water. Averaged over the solar
spectrum, Welch et al. (1980) calculate that the cloud absorbs 10.0%, scatters
73.8% out the top of the cloud, and scatters 16.2% out the bottom of the 2-km-
thick cloud.
In the 8.5-12.5-μηι window, cloud droplets are Mie scatterers. The scattering
efficiency is roughly in the range 1-3, but in contrast to visible wavelengths, the
absorption efficiency is on the order of 1. Therefore, clouds absorb nearly all of
the infrared radiation incident on them. They act essentially as blackbodies.
In the microwave portion of the spectrum (χ ~ 0.01), absorption due to cloud
drops is very small. The transmittance of a typical nonraining cloud is greater
than 90%. Scattering is negligibly small. Raindrop-size particles, however, interact
strongly with microwave radiation. Therefore, clouds are nearly transparent in the
microwave region, but raining clouds are not. This forms one basis of microwave
detection of precipitation (see Chapter 9).
Cirrus clouds have a higher transmittance than water clouds because ice clouds
have far fewer particles per unit volume than water clouds and because water is
a better absorber than ice. Cirrus clouds can also be vertically thinner than water
clouds. In general thin cirrus clouds are difficult to detect with satellite radiometers,
yet their effects are not negligible. They can cause problems in the retrieval of
atmospheric soundings (see Chapters 6 and 8). The detection of cirrus clouds
with satellite instruments is an area of active research.
write the formula for the radiance reflected from a small element of surface:
surface. Taking into account the effect of incident angle, L^osd, is available to
be reflected. A fraction γ ( ^ , φ ; 0 ,φ ) is reflected into direction (0 ,φ ). Integrating
Γ ί ί Γ Γ Γ Γ
over all incident solid angles gives the reflected radiance in direction (0 ,φ ). Γ Γ
Α= ^· (3.65c)
Albedo is a unitless ratio between zero and one. As defined in Eqs. 3.65, it is a
function of neither the incoming nor the outgoing angles; however, this does not
mean that the albedo is constant. If the incoming radiation changes, the albedo
will change. This is most easily understood by restricting the incoming radiation
to direct-beam solar radiation, which comes from a very narrow range of angles.
In this case,
£ = L n sun sun cos0 , sun (3.66a)
L (0 ,(/> ) =
r r r L n sun sun c o s 0 y ( 0 , φ ; 0 , φ ),
sun r r Γ sun 51ιη (3.66b)
RLTT RTT/2
Μ= L sun il sun cos0 sun γ (0 , φ ; 0 , φ ) cos0 sin0 de, άφ ,
Γ Γ Γ sun δ1ιη r r τ (3.66c)
where i l is the solid angle of the sun subtended at the Earth. Since the direction
sun
of the sun remains in the equation, the albedo is a function of solar direction.
As an example of reflecting surfaces, two limiting cases are useful. A Lambertian
or isotropic reflector reflects radiation uniformly in all directions. If its albedo is A,
then its bidirectional reflectance is a constant ΑΙ π. Flat white paint approximates a
perfect {A = 1, independent of wavelength) Lambertian reflector. A specular
reflector is like a mirror; its bidirectional reflectance is strongly peaked. Solar
radiation from a perfect specular reflector would be observed only at the zenith
angle equal to the solar zenith angle and at the azimuth angle equal to the solar
azimuth angle plus 180°. Water surfaces are similar to specular reflectors, except
that real water surfaces are always somewhat rough, so the solar reflection is
blurred and larger than the sun. This is called sun glint or sun glitter (see Chap-
ter 5).
Although we have not used the subscript λ with the reflectance quantities
discussed here, they are functions of wavelength. They can be integrated over the
passband of a satellite radiometer or over the solar spectrum. To so integrate
albedo, integration is performed separately for £ and Μ over the passband, and
3.7 Solar Radiation g]
Hobbs (1977).
then the ratio is taken. Readers of the literature should be careful to determine
whether "albedo," in particular, refers to a monochromatic quantity or to one
integrated over some passband. The albedo of various surfaces, integrated over
solar wavelengths, is given in Table 3.3.
A function closely related to albedo and bidirectional reflectance is the aniso-
tropic reflectance factor
ί (β ,φ ;βΐ,Φί)=57Γ(β ,Φ ;βί,Φί).
Γ Γ Γ Γ Γ (3.67)
the surface to that from a Lambertian surface also with albedo A; it is greater
than one where the surface reflects more than a Lambertian surface; it is less than
one where the surface reflects less than a Lambertian surface. Since bidirectional
reflectance is usually applied to solar radiation, it is convenient to use the azimuth
angle of the sun as the reference azimuth. ξ is therefore usually written as τ
and ocean. Note that although isotropic reflection (ξ ~ 1) is not a bad assumption Τ
in some cases, sun glint is evident in the water reflectance. Most surfaces deviate
significantly from Lambertian surfaces at low solar elevation angles (high solar
zenith angles). Finally, note that Eqs. 3.66 and 3.67 require that for any incident
direction (0 ,φ ) 8υη 51ιη
J
'Itt γττ/2
£ (0 , φ ; 0 , φ ) cos0 sin0 άθ, άφ = ττ.
r r Γ sun 5υη r r τ (3.68)
0 J
o
This equation is useful for checking experimentally determined values of ξ . Τ
The solar radiation reaching the Earth originates (for our purposes) from a
layer of the sun called the photosphere, which coincides with the visible disk of
αθ·-25.8· SOLAR ZENITH ANGLE 66,4«-72.5* QO*-25.8* SOLAR ZENITH ANGLE 66,4»-72.5'
0.667 ALBEDO 0-677 0.425 ALBEOO 0.560
0.0·-25.8· SOLAR ZENITH ANGLE 66.4·-72.5· Ο0·-25.8· SOLAR ZENITH ANGLE 66.4 -72.5
e e
0.160 ALBEDO °- 2 0 5
0.076 ALBEDO 0.161
FIGURE 3.22. Anisotropic reflectance factor for snow, middle cloud (overcast), clear land, and clear
ocean from the Nimbus 7 Earth Radiation Budget Experiment. The radial lines represent the viewing
azimuth angle relative to the sun: At the top of each plot (0° relative azimuth), reflection is back
toward the sun; at the bottom of each plot (180° relative azimuth), reflection is away from the sun.
The circles represent the viewing zenith angle: The center (0° zenith angle) is looking straight down
on the scene; the outer circle (90° zenith angle) represents viewing parallel to the surface. Shaded
areas are those in which the scene reflects less radiation (appears darker) than an isotropic (Lambertian)
reflector; unshaded areas reflect more radiation (appear brighter) than an isotropic reflector. [Plotted
from data in Suttles et al. (1988).]
3.7 Solar Radiation
the sun and has a radius of 6.96 x 10 km. The radiation leaving the photosphere 5
is very nearly that of a 6000-K blackbody. However, before reaching Earth, solar
radiation must traverse the solar atmospheric layers of the chromosphere and the
corona. The gases in these layers are both cooler and warmer than 6000 K;
therefore, they both absorb and emit radiation at their characteristic wave-
lengths. After leaving the solar atmosphere, the radiation travels (on average) the
1.49598 x 10 km to Earth. Because the solid angle subtended by the sun at the
8
Earth is so small (6.8 x 10~ sr) solar radiation all comes from essentially the
5
same direction. It is customary, therefore, to use solar irradiance rather than solar
radiance. The irradiance reaching the top of the Earth's atmosphere is the radiant
exitance leaving the top of the sun's atmosphere times the square of the ratio of
the radius of the photosphere to the Earth-sun distance (L il ). The radiation sun sxin
reaching the Earth's surface is further modified by scattering and gaseous absorp-
tion in the atmosphere. Figure 3.23 shows the solar spectral irradiance reaching
the top of the Earth's atmosphere and that reaching the surface. Shown for
comparison is the spectral irradiance which would reach the Earth if the sun were
a 6000-K blackbody.
Solar irradiance reaching the Earth peaks in the visible portion of the spectrum
near 0.48 μηι, whereas infrared radiation emitted by the Earth peaks near 10
μηι. The Earth emits essentially no visible radiation; likewise, the Earth receives
negligible amounts of 10 μηι solar radiation, due to a combination of the sun's
0 3 0 5 0 2 H 0 Η θ'
2 2 H0 '
2 H 02 H2O-CO2 ' H^Ö
FIGURE 3.23. Solar spectral irradiance. The dashed curve shows the approximate irradiance that would
be received at the Earth if the sun were a 6000-K blackbody. The top solid curve shows the spectral
irradiance at the top of the atmosphere. (The integral under this curve is the solar constant.) The
bottom solid curve represents the approximate solar irradiance reaching the Earth's surface after
absorption and scattering in the atmosphere. The shaded area represents absorption by atmospheric
gases, and the difference between the top solid curve and the envelope of the shaded area represents
scattering. [Adapted from Liou (1980). Reprinted by permission of Academic Press.]
J4 Chapter 3 Radiative Transfer
temperature and its very great distance from Earth. As a result of this separation
in wavelength, solar radiation is often called shortwave radiation, and terrestrial
radiation is called longwave radiation. The separation between solar and terrestrial
radiation is not quite complete, however. Suppose that we represent the Earth as
a 250-K blackbody and that we represent the sun as a black, 5774-K sphere
whose radius is that of the sun and whose distance is the mean Earth-sun distance.
Then solar radiation and terrestrial radiation are equal in magnitude at about
5.7 μπι. Solar radiation is one tenth terrestrial radiation at 7.7 μιη, and terrestrial
radiation is one tenth solar radiation at 4.5 μτη. During daylight hours, satellite
data must be carefully interpreted near these wavelengths.
The annual average total irradiance reaching the top of the Earth's atmosphere
is known as the solar constant (S ). Accurate determination of the solar constant
sun
during the year due to the eccentricity of the Earth's orbit about the sun.
Bibliography
Clough, S. F., F. Kneizys, W. Gallery, L. Rothman, L. Abreu, and J. Chetwynd (1982). FASCOD.
U.S. Air Force Geophys. Lab. AFGL-TR-78-0081, Hanscom AFB, MA.
Goody, R. M., and Y. L. Yung (1989). Atmospheric Radiation: Theoretical Basis, 2nd ed. Oxford
Univ. Press, Oxford and N e w York.
Hansen, J. E., and L. D. Travis (1974). Light scattering in planetary atmospheres. Space Sei. Rev.,
16, 5 2 7 - 6 1 0 .
Herzberg, G. (1950). Atomic Spectra and Atomic Structure. Dover, N e w York.
Irvine, W. M., and J. B. Pollack (1968). Infrared optical properties of water and ice spheres. Icarus,
8, 3 2 4 - 3 6 0 .
Kneizys, F., Ε. Shuttle, W. Gallery, J. Chetwynd, L. Abreu, J. Selby, S. Clough, and R. Fenn (1983).
Atmospheric Transmittance I Radiance Computer Code LOWTRAN 6. U.S. Air Force Geophys.
Lab. AFGL-83-0187, Hanscom AFB, MA.
Kondratyev, K. Ya. (1969). Radiation in the Atmosphere. Academic Press, N e w York.
Kyle, T. G., and A. Goldman (1975). Atlas of Computed Infrared Atmospheric Absorption Spectra.
National Center for Atmospheric Research, N C A R - T N / S T R - 1 1 2 , Boulder, CO.
Liebe, Η. J., and G. G. Gimmestad (1978). Calculation of clear air EHF refractivity. Radio Science,
13, 2 4 5 - 2 5 1 .
Liou, K.-N. (1980). An Introduction to Atmospheric Radiation. Academic Press, N e w York.
McCartney, E. J. (1983). Absorption and Emission by Atmospheric Gases. Wiley, N e w York.
McClatchey, R. Α., R. W. Fenn, J. E. A. Selby, F. E. Volz, and J. S. Garing (1972). Optical Properties
Based on measurements from the Nimbus 7 Earth Radiation Budget (ERB) experiment, the
2 0
Active Cavity Radiometer Irradiance Monitor (ACRIM) on board the Solar M a x i m u m Mission
satellite, and the Earth Radiation Budget Experiment (ERBE) on board the Earth Radiation Budget
Satellite, N O A A 9, and N O A A 10. A systematic difference between the instruments of approximately
7 W m ~ has yet to be resolved (see Section 10.1).
2
3.7 Solar Radiation g5
of the Atmosphere (3rd ed.). U.S. Air Force Cambridge Res. Lab. AFCRL-TR-72-0497, Hanscom
AFB, MA.
McClatchey, R. Α., W. S. Benedict, S. A. Clough, D. E. Burch, R. F. Calfee, K. Fox, L. S. Rothman,
and J. S. Garing, (1973). AFCRL Atmospheric Absorption Line Parameters Compilation. U.S.
Air Force Cambridge Res. Lab. AFCRL-TR-73-0096, Hanscom AFB, MA.
McMillin, L. M., and Η. E. Fleming (1976). Atmospheric transmittance of an absorbing gas: A
computationally fast and accurate transmittance model for absorbing gases with constant mixing
ratios in inhomogeneous atmospheres. Appl. Optics, 15, 3 5 8 - 3 6 3 .
Raschke, E. (ed.), (1978): Terminology and Units of Radiation Quantities and Measurements. Interna-
tional Association of Meteorology and Atmospheric Sciences (IAMAS, formerly the International
Association of Meteorology and Atmospheric Physics) Radiation Commission, Boulder, CO.
(Printed by the National Center for Atmospheric Research, Boulder, C O 80307.)
Smith, W. L. (1985). Satellites. In D . D. Houghton (ed.), Handbook of Applied Meteorology. John
Wiley & Sons, Inc., N e w York.
Suttles, J. T., R. N . Green, P. Minnis, G. L. Smith, W. F. Staylor, B. A. Wielicki, I. J. Walker, D . F.
Young, V. R. Taylor, and L. L. Stowe (1988). Angular Radiation Models for Earth-Atmosphere
System, Volume I: Shortwave Radiation. N A S A Ref. Pub. RP 1184, vol. I, Langley Research
Center, Hampton, VA.
Valley, S. L. (ed.) (1965). Handbook of Geophysics and Space Environments. Air Force Cambridge
Research Laboratory, Hanscom AFB, M A .
van de Hülst, Η. C. (1957). Light Scattering by Small Particles. John Wiley &c Sons, N e w York.
Wallace, J. M., and P. V. Hobbs (1977). Atmospheric Science: An Introductory Survey. Academic
Press, N e w York.
Welch, R. M., S. K. Cox, and J. M. Davis (1980). Solar radiation and clouds. Meteorol. Monogr.,
17, 96 pp.
4
Meteorological
Satellite
Instrumentation
because these are the sources of the data that readers will most likely use. Next
we discuss important but not yet operational instrumentation on other satellites.
In the final section of the chapter, we list data archives, the locations where
satellite data are stored.
87
a ARRAY
DRIVE
HIGH-ENERGY
PROTON A N D
ELECTRONICS ALPHA PARTICLE
SOLAR A R R A Y DETECTOR MEDIUM ENERGY
DRIVE MOTOR PROTON A N D ELECTRON
DETECTOR
ADVANCED
REACTION
VERY HIGH
SYSTEM
RESOLUTION
SUPPORT
RADIOMETER
STRUCTURE
BATTERY STRATOSPHERIC
M O D U L E S (4) SOUNDING UNIT
THERMAL CONTROL
' P I N W H E E L L O U V E R S (15)
LEGEND
FIGURE 4.1. The TIROS N-series satellite (a) TIROS Ν ; (b) Advanced TIROS N . [After Schwalb
( 1 9 7 8 , 1982).]
4.1 Operational Polar-Orbiting Satellites 89
Satellite Orbit AVHRR HIRS/2 MSU ssu ERBE SBUV/2 SEM DCS SScR
TIROS Ν PM 1 X X X X X
NOAA 6 AM 1 X X X X X
NOAA 7 PM 2 X X X X X
NOAA 8 AM 1 X X X X X X
NOAA 9 PM 2 X X X X X X X X
NOAA 10 AM 1 X X X X X X X
NOAA 11 PM 2 X X X X X X X
NOAA 12 AM 1 X X X
NOAA 13 PM 2 X X X X X X X
NOAA 14 AM 2 X X X X X
data collected by the instruments to Earth stations, and radio receivers, which
acquire operational commands from the ground. The bus contains the computers
that process the data from the instruments and the tape recorders on which data
are recorded for later transmission to Earth. The bus must supply electrical power
to itself and the instruments, control the environment (particularly temperature)
for the instruments and electronics, maintain proper orientation, and be capable
of adjusting its orbit. These are difficult tasks in the harsh environment of space.
Satellite bus technology is a subject unto itself, and we will not pursue it further.
We concentrate on the instruments which make the measurements.
Table 4.2 lists the instruments carried on the TIROS Ν series satellites. Several
of these instruments are not directly related to meteorology. For example, the
Search and Rescue (S&R ) system detects signals from downed aircraft and other
2
vessels and gives a precise estimate of their location to aid in rescue operations.
The Space Environment Monitor (SEM) measures energetic particles (protons,
electrons, alpha particles) for solar and ionospheric studies. Finally, the Data
Collection System (DCS) relays meteorological and other data transmitted from
ground-based instruments. In the following sections, the meteorological instru-
ments are described in detail.
aperture radar. To avoid confusion, we use S&R for Search and Rescue.
4.1 Operational Polar-Orbiting Satellites
91
TABLE 4.3. Advanced Very High Resolution Radiometer (AVHRR) System Parameters
Telescope
Type Afocal Mersenne Afocal Mersenne
Diameter 2 0 . 3 2 cm 2 0 . 3 2 cm
Field of view 1.3 mrad 1.3 mrad
Ground resolution*
Nadir 1.1 km 1.1 km
End of scan 2.3 x 6.4 km 2.3 x 6.4 km
Scan mirror rotation 1 2 π rad s _ 1 127Γ rad s" 1
1 0 . 5 5 - 0 . 6 8 μπι 0 . 5 8 - 0 . 6 8 μηι
2 0 . 7 5 - 1 . 1 0 μηι 0 . 7 2 5 - 1 . 1 0 μπι
3 3 . 5 5 - 3 . 9 3 μπι 3 . 5 5 - 3 . 9 3 μπι
4 1 0 . 5 - 1 1 . 5 μπι 1 0 . 3 - 1 1 . 3 μπι
5 channel 4 repeated 1 1 . 5 - 1 2 . 5 μπι
Data rate c 7 5 0 kbits s" 1 7 5 0 kbits s" 1
Instrument size 2 7 x 3 7 x 79 cm 2 7 x 3 7 x 79 cm
Instrument mass 3 0 kg 3 0 kg
Power consumption 29 W 29 W
SECONDARY
SCAN PRIMARY
MIRROR MIRROR
MIRROR
TO OPTICS
5.08cm AND
DETECTORS
SPACECRAFT
VELOCITY
VECTOR
U — 2 0 . 3 2 cm *\
FROM EARTH
FIGURE 4.3. Schematic diagram of the A V H R R scan mirror and telescope.
the physical properties of the two versions of the instrument that have been flown
to date. The AVHRR, like its predecessor, the Very High Resolution Radiometer
(VHRR), which flew on the previous generation of NOAA satellites (the ITOS
series), consists of a rotating scan mirror, a telescope, internal optics, detectors,
and electronics.
The scan mirror is elliptical with a major axis of 29.46 cm and a minor axis
of 20.96 cm. It is aligned at a 45° angle to the axis of the telescope, which is
nominally parallel to the satellite velocity vector (Fig. 4.3). As the mirror rotates,
it scans the field of view of the telescope across the Earth perpendicular to the
satellite ground track. The AVHRR scan mirror rotates at a rate of 360 revolu-
3
tions per minute (rpm). Since the subsatellite point of the TIROS Ν series satellites
moves at the rate of about 392 km min" , the distance between successive scan
1
lines at the subsatellite point is about 1.1 km. The distance between successive scan
lines is one parameter that describes the spatial resolution of a satellite instrument.
The afocal Mersenne telescope consists of parabolic primary and secondary
mirrors. The primary mirror is 20.32 cm in diameter, and the beam reflected
from the secondary mirror is 5.08 cm in diameter. Unlike other instruments, the
reflected beam is collimated (not focused).
After passing through the hole in the primary mirror, the beam encounters an
array of beam splitters, mirrors, lenses, and filters (Fig. 4.4) whose job is to deliver
a filtered portion of the beam to each of the detectors. The optics are designed
to minimize stray radiation and polarization effects that would contaminate the
signal. Also, the optics are carefully aligned so that each of the channels views
the same spot on Earth within an angle of 0.1 mrad.
Two versions of the AVHRR have been flown. The AVHRR/1 has 4 channels,
and the AVHRR/2 has 5 channels. Channels 1 (0.6 μιτι) and 2 (1.1 μιη) on both
instruments use silicon detectors that are 2.54 mm on each side. In front of the
Since the spacecraft is moving, however, a plot of the scan spots will make a slight angle with
3
I
M3,M4 FLAT FOLDING MIRRORS
L1 FAR-INFRARED FOCUS LENS
W1.W2 COOLER WINDOWS
D2/F3 INFRARED, DICHROIC &
CHANNEL 3 FILTER
F1.F2.F3 BANDPASS FILTERS
L2, L3, L7 APLANAT LENSES
L4 FOCUS ACHROMAT LENS ASSY.CH.1
L5 FOCUS ACHROMAT LENS ASSY.CH.2
D3 BEAMSPLITTER, INCONEL ...
W3, W4 COOLER WINDOWS M 4
CHANNEL 3 DETECTOR.
CHANNEL 4 DETECTOR
FIGURE 4.4. The A V H R R posttelescope optical system. [Courtesy of N O A A and ITT Aerospace.]
detector is a 0.6-mm-square field stop, which blocks radiation except for that
coming from a rectangular scan spot on Earth. Channel 3 (3.7 μηι) uses an indium
antimonide (InSb) detector that is 0.173 mm square. Channel 4 on the AVHRR/1
(11 μιτι) and channels 4 (11 μηι) and 5 (12 μηι) on the AVHRR/2 use mercury
cadmium telluride (HgCdTe) detectors that are also 0.173 mm square. To lessen 4
the amount of thermal noise, the detectors for channels 3, 4, and 5 are cooled
to 105 Κ by exposing them to space (2.7-K equivalent blackbody temperature)
through the side of the instrument housing. NOAA Κ will have a third version
of the AVHRR (AVHRR/3) on board. AVHRR/3 will add a sixth channel (called
channel 3A) sensitive to radiation between 1.58 and 1.64 μπι. Channel 3A will
operate during daylight, and channel 3 will operate at night.
The properties of the detectors and the optics in front of them determine the
spectral response of each channel. Figure 4.5 shows the relative response functions
for the AVHRR/2 which flew on NOAA 9. The AVHRR/1 response functions
are very similar, but the AVHRR/1 lacks channel 5. These response functions
are slightly different for the instrument on each satellite. They are measured in the
4
For more information o n the properties of various types of detectors, the reader is referred to
Chen ( 1 9 8 5 ) .
Chapter 4 Meteorological Satellite Instrumentation
94
w 0.8-
ρ
ft 0.6-
Q
Μ
0.4-
<
8 0.2-
o.o- I I I I I
1 I III
0.506 0.8 1.0 2 3 4 5 7 9 II 13
WAVELENGTH (/XM)
FIGURE 4.5. A V H R R spectral response functions. [Adapted from Kidwell ( 1 9 8 6 ) /
Twelve points are used for channel 1, but for channel 2, only 6 - 7 points (lamps) are necessary
5
percent of the mean annual spectrally weighted solar irradiance at the top of the
atmosphere. A spectrally averaged radiance can be calculated by multiplying
6 7
this percentage by the mean annual spectrally averaged solar irradiance and
dividing by π. Typically, channels 1 and 2 have a 1 0 : 1 signal-to-noise ratio for
a very dark scene which reflects only 0.5% of the solar irradiance.
Until now, channels 1 and 2 have not been calibrated after launch; the labora-
tory slope and intercept were used throughout the lifetime of the instrument. As
of this writing, however, NESDIS is planning to implement a system whereby
AVHRR observations of Earth targets such as White Sands, New Mexico, of
known or independently measured reflectance are used to update the slope and
intercept every few weeks. This system should alleviate the inevitable postlaunch
degradation of the sensor system and improve radiation budget estimates, for ex-
ample.
AVHRR channels 3, 4, and 5 are calibrated in flight by viewing hot and cold
objects. During each rotation of the mirror, the telescope views both cold space
and the instrument housing, which is painted black and equipped with platinum
resistance thermometers to accurately measure the housing temperature (roughly
290 K). The radiance of the instrument housing (calculated from its measured
temperature) plus the digital counts for the housing and for space (essentially
zero radiance) allow digital counts for Earth scenes to be linearly converted to
radiance (Lauritson et al., 1979). The equivalent blackbody temperature of the
scene can then be determined with the Planck function.
NOAA specifications for AVHRR channels 3 - 5 require the equivalent black-
body temperature of a 300 Κ scene to be determined within ±0.12 K. This is
equivalent to radiance errors of 2.1, 16.9, and 14.6 mW m " s r μιτΓ at 3.7, 2 - 1 1
11, and 12 /xm, respectively. Because the sensors basically measure radiance, the
radiance error for each channel is approximately constant. The temperature error,
however, varies with scene temperature as specified by the Planck function. At
250 K, for example, the temperature errors will be approximately 1.1, 0.20, and
0.18 Κ at 3.7, 11, and 12 /xm, respectively.
The output of each sensor is continuous; a voltage can be measured at any
time. This continuous output is discretized by electronically sampling (measuring)
the voltage at the rate of 40 kHz (once each 25 μ$). Each pixel in the data stream
represents one such sample. Since the scan mirror rotates at the rate of 127Γ rad
s , it turns about 942 /xrad between samples. From a height of 850 km, a scan
_1
spot at the subsatellite point moves about 0.80 km between samples. This distance
between adjacent scan spots is a second parameter that is used to describe the
spatial resolution of a satellite instrument. For each rotation of the scan mirror,
the AVHRR records 2048 samples centered on the subsatellite point. This means
In the documentation (e.g., Kidwell, 1986) this value is often called albedo, but it takes into
6
account neither solar zenith angle, variation of the Earth-sun distance, nor anisotropic reflectance
of the scene; therefore, it is not an albedo as defined in Chapter 3.
Alternately, the albedo of the scan spot can be calculated (see Chapter 10). If the scan spot being
7
observed is assumed to be a Lambertian reflector, its albedo can be calculated by dividing the percentage
by the cosine of the solar zenith angle at the scan spot.
96 Chapter 4 Meteorological Satellite Instrumentation
Channel 1 Channel 5
• r ι ^
l.35mrad A
2 3
ANGLE (mrod)
FIGURE 4.6. A V H R R angular response functions. [Courtesy of N O A A and I T T Aerospace.]
that it scans ±55.3° from nadir. From a height of 850-km, it scans about 1500
km from the subsatellite track. All five channels are measured simultaneously.
The final parameter that is used to describe the resolution of a satellite instru-
ment is the field of view (FOV, also called the instantaneous field of view or
IFOV) of the instrument. In the case of the AVHRR, the active area of the
8
detectors themselves (or the field stops in front of the sensors) determines the
FOV. All channels of the AVHRR have fields of view of 1.3 ± 0.1 mrad. From
850 km altitude, this translates to a scan spot 1.1 km across at the subsatellite
point. At either end of a scan line, each pixel measures about 2.3 km in the
direction of satellite motion (along-track) and about 6.4 km in the direction
perpendicular to the satellite track (across-track).
Just as a radiometric instrument measures radiation covering a range of wave-
lengths, not just from a single wavelength, so an instrument measures radiation
coming from a range of directions, not just from a single direction. The angular
response curve for an instrument is the relative response as a function of angle
from the central angle. Figure 4.6 shows an example of the angular response
curves for two of the AVHRR channels. Before launch, the spectral and angular
responses of all AVHRR instruments are measured. Similar to the case of spectral
response, the field of view of a satellite radiometer is specified by giving the full
width at half maximum of the response curve: 1.40 mrad for channel 1 of the
instrument depicted in Fig. 4.6, 1.35 mrad for channel 5.
The angular response curve for the AVHRR visible channel is steep near the
half-power points; only about 7% of the area under the curve is outside of these
points. The curve for the AVHRR infrared channel is broader; about 16% of the
See, however, the discussion of the effects of motion on scan spot size in Section 4.1.6.2.
4.1 Operational Polar-Orbiting Satellites
97
area under the curve is outside the half-power points. This broadening of the
9
can be received and processed by anyone with the proper equipment. Because the
data rate is rather high, however, the necessary equipment is relatively expensive.
For those interested only in acquiring visible and infrared images, the AVHRR
data are processed on board the satellite into a much lower data rate data stream
that can be received with inexpensive equipment. Two channels of the AVHRR,
usually channels 1 and 4, are reduced in resolution to about 4 km by averaging
along the scan line and by neglecting some scan lines. These data are transmitted
to the ground in an analog (television-like) signal by a separate 5-W transmitter
operating at a frequency of 137.50 or 137.62 MHz. These transmissions are called
automatic picture transmission (APT). Barnes and Smallwood (1982) describe the
APT transmissions.
Since APT and HRPT can be received only when the satellite is within range
of a receiving station, five pairs of digital tape recorders are used on the satellite
to record data for playback to ground stations. NOAA operates two Command
and Data Acquisition (CDA) stations, one at Gilmore Creek, Alaska, and the
other at Wallops Island, Virginia. In addition, the Centre National d'Etudes
Spatiales (CNES) operates a receiving station near Lannion, France. All these data
are transmitted to the Satellite Operations Control Center (SOCC) in Suitland,
Maryland. Each tape recorder can store either one full orbit of 4-km-resolution
data, called global area coverage (GAC) data, or 10 min of full-resolution HRPT
11
has much coarser resolution (42 km) than the AVHRR (1.1 km). These differences
are due to the different requirements of the two instruments. The AVHRR is
designed to make images, in which the horizontal structure of the atmosphere is
most important, whereas the HIRS/2 is used for soundings, in which the vertical
structure of the atmosphere is most important. Table 4.4 contains the HIRS/2
system parameters, and Table 4.5 summarizes the HIRS/2 channels.
The HIRS/2 utilizes two carbon dioxide bands for temperature sounding: seven
channels are located in the 15-μπι band, and six channels are located in the 4.3-
jLtm band. Older sounding instruments [such as the Vertical Temperature Profile
Radiometer (VTPR), which flew on the NOAA satellites prior to the TIROS Ν
series] had only 15-/xm temperature sounding channels. The 4.3-μπι channels
were added to improve sensitivity (change in radiance for a given change in
atmospheric temperature) at relatively warm temperatures. Moisture is sensed
with three channels in the 6.3-μηι band of water vapor. The 9.7-μπι channel is
designed to sense ozone. Three channels are in atmospheric windows: the 11.11-
and 3.76-/zm channels help determine the surface (skin) temperature, whereas
the 0.69-μτη channel is used to detect clouds. A detailed description of the process
by which atmospheric soundings are retrieved from these radiances can be found
in Chapter 6.
TABLE 4.4. High Resolution Infrared Radiation Sounder (HIRS/2) System Parameters
Parameter Value
Telescope
Type Cassegrainian
Diameter 15.24 cm
Field of view 21.8 mrad (1.25°)
Ground field of view
Nadir 18.5 km diameter
End of scan 31.8 km along-track x 62.8 km cross-track
Scan line
Time per scan 6.4 s
Angle from nadir ±49.5°
Distance from nadir ± 1 1 1 5 km
Number of steps 56
Scan step
Time 100 ms
Angle 1.8°
Distance between scan spots
Cross-track 2 6 . 4 km
Along-track 41.8 km
Gap between consecutive passes at equator 5 4 0 km
Number of channels 20
Calibration 2 stable blackbodies and space
Data rate 2 8 8 0 bits s _ 1
Instrument mass 2 0 . 4 kg
Instrument volume 0.040 m 3
a
Values in this table will change slightly with satellite (see Kidwell, 1 9 8 6 ) .
b
Noise equivalent radiance difference (rms error).
c
Noise equivalent temperature difference (rms error) at the mean scene temperature.
The scan mirror in the AVHRR rotates continuously; the HIRS/2 scan mirror,
however, moves in steps. Between each step, the mirror moves 1.8°. The instrument
then stares at a location for approximately 65 ms while all 20 channels are
sampled. During the next 35 ms, the scan mirror steps to the next position. The
total time between scan spots is 100 ms. Fifty-six scan spots between nadir angles
of ±49.5° are sampled during each scan line. The total time for a scan line,
including the time for the mirror to rotate back to the first position, is 6.4 s. The
instantaneous field of view of the radiometer is 1.25°. This scanning geometry
results in the scan pattern shown in Fig. 4.7.
Only three detectors are used on the HIRS/2. The visible channel (20) is
sampled by a silicon detector at ambient temperature. Channels 1—12 are called
the longwave channels and are sampled by a single HgCdTe detector cooled to
105 K. The shortwave channels (13—19) are sampled by a single InSb detector
also cooled to 105 K. Sampling of multiple channels with a single detector is
accomplished with a rotating filter wheel. Figure 4.8 shows a schematic diagram
of the HIRS/2 optics. The angular length of each filter and the rotation rate of
100 Chapter 4 Meteorological Satellite Instrumentation
the filter wheel determine the integration time for each channel and are chosen
to provide an acceptable signal to noise ratio.
Unlike the AVHRR, the HIRS/2 is not calibrated on each scan line. Instead,
every 256 s the instrument goes into calibrate mode. First it looks at space, then
at the internal hot and cold calibration sources. Each of these is viewed for a
time equivalent to one full scan line. During the calibration period, no Earth data
are collected. This results in 3-line gaps in the scan pattern (Fig. 4.7).
Voltages from each of the HIRS/2 channels are digitized using 13 bits and are
processed by the TIROS Information Processor (TIP) as are data from the other
sounders, the SEM, the DCS, the SBUV/2, and the ERBE. These data are transmit-
ted continuously over the beacon frequencies of 136.77 or 137.77 MHz. The
data are also included in the HRPT data. Up to 250 min of these data can be
recorded on each of the tape recorders for later playback.
4.1 Operational Polar-Orbiting Satellites 101
Parameter Value
θ (4.1)
where λ is the wavelength of radiation, and D is the diameter of the lens. The
same principle applies to a microwave receiving antenna. Consider an infrared
instrument (λ ~ 10 μτη) and a microwave instrument (λ ~ 1 cm) with comparably
sized optics. Other factors being equal, the size of a scan spot for the microwave
instrument will be approximately 1000 times that of the infrared instrument, due
to the ratio of their wavelengths. That the MSU resolution is not 1000 times
worse than the HIRS/2 resolution indicates that the optics of the two instruments
are not comparable. However, it should be kept in mind that the Rayleigh criterion
severely limits the resolution of all microwave instruments. Only by using very
large antennas can the resolution of microwave radiometers be made comparable
to the resolutions of visible or infrared radiometers.
4.1 Operational Polar-Orbiting Satellites 103
FIGURE 4.9. Photograph of the Microwave Sounding Unit. [Courtesy of James C. Fischer, N O A A /
NESDIS.]
104 Chapter 4 Meteorological Satellite Instrumentation
Although externally the HIRS/2 and the MSU are similar, internally they are
quite different. In the MSU, radiation from each of the two scan mirrors enters
a feedhorn where it encounters a transducer that separates the beam into two
beams (4 channels total). Each of these beams travels via a waveguide to a device
called a Dicke switch, which switches between viewing radiation from the feedhorn
or from an internal microwave source of known temperature. In the MSU this 12
switching occurs at the rate of 1 kHz. The radiation is then mixed with an
internally generated signal at the frequency of the channel and processed by a
superheterodyne (radio) receiver. The output of the receiver is proportional to
the difference between the brightness temperature of the scene being viewed and
the temperature of the internal radiation source.
N o t e that this detector system essentially amounts to a dipole antenna. The MSU and all other
1 2
microwave radiometers measure polarized radiation. The polarization vector for the MSU rotates
with the scan angle.
4.1 Operational Polar-Orbiting Satellites 105
D E T E C T O R BLOCK
SPACE VIEW
D I R E C T I O N O F FLIGHT
1 SUN SHIELD
EARTH VIEW 2 MIRROR SCAN MECHANISM
3 P R E S S U R E M O D U L A T O R CELL
4 D E T E C T O R BLOCK
5 ELECTRONICS STRUCTURE
Parameter Value
Telescope None
Optics Simple 2-lens, 15 mm diameter, / / 2 . 4 ,
imaging system for each channel
Field of view 10°
Ground field of view
Nadir 148 km diameter
End of scan 187 km along-track x 2 4 9 km cross-track
Scan line
Time per scan 32 s
Angle from nadir ±35°
Distance from nadir ± 6 1 7 km
Number of steps 8
Scan step
Time 4s
Angle 10°
Distance between scan spots (at nadir)
Cross-track 148 km
Along-track 2 0 9 km
Gap between consecutive passes at equator 1330 km
Number of channels 3
Calibration Stable blackbody and space
Data rate 4 8 0 bits s" 1
Instrument mass 15 kg
Instrument size 26.5 x 36.5 x 26.3 cm
Average power consumption 18 W
The SSU is a scanning radiometer like all the instruments discussed above. Its
scan mirror steps in 10° increments for 8 steps from 35° left to 35° right of the
satellite track. Each step takes 4 s, and an entire scan line takes 32 s. Since the
field of view of the radiometer is 10°, the scan spots are contiguous in the cross-
track direction (Fig. 4.12), but there is an underlap in the along-track direction.
Also, the SSU scans only about two thirds as far out as does the HIRS/2 and the
MSU. Thus the outermost HIRS/2 and MSU scan spots have no corresponding
SSU data. Special handling of the SSU data are necessary in the retrieval process
to correct for this (see Chapter 6).
The SSU utilizes an interesting method for filtering the incoming radiation. All
three of its channels sense radiation within 50 c m of the center of the Q branch
- 1
of the 15-μηι C 0 band. Each channel has a cell containing C 0 in its optical
2 2
path. The C 0 cell acts as a filter with a response function which is very similar
2
to that shown in Fig. 3.17; that is, it is perfectly tuned to sense C 0 . This is 2
called gas correlation spectroscopy. The pressure inside the cells is modulated at
40 Hz. The radiation on the detector is a weighted average of the radiance emitted
by the atmosphere and by the gas in the cell. As the cell pressure increases, the
absorption lines of the C 0 in the cell broaden and filter radiation from the
2
atmosphere; that is, more of the radiation striking the detector is from the gas
4.1 Operational Polar-Orbiting Satellites 107
in the cell. As the cell pressure decreases, the lines become narrower, and the
detector better senses more radiation from the atmosphere. The mean cell pressure
determines the level to which the channel is most sensitive (Table 4.8). The
amplitude of the 40-Hz signal from each of the three channels yields information
on the vertical temperature structure.
a
The hectopascal (hPa) is used throughout the book as the unit of pressure. It is numerically equal
to the non-SI unit millibar.
Chapter 4 Meteorological Satellite Instrumentation
108
The detector for each of the three SSU channels is a flake of triglycine sulfate.
Once each 256 s, the SSU enters a calibration mode in which it first looks at
space, then at an internal blackbody source maintained near 15°C. Calibration
takes 32 s, during which time no Earth data are collected. This produces a one-
line gap in the scan pattern.
The radiances falling on the detectors are integrated for 3.6 s of each 4-s step
and are digitized to 12 bits for transmission to Earth as part of the TIP data stream.
much like a prism. The radiation again reflects from the Ebert mirror and leaves
via an exit slit. The grating is rotated to select which wavelength leaves the
Monochrometer. The entrance slit for the second half of the Monochrometer is
the same as the exit slit for the first half. The two gratings are mounted on a
common shaft to coordinate spectral scanning. The use of two monochrometers
in series improves rejection of stray light and sharpens spectral response in compar-
ison to a single monochrometer. The spectral resolution of each channel is 1.1 nm.
The Monochrometer has two operational modes. In discrete mode, the gratings
are rotated to 12 discrete positions to sample 12 discrete wavelengths (see Table
4.9). Each wavelength is sampled for 1.25 s, and 0.75 s is allowed to step to the
next position. All 12 wavelengths are sampled in 24 s. Eight seconds are allowed
for the gratings to rotate back to the first position to begin a new wavelength
scan. Each wavelength scan thus takes 32 s. In sweep mode, the gratings are
stepped such that the wavelength is scanned through the range 160-407 nm in
0.148-nm increments. Each step takes 0.1 s, so that the entire wavelength scan
See Halliday and Resnick (1962). Chapter 45 reviews the optical properties of gratings.
CALIBRATION LAMP MECHANISM DRIVE FROM
j < ELECTRONICS AND LOGIC MODULE
SOLAR IRRADIANCE
I ^X. CHOPPER ENCODER
/ /^CALIBRATION LAMP READOUT TO
S T E P P E R M T 0 R
S / TWO-POSITION ° ELECTRONICS AND
L 0 6 I C
/ / CALIBRATION LAMP , C
MODULE
\ / S\ I *L PHASE-LOCKED LOOP
/ \ \ j LAMP CHOPPER MOTOR DRIVE
V \ \ ι POWER FROM ELECTRONICS AND
) \ \ I 1 SUPPLY LOGIC MODULE
( \ \ GEAR « 1 γ , 1
FRAIN
V r^\\ FILTER
Λ * ^ W \^ C L O U D C O V E R
_ \ \ FOUR-POSITION » ^_ I „ Π/7 I \ RADIOMETER
\ \ INSTRUMENT DIFFUSER "5 |J\\ I \ W
"~^\
SCENE \ \ (SHOWN DEPLOYED) Τ CCR p/WS I 1 λ
RADIANCE \ \ I ENCODER DIODE 1^ IV. VOLTAGE/
--y\ T r r Q— €4-Η>Η>- -
/ \ \ \ "Zm" ι ENTRANCE SLIT ™ ' ·
I VS ILJ___LJ j grating
I W CHOPPER •
\ ι ·—ι ΓΊ ι <T ANALOG
\ GFAR | | ( INSTRUMENT DIFFUSER ROOF MIRROR ι ^ HOUSEKEEPING
\ J TRAIN I , yr' STOWED) RFFI FCTIONS)|-**C-]
( T W 0 ^ ^ / V ^
\~Ί ΙI ι 1 ι 1 V 7 EBERT
, BRUSHLESS OPTICAL ' MIRRORS
j , TORQUE —ENCODER "] ^ ^ / v
FIGURE 4.13. Diagram of the SBUV/2 instrument. [Courtesy of N O A A and Ball Corp., Aerospace Systems Division.]
110 Chapter 4 Meteorological Satellite Instrumentation
EBERT MIRRORS
FIGURE 4.14. SBUV/2 Monochrometer. [Courtesy of N O A A and Ball Corp., Aerospace Systems Di-
vision.]
takes about 168 s. During the next 24 s, the gratings are rotated back to the
initial position.
Retrieval of ozone profiles requires a knowledge of the fraction of solar radia-
tion reflected at each wavelength. Not only must the radiance reflected by the
Earth's atmosphere be known, but also the incident solar radiation. Ground-
based estimates of the solar radiation at the top of the atmosphere in the ultraviolet
portion of the spectrum are not useful because of the opacity of ozone. Therefore,
the SBUV/2 measures the ultraviolet solar irradiance itself. A Lambertian diffuser
plate painted with BaS0 (Kodak White) can be swung into place such that once
4
each orbit solar radiation is reflected into the instrument. The Monochrometer
can be operated in either discrete mode or sweep mode during this time.
The SBUV/2 is calibrated with an on-board mercury lamp that has eight
emission lines in the wavelength range of the instrument: 184.0, 253.7, 302.2,
313.2, 365.0, and 404.7 nm. The lamp can be positioned such that it shines
directly into the optics, or such that it first reflects from the diffuser plate, thus
checking the properties of the diffuser. During these measurements, the SBUV/2
can be operated in either discrete mode or sweep mode. Normally, discrete mode
is used during calibration and 12 wavelengths around the 253.7-nm line are mea-
sured.
The Cloud Cover Radiometer (CCR) is a separate radiometer used to detect
clouds. It operates at 379 nm using a 3.0-nm-wide filter. The CCR is aligned to
4.1 Operational Polar-Orbiting Satellites ]]]
Parameter Value
Monochrometer
Instrument type Spectral scanning
Double Ebert-Fastie Monochrometer
Optics
Depolarizer 4 elements
Chopper wheel 2 0 Hz
Mirror Ebert, 25 cm focal length
Entrance slit 0.13 x 3.0 cm
Grating 2 4 0 0 lines m m " 1
Instrument mass
Sensor module 2 5 . 6 kg
Electronics and logic module 11.2 kg
Instrument size
Sensor module 31.1 x 3 5 . 6 x 50.5 cm
Electronics and logic module 2 2 . 2 x 3 3 . 0 x 3 3 . 4 cm
Average power consumption 12 W
view the same scene as the Monochrometer, and it has the same 11.3° field of
view. The CCR shares the chopper wheel, mercury lamp and diffuser plate with
the Monochrometer.
Data from both the Cloud Cover Radiometer and the Monochrometer are
digitized to 8 bits for transmission to Earth as part of the TIP data stream.
|]2 Chapter 4 Meteorological Satellite Instrumentation
Parameter Value
Telescope None
Number of channels 5
Detector Active cavity radiometer (5)
Calibration 1. T w o internal blackbodies
2. On-board tungsten lamp
3. Direct viewing of sun
Digitization 13 bits (each channel)
Data rate 130 bits s" 1
Instrument mass 3 6 . 7 kg
Instrument size 4 5 . 7 x 50.3 x 62.5 cm
Average power consumption 20 W
4.1 Operational Polar-Orbiting Satellites 113
Baseplate
Pedestal
Main electronics
Azimuth drive
Solar aspect assy. & cable wrap
sensor (SAS)
Elevation drive assembly
Solar monitor
Instrument head
assembly (SMA)
Contamination
covers
FIGURE 4.15. Diagram of the ERBE Non-Scanner. [After Luther et al. (1986).]
The sensors of the ERBE-NS are active cavity radiometers similar to those
employed on the Solar Maximum Mission (Willson, 1979). Figure 4.16a details
the Earth sensor module. It consists of two silver conical cavities wrapped with
resistance heater windings on the outside, painted with a black coating on the
inside, and mounted inside silver cylinders. The cylinders are wrapped with resis-
tance temperature sensors and soldered back-to-back inside a temperature-con-
trolled heatsink. The heatsink can be commanded to any desired temperature
Reference cavity-
Active cavity
Heat sink
Elevation beam
Insulator
Filter dome
-Primary aperture
^-Duplex cone assembly
L
Mounting bracket
FIGURE 4.16. ERBE-NS active cavity radiometers: (a) Earth sensor module; (b) solar monitor. [After
Luther et al. (1986).]
114 Chapter 4 Meteorological Satellite Instrumentation
between ambient (~15°C) and 60°C within ±0.02°C. The reference cavity views
only the heatsink; thus the resistance of its temperature sensor is determined by
the heatsink temperature. The active cavity views radiation from the Earth (or
sun or internal calibration source), field-of-view-limiting aperture, and filter. The
active cavity is maintained at a constant temperature approximately 0.5 Κ above
the temperature of the reference cavity by supplying power to the heater windings
around the active cavity. If the irradiance from the Earth increases, less power
must be supplied by the heater windings to maintain the desired temperature
difference between the active and reference cavities. On the other hand, a decrease
in Earth irradiance requires an increase in heating to maintain the temperature
difference. The output of the sensor is the voltage across the active cavity heater
windings which is required to maintain a constant temperature difference. This
voltage (0-10 V, full scale) is digitized using 13 bits for transmission to Earth.
The detector time constant is 1-2 s.
The solar monitor (Fig. 4.16b) is very similar to the Earth monitors, except
that it has a shutter which covers the sensor for 32 s and then uncovers it for an
equal length of time. The shutter eliminates the need for actively controlled
heatsink temperature.
The four Earth-viewing sensors are differentiated by filters and field-of-view-
limiting apertures. Two sensors have no filters; they absorb essentially all radiation
incident on them in the wavelength range of 0.2 μηι to beyond 50 μ,ιτι. These
are called total channels. The other two sensors are covered by hemispherical,
fused-silica (Suprasil-Wl) domes that pass radiation in the wavelength range
0.2-3.5 μτη (see next section). These are called shortwave channels. Longwave
radiation emitted by the Earth can be calculated by subtracting the irradiance
measured by the shortwave channels from that measured by the total channels.
Two of the sensors have wide-enough fields of view (135.8°) that they sense the
Earth from horizon to horizon. These are called the wide-field-of-view (WFOV)
channels. They are also called flat-plate sensors because they sense essentially the
same radiance as would a flat plate, which has a 180° FOV. The other two sensors
have fields of view of 66.6°, which allows them to view a spot on Earth with
about a 1100-km diameter. These are called medium-field-of-view (MFOV) chan-
nels. The solar monitor has no filter, but has a field of view of 18.0°. The ERBE-
NS channels are summarized in Table 4.11.
The ERBE instruments are the most accurately calibrated instruments of their
type ever flown in space. During assembly, all important optical, electrical, and
thermal characteristics of the instruments were measured. Before launch, the
instruments were calibrated in a specially built vacuum chamber at TRW. Approxi-
mately every 14 days in orbit, the normally Earth-viewing sensors are commanded
to directly view the sun through a special port. Also approximately every 14 days,
the instruments are rotated to the stow position, where they view calibrated
blackbodies or tungsten light sources, both of which were carefully calibrated
prior to launch and are monitored in flight by calibrated platinum resistance
thermometers or photodiodes. In addition, a variety of checks of the electronics
can be made. Finally, during data reduction, several consistency checks, which
compare the instruments, such as the solar monitor and the Earth-viewing sensors
in solar calibration mode, are made. The measurement uncertainty specifications
to which each nonscanner sensor was designed and built are listed in Table 4.11.
Parameter Value
Instrument mass 3 2 kg
Instrument size 3 7 . 6 x 4 0 . 4 x 5 8 . 9 cm
Average power consumption 28 W
a
From a height of 8 5 0 km on a N O A A satellite. The resolution of the instruments from the 610-
km height of the ERBS is slightly different.
Chapter 4 Meteorological Satellite Instrumentation
116
Pedestal
(contains main
electronics)
Azimuth drive assembly
Solar viewing ports
Radiator
Electronics -
Supra si I window
(short wave Filter diamond (long wave)
channel only) >
Suprasil-Wl (short wave)
Active flake
Compensating flake
Thermistor
Cap assembly
Pivoted baffle
Detector assembly
80
70 ... \
v
i; :
SPECTRAL 6 0
- •\ "V \
\
RESPONSE 50
SHORTWAVE :
<« 40
CHANNEL i
30
V
20 :
. LONGWAVE
10 : CHANNEL
0 ι V ι ^"
0.1 1 10 100
WAVELENGTH ( ^ m )
FIGURE 4.19. ERBE Scanner spectral response functions. [After Smith et al. (1986).]
4.1 Operational Polar-Orbiting Satellites ]"\f
from a pedestal that contains the main electronics (Fig. 4.17). The instrument
consists of three optical modules that are identical except for filters (Fig. 4.18).
Light enters each module through a baffled entrance tube where it encounters a
23.03-mm focal length, /"/1.84 Cassegrainian telescope. The telescope mirrors are
aluminized glass, coated to enhance ultraviolet reflection. Before encountering
the detector, the radiation in the shortwave module passes through a suprasil
filter, while the radiation in the longwave module passes through a multicoated
diamond filter. The total module has no filter. Fig. 4.19 shows the spectral response
function of each module, and Table 4.13 describes the channels.
Just above the detector is a hexagonal field stop aperture (Fig. 4.20) that is 3°
in the cross-track direction and 4.5° in the along-track direction. This field stop
was chosen to minimize problems caused by motion of the scan mechanism during
sampling (see below).
The detector is a thermistor bolometer. A bolometer is a device that measures
radiation by measuring the temperature of an object which is warmed by absorp-
tion of the radiation. A thermistor is a solid-state device whose resistance changes
with temperature. Each ERBE-S optical module contains two thermistor flakes
about 2 mm square. One is in a temperature-controlled cavity. The other is directly
beneath the field stop aperture and is heated by radiation from the telescope. The
resistances of the two thermistors are constantly compared in a bridge network.
The voltage difference across the bridge, which results when there is an imbalance
between the cavity temperature and the active flake temperature, is digitized using
12 bits for transmission to Earth. This voltage is converted to radiance with the
help of cold and warm calibration sources of known radiance.
The scan mechanism moves at the rate of 66.7° s" and is sampled every 33
1
ms, that is, every 2.2°. Since the scan head moves continuously, the area scanned
in 33 ms will be octagonal (Fig. 4.20), and adjacent scan spots will overlap by
about 4 8 % . The time between adjacent scan lines is 4 s. Thus adjacent scan lines
also overlap (by about 60%) because during the scan period the subsatellite point
moves only about 26.1 km, but the along-track field of view is about 66.8 km.
The scan head can be rotated azimuthally so that it scans in the along-track
direction rather than in the normal cross-track direction. In this mode, each spot
along the subsatellite track is viewed once each 4 s for the approximately 10 min
that the point is in view of the satellite. These observations can be used to
check the anisotropic factors necessary to convert the ERBE-S observations into
t 0 s
t=33ms t =66ms
FIGURE 4.20. ERBE Scanner scan spots, (a) Instantaneous field of view from the ERBE-S aperture,
(b) A motion-produced ERBE-S scan spot (solid line), that is, the area sensed between t w o consecutive
samples separated by 33 ms as the scanner moves along the scan line. N o t e that the center is
continuously sampled; other areas are sampled for less time, (c) Overlap of t w o consecutive scan
spots (48%).
quantities used in radiation budget studies (see Chapter 10). Unfortunately, this
mode has not been used much in practice.
Like the ERBE-NS, the ERBE-S is very carefully calibrated both in space and
before launch, and the data are validated after collection. During each scan line,
the instrument views cold space and an internal radiance source that includes a
temperature-controlled blackbody and a tungsten lamp that is monitored by a
photodiode. Approximately once every 14 days, the instrument is rotated to view
sunlight reflected from a diffuser plate. Prior to launch, the ERBE-S was carefully
calibrated and characterized in the calibration chamber used for the ERBE-NS.
Finally, an elaborate system of data validation has been devised (Barkstrom,
1986). As examples, data from the longwave and shortwave channels of the
ERBE-S are compared with data from the total channel, and data from the ERBE-
S are integrated over the field of view of the ERBE-NS channels and compared.
This cross-checking helps ensure that the data are as error-free as possible.
4.1 Operational Polar-Orbiting Satellites 119
The former Soviet Union has had an active series of operational polar-orbiting
meteorological satellites since 1969. There have been three series of Meteor satel-
lites. Through 1991, 31 Meteor-Is, 20 Meteor-2s, and 5 Meteor-3s have been
launched (see Appendix A). Both Meteor-2 and Meteor-3 satellites are in current
operational use, although the Meteor-2 satellites are being phased out. These
satellites are described in the 1989 World Meteorological Organization report,
Information on Meteorological and Other Environmental Satellites. Tables 4.14
and 4.15 contain descriptions, excerpted from the report, of the Meteor-2 and
A scanning telephotometer for direct transmission of images in the visible spectrum ( 0 . 5 - 0 . 7 μηι),
giving a swath width of 2 1 0 0 km at a resolution of 2 km
"From W M O (1989).
A scanning telephotometer for direct transmission of images in the visible spectrum ( 0 . 5 - 0 . 7 μηι),
giving a swath width of at least 2 6 0 0 km at a resolution of 1.0 x 2.0 km or better at the subsat-
ellite point
A scanning telephotometer for global survey mode in the visible spectrum ( 0 . 5 - 0 . 7 μηι), giving a
swath width of at least 3 1 0 0 km and a resolution at the subsatellite point of 0.7 x 1.4 km or
better
A scanning IR radiometer for direct image transmission and global survey mode (spectral range
10.5—12.5 μηι), giving a swath width of at least 3 1 0 0 km; nadir point resolution 3 x 3 km
A scanning IR radiometer for global survey mode (spectral range 8 - 1 2 μηι), providing a swath
width of at least 3 1 0 0 km; resolution at the nadir point 10 x 10 km
A multispectral scanning IR radiometer (10 channels between 9.6 μηι and 15.2 μηι and 18 μηι)
for atmospheric temperature and humidity sensing; swath width 1000 km, surface resolution
5 0 km
A radiation measuring complex for recording fluxes of protons and electrons with threshold ener-
gies from 0.15 M e V to 9 0 M e V
There is an experimental complex for measuring total ozone content and vertical ozone distribu-
tion in individual regions
' F r o m W M O (1989).
120 Chapter 4 Meteorological Satellite Instrumentation
4.2.1 GMS
4.2.2 Meteosat
Launched by the European Space Agency from Kourou, French Guiana, Meteo-
sat is positioned at the prime meridian. The Meteosat spin scan radiometer returns
images in the visible and 11-μηι bands as well as the 6.7-μιη water vapor band.
In fact, Meteosat 1, launched in 1977, provided the first 6.7 μηι data from
geostationary orbit (Morgan, 1981). Beginning with Meteosat 4, these satellites
are also called Meteosat Operational Program (MOP) satellites. Table 4.17 de-
scribes the characteristics of Meteosat.
4.2.3 Insat
Insat is positioned at 74° east. Insat IB was launched from the Space Shuttle
as part of the Indian National Satellite System. Insat is the first of a new breed
of geostationary meteorological satellites which do not rely on spinning to provide
attitude stabilization. They are called three-axis stabilized-, their instruments al-
ways point toward Earth, as do the instruments on the low Earth orbiters. Insat
4.2 Operational Geostationary Satellites
121
FIGURE 4.21. Areas viewed by geostationary meteorological satellites. The solid line shows the limb;
a satellite sees nothing outside this area. The dashed line encloses the area of useful data where the
satellite is at least 10° above the horizon.
was built by Ford Aerospace (now Space Systems/Loral), which is building the
GOES I - M satellites based on the three-axis-stabilized Insat design. Table 4.18
describes the characteristics of Insat.
Value
a
Information from Rao et al. (1990).
"122 Chapter 4 Meteorological Satellite Instrumentation
Value
1
Information from Rao et al. (1990).
4.2.4 GOES
The series of geostationary satellites operated by NOAA began with the Syn-
chronous Meteorological Satellite 1 (SMS 1) launched 17 May 1974. Like TIROS
N, both SMS 1 and SMS 2 were considered experimental satellites and were
initially operated by NASA. The Geostationary Operational Environmental Satel-
lites (GOES) were commissioned and are operated by NOAA. SMS 1 and 2 and
GOES 1, 2, and 3 were designed and built by Ford Aerospace based on NASA's
Applications Technology Satellites 1 and 3 (ATS 1 and 3). GOES 4, 5, 6, and 7
were designed and built by Hughes Aircraft. GOES 8, launched as this book was
nearing publication, and its successors (GOES J - M ) are built by Space Systems/
Loral. Table 4.19 summarizes the physical characteristics of these satellites. Figures
4.22 and 4.23 illustrate the GOES 7 and 8 satellites, respectively.
Value
Detectors 1
Wavelengths 0 . 5 5 - 0 . 7 5 μπι 1 0 . 5 - 1 2 . 5 μια
Digitization 10 bits 10 bits
Angular field of view 78 x 67 /xrad 2 4 3 X 3 2 6 jitrad
Resolution at nadir 2.75 x 2.75 km 11 x 11 km
Frame size 20° x 20° 20° x 20°
Frame time 23 min 23 min
Accuracy S/N > 13 Ν Ε Δ Τ < 0.19 Κ
for 2 . 5 % albedo at 3 0 0 Κ
a
Information from Rao et al. (1990).
4.2 Operational Geostationary Satellites ]23
The spin axis is maintained closely parallel to the Earth's spin axis. The satellite
rotates at the rate of 100 rpm (10.4720 rad s ) , which provides east—west optical
-1
scanning (thus the name spin scan radiometer). North-south optical scanning is
provided by a scan mirror that moves in 192-/xrad steps. At geostationary altitude,
the Earth subtends only 17.4°. Therefore, the mirror need only scan ±10° from
the subsatellite point to view the entire hemisphere. A complete hemispherical
14
scan is made in 1821 rotations of the satellite with one mirror step after each
rotation. At 100 rpm, a hemispherical image, called a full-disk image, is completed
in 18.21 min. Usually, full-disk images are terminated short of 1821 scans thus
These figures actually refer to the radiation beam; the mirror itself moves only half this amount,
1 4
NORTH
FIGURE 4.23. Sketch of GOES 8. The face of the main body, which contains the imager and sounder
telescopes continually points toward Earth. The b o o m on which the solar sail is mounted is parallel
with the Earth's spin axis. [After Komajda and McKenzie (1987).]
eliminating a portion of the Southern Hemisphere. At the end of each image, the
mirror rapidly retraces to its northernmost position to start a new image. 15
Meteosat does not retrace; every other image starts in the south and proceeds northward.
126 Chapter 4 Meteorological Satellite Instrumentation
G O E S - E a s t (75° West)
SMS 1 VISSR 2 7 Jun 1 9 7 4 - 8 Jan 1976
GOES 1 VISSR 8 Jan 1 9 7 6 - 1 5 Aug 1 9 7 7
GOES 2 VISSR 15 Aug 1 9 7 7 - 2 6 Jan 1979
SMS 1 VISSR 2 6 Jan 1 9 7 9 - 1 9 Apr 1979
SMS 2 VISSR 19 Apr 1 9 7 9 - 5 Aug 1981
GOES 5 VAS 5 Aug 1 9 8 1 - 2 9 Jul 1984
GOES 7 VAS 25 Mar 1 9 8 7 - 2 1 Jan 1989
G O E S - W e s t (135° West)
SMS 2 VISSR 10 Mar 1 9 7 5 - 1 3 Jul 1978
GOES 3 VISSR 13 Jul 1 9 7 8 - 5 Mar 1981
GOES 4 VAS 5 Mar 1 9 8 1 - 1 Jun 1983
GOES 6 VAS 1 Jun 1 9 8 3 - 1 Aug 1984
GOES 6 VAS 2 6 Mar 1 9 8 7 - 2 1 Jan 1989
FIGURE 4.24. Sketch of the VAS instrument. [Courtesy of N O A A and Santa Barbara Research Center.]
TABLE 4.21. VISSR Atmospheric Sounder (VAS) Characteristics
Parameter Value
Size
Length 1.5 m
Radial dimension 0.65 m
Mass
Scanner 64.3 kg
Electronics module 10.5 kg
Average power consumption 40 W
a
From a height of 3 5 , 7 9 0 km above the equator.
b
Megabits per second.
]28 Chapter 4 Meteorological Satellite Instrumentation
GOES-West images start at 15 and 45 min past the hour. During special rapid-
scan periods, sectors of limited north-south extent (but full east-west extent)
are scanned in less time. Scanning periods of 15, 7.5, 5, and 3 min have been used.
After being reflected by the VAS scan mirror, radiation enters a 40.64-cm
diameter, /"/7.2, Ritchey-Chretien telescope. At the principal focal plane, the
radiation is intercepted by an array of eight optical fibers connected to eight
photomultiplier tubes sensitive to radiation in the wavelength range 0.54-0.70
μηι. This array of optical fibers is arranged in a north-south line, and each fiber
has a field of view of 25 μτζά (E-W) by 24 ^rad (N-S). Thus with each rotation
of the satellite, eight lines of visible data are swept out. The 25 x 24-^rad field
of view results in a ground resolution of about 0.9 km at the subsatellite point.
The visible channels are sampled at the rate of 500 kHz, which, at a satellite
16
rotation rate of 10.47 rad s , means that visible scan spots are separated by 21
_1
/xrad or 0.75 km east-west. Visible scan lines are separated by 24 μτζά or 0.86
km north-south. Thus visible scan lines are contiguous, but adjacent scan spots
overlap slightly.
The VAS visible data are digitized using 6 bits, but the output count is not
proportional to radiance; rather, it is approximately proportional to the square
root of the radiance. This choice was made because the noise from the photomulti-
plier tubes is proportional to the square root of the radiance. By making the
output proportional to the square root of radiance, the noise level, expressed in
counts, is nearly independent of the count value. The visible channels are calibrated
before launch, and calibration curves are available from NESDIS.
On each rotation of the satellite, 15,288 samples are taken for each of the
eight visible channels. Since a complete full-disk image is made in 1821 rotations,
it contains 14,568 lines (8 x 1821). A single full-disk visible image, therefore,
consists of 15,288 x 14,568 x 6 = 1,336,293,504 bits (159.3 megabytes), not
counting the documentation and navigation information which is normally part
of the image file. Archiving all of these data is a challenge.
Since there are eight visible channels, each digitized to 6 bits, and each sampled
at the rate of 500 kHz, the visible data are taken at the rate of 24 Mbits s - 1
(megabits per second). The addition of the infrared data boosts the total data
rate to 28 Mbits s" . This rate only occurs, however, while the instrument is
1
viewing the Earth. Ninety-five percent of the time the instrument views space and
collects no data. To decrease the data rate so that users may receive it more easily,
the following arrangement is employed. The high-data-rate, raw VAS data are
transmitted in real time to the Command and Data Acquisition (CDA) station
located at Wallops Island, Virginia. The data are transmitted back to the satellite
at a data rate of approximately 2.1 Mbits s ; which means that the data, which
_1
were collected only while the instrument viewed Earth, take one full rotation of
the satellite to transmit. These data, called stretched data, are retransmitted by
the GOES satellite to anyone with the proper receiving equipment.
Actually, the satellite spin rate is allowed to vary by ± 5 % , but the sampling rate is adjusted
1 6
Weighting NEAL
Wave- Half-power function
%nction (mW m~ sr~ cm)
2 l
detectors, used to monitor wavelengths longer than 6 /xm, and two InSb detectors,
used for the shorter-infrared wavelengths. Two of the HgCdTe detectors are 80
μτη square. They have fields of view of 192 ^rad and, therefore, Earth resolutions
of 6.9 km. These detectors are called small detectors. The two remaining HgCdTe
detectors and both of the InSb detectors are 160 μηι square, have 384-^rad fields
of view and Earth resolutions of 13.7 km. These detectors are called large detectors.
The infrared detectors are offset from the visible detectors and from each other
both in the north-south and the east-west directions to facilitate imaging. The
data are properly registered at Wallops Island during the stretching processes. To
decrease noise, the infrared detectors are cooled to 94 Κ by a radiant cooler,
which radiatively exhausts heat energy to cold space.
During each rotation of the satellite, two infrared detectors are active and are
sampled at the rate of 125 kHz. Since a scan spot near nadir moves only 3 km
between samples, the infrared channels are oversampled, that is consecutive scan
spots overlap. The small detectors are oversampled by about 57%, and the large
detectors are oversampled by about 78%. One occasionally hears that the resolu-
tion of the small detectors is 3 x 7 km ( 4 x 8 km on the older VISSR instrument).
It should be remembered that two separate aspects of resolution have been com-
bined to make this statement. The instantaneous field of view (half-power points)
for the VAS small detectors is 7 x 7 km, but the oversampling causes near-nadir
scan spots to be only 3 km apart.
The infrared data are calibrated every scan line by viewing space and an internal
blackbody. The data are digitized using 10 bits in a manner such that the counts
(0-1023) are linearly proportional to radiance. In the older VISSR instruments,
only the 11 μηι channel was used, and the data were digitized to 8 bits in a
scheme tied to equivalent blackbody temperature rather than to radiance. In this
scheme, count 000 corresponded to an equivalent blackbody temperature of
330 K, and the temperature decreased 0.5 Κ with each count until count 176
(242 K). Thereafter, the temperature decreased 1 Κ with each count until the
maximum count of 255 (163 K) was reached. This digitization scheme has the
disadvantages that (1) it is not very sensitive at colder temperatures (1-K steps),
and (2) certain count values do not occur due to rounding problems in the
conversion of sensor voltage (proportional to radiance) to equivalent black-
body temperature.
The VAS has three operating modes: VISSR, Multispectral Imaging (MSI) and
Dwell Sounding (DS). VISSR mode is a backup mode. In VISSR mode, the on-
board computer, called the VAS Processor, is turned off, and the instrument
operates much like the VISSRs on the first generation of GOES satellites. Normally
the filter wheel is set to channel 8 (11 μ,ιη, although it can be set to any position),
the two small detectors are active, and full-resolution visible and 3 x 7-km IR
data are collected. 17
MSI is the normal VAS imaging mode. Here, full-resolution visible data are
collected along with two, three, or four channels of infrared data. Two images
7-km resolution and two images at 3 x 14-km resolution, or four images may
be collected at 3 x 14-km resolution. In this mode, the filter wheel is under control
of the VAS Processor, and the scan mirror steps 192 μταά on each revolution of
the satellite.
Dwell sounding mode is used to make soundings; no visible data are collected.
In this mode, the filter wheel and the scan mirror are under control of the VAS
Processor. All 12 positions of the filter wheel are sampled sequentially. Normally,
the small detectors are used for channels 3 - 5 , 7, and 8, and the large detectors
are used for the remaining channels. The scan mirror does not step with each
19
rotation of the satellite. In order to reduce noise, the instrument dwells on the
same scan line for several programmable rotations. The number of rotations of
the satellite at each mirror position for each filter wheel position is called the spin
budget. A typical spin budget is listed in Table 4.22. The data for each channel
are averaged over the number of spins. If there are Ν spins for a particular channel,
the noise in the averaged radiance is reduced by a factor of N " . The noise is 1 / 2
further reduced by averaging the radiance over an area. If Μ pixels are averaged,
the noise in the area-averaged radiance is reduced by a factor of M ~ . Because 1/2
of the multiple spins on each scan line, only a limited latitude band can be dwell
sounded during the 30-min GOES duty cycle. The normal operational schedule
calls for a combination of MSI and DS modes to satisfy a large variety of data
users (see Gibson, 1984).
Since 1978, digital GOES data have been archived by specially adapted video
tape recorders at the Space Science and Engineering Center of the University of
Wisconsin under contract to NESDIS. In addition, a number of hard copy prints
and negatives are archived by NESDIS. GOES data from these archives may be
requested from the Satellite Data Services Division of NESDIS (see Section 4.4).
1 8
Prior to GOES 7, this two-image option was not available.
1 9
The small detectors could also be used for channels 9 and 10.
132 Chapter 4 Meteorological Satellite Instrumentation
The VAS instrument combines imaging and sounding in one instrument, which
means that imaging must be abandoned while soundings are being made. The
Imager and Sounder on GOES I - M are separate instruments capable of indepen-
dent, simultaneous operation. This increases the rate at which soundings are
available.
The GOES I - M Imager (Tables 4.23 and 4.24, Fig. 4.25) is built by ITT
Aerospace/Communication Division and is similar to the AVHRR. Like AVHRR,
the Imager has 5 channels (0.65, 3.9, 6.75, 10.7, and 12.0 μπι; see Table 4.24),
however, a water vapor channel (6.75 μπι) has been substituted for the 0.9-μπι
AVHRR channel. All five channels are scanned during every Imager frame.
The Imager scans west-to-east on one scan line followed by an east-to-west
scan on the next line. Visible data are swept out eight lines at a time with eight
silicon photodiode detectors. The 3.9-, 10.7-, and 12.0-μιη channels are swept
out two scan lines at a time, and the 6.75-μιη channel is swept out one line at a
time. At the end of each scan line, the mirror steps down (south) 224 μτζά (8
km) to begin a new scan.
The ground resolution of some of the channels is improved over that possible
with VAS. The 3.9-μιη channel has approximately four times the resolution, and
the 10.7- and 12.0-μηι channels have about twice the resolution, whereas the
0.65- and 6.75-μιη channels have about the same resolution.
The radiometric accuracy of some channels also increase (see Tables 4.22 and
4.24). Also, the digitization for the visible channel increases from 6 bits to 10
bits and the output is linear in radiance instead of proportional to the square
root of radiance. This should improve the low-light detection capabilities of the
visible channel.
In addition, the scan patterns for the GOES I—Μ Imager are much more flexible
than those currently available. For example, it is possible to suspend a full-disk
scan to perform a rapid scan of a small area and then to resume the full-disk
scan. Currently, rapid scanning must be halted during full disk scanning.
4.2 Operational Geostationary Satellites 133
Parameter Value
In-orbit calibration
IR (channels 2 - 5 ) Space and internal 290-K blackbody
Visible (channel 1) None
Size
Sensor module 115 x 80 x 75 cm
Electronics module 6 7 x 43 x 19 cm
Power supply 2 9 x 2 0 x 16 cm
Mass (3-module total) 120 kg
Power consumption 119 W, daily average
a
After Koenig (1989).
Channel 1 2 3 4 5
sr cm)
-1
a
From Koenig (1989) and Menzel and Purdom (1994).
134 Chapter 4 Meteorological Satellite Instrumentation
Parameter Value
a
After Koenig (1989).
4.2 Operational Geostationary Satellites ]35
Central
wavelength
? relaunch NE AL Bandwidth
Channel Detector {μτη) (cm- )1 (mW m~ sr~ cm)
2 x (cm~ )
l
north-south direction to begin the next scan. The scans are alternately east-to-
west and west-to-east.
The Sounder has roughly the same radiometric accuracy as the VAS instrument,
but the Sounder does not have to dwell on a point as VAS does to reduce noise.
This improves flexibility in scanning. Also, the digitization increases from 10 bits
to 13 bits per channel.
A significant improvement in the GOES I - M Sounder is that it has more
channels (19 versus 12) to retrieve atmospheric parameters. In particular, it has
more surface-sensing channels than VAS (6 versus 3) and more channels in the
less-water-vapor-sensitive shortwave IR (<6 μπι) channels (6 versus 3). These
improvements should allow the GOES I - M Sounder to better separate surface,
moisture, and temperature effects (see Hayden, 1989).
The extensive changes in the way the GOES I - M instruments work requires
a new data format called GVAR (GOES Variable; see Komajda and McKenzie,
1987). This change requires the modification of the ground equipment for all
users who receive the retransmitted data.
An interesting point is that the motion of the mirrors causes perturbations in
the satellite attitude. The independent motion of the Sounder mirror would cause
the Imager scan lines to be crooked (and vice versa). To prevent this, the GOES
I - M satellite has a Mirror Motion Compensation system that automatically antici-
pates attitude changes of the satellite (caused by the Sounder scan mirror, for
example) and makes mirror adjustments (in the Imager, for example) to compen-
sate for them (see Savides and Reseck, 1989).
4.3.1 Interferometers
The Infrared Interferometer Spectrometer (IRIS) flew on Nimbus 3 and 4.
Rather than measure several discrete wavelengths by means of filters, IRIS mea-
sured a broad spectrum from 500 to 2000 cm" (5-20 μηι) in 5 - c m steps.
1 -1
4.3 Other Satellite Instruments 137
Nimbus
1 2 3 4 5 6 7
APT APT HRIR BUV ESMR ERB czcs
AVCS AVCS IDCS FWS ITPR ESMR ERB
HRIR HRIR IRIS IDCS NEMS HIRS LIMS
MRIR IRLS IRIS SCMR LRIR SAM II
MRIR IRLS SCR PMR SAMS
MUSE MUSE THIR SCAMS SBUV/TOMS
RTTS SCR T&DRE SMMR
SIRS SIRS THIR THIR
THIR TWERLE
RTTS
* Abbreviations:
APT Automatic Picture Transmission System
AVCS Advanced Vidicon Camera System
BUV Backscatter Ultraviolet Spectrometer
CZCS Coastal Zone Color Scanner
DRIR Direct Readout Infrared Radiometer
ERB Earth Radiation Budget Experiment
ESMR Electrically Scanning Microwave Radiometer
FWS Filter Wedge Spectrometer
HIRS High Resolution Infrared Radiation Sounder
HRIR High Resolution Infrared Radiometer
IDCS Image Dissector Camera System
IRIS Infrared Interferometer Spectrometer
IRLS Interrogation, Recording, and Location System
ITPR Infrared Temperature Profile Radiometer
LIMS Limb Infrared Monitor of the Stratosphere
LRIR Limb Radiance Inversion Radiometer
MRIR Medium Resolution Radiometer
MUSE Monitor of Ultraviolet Solar Energy
NEMS Nimbus Ε Microwave Spectrometer
PMR Pressure Modulator Radiometer
RTTS Real Time Transmission Systems
SAM II Stratospheric Aerosol Measurement II
SAMS Stratospheric and Mesospheric Sounder
SBUV Solar Backscatter Ultraviolet
SCAMS Scanning Microwave Spectrometer
SCMR Surface Composition Mapping Radiometer
SCR Selective Chopper Radiometer
SIRS Satellite Infrared Spectrometer
SMMR Scanning Multichannel Microwave Radiometer
T&DRE Tracking and Data Relay Experiment
THIR Temperature Humidity Infrared Radiometer
TOMS Total Ozone Mapping Spectrometer
TWERLE Tropical Wind Energy Conversion and Reference Level Experiment
]38 Chapter 4 Meteorological Satellite Instrumentation
The Limb Radiance Inversion Radiometer (LRIR), the Limb Infrared Monitor
of the Stratosphere (LIMS), and the Stratospheric and Mesospheric Sounder
(SAMS) are infrared sounders, but they scan the atmosphere at the edge of the
Earth (the limb) instead of looking down (Fig. 4.28). These instruments have
good vertical resolution and can make accurate measurements of the temperature
and trace-gas concentration in the mesosphere, stratosphere, and (in the absence
of clouds) the upper troposphere (see Chapter 6).
The Stratospheric Aerosol Measurement I Experiment (SAM I, flown on Apollo-
Soyuz), SAM II (flown on Nimbus 7), the Stratospheric Aerosol and Gas Experi-
ment (SAGE I, which flew on a dedicated satellite), and SAGE II (which flies on
the Earth Radiation Budget Satellite) are called solar occultation experiments.
They sample the atmosphere by measuring the extinction of sunlight during
WAVENUMBER (cm" ) 1
FIGURE 4.27. Infrared spectrogram of the Earth's atmosphere at 15.1°N and 144.7°W on 2 7 April 1970
taken by the IRIS instrument. [After Liou (1980). Reprinted by permission of Academic Press.]
4.3 Other Satellite Instruments 139
satellite sunrise and sunset. They can measure the concentrations of aerosols
(Chapter 8) as well as gases that have absorption spectra in the solar spectrum,
such as ozone, water vapor, and nitrogen dioxide.
at 19.35 GHz (1.55 cm). ESMR-6 measured both vertically and horizontally 21
polarized radiation at 37.0 GHz (0.81 cm). These instruments are unique in the
fact that they had phased array antennas, which means that the beam was scanned
electronically; no moving parts were required. ESMR-5 scanned through the
satellite subpoint and had a spatial resolution of 25 km at nadir. ESMR-6 scanned
in a cone ahead of the satellite in such a way as to keep the zenith angle of the
satellite, as measured at the scan spot, nearly constant. ESMR—6 had a spatial
resolution of about 20 x 40 km independent of scan position. The ESMRs were
meteorologically significant because they can be used to estimate precipitation
rate (see Chapter 9). They are also sensitive to sea ice.
The Scanning Multichannel Microwave Radiometer (SMMR) flew on both
Nimbus 7 and Seasat. It measured both horizontally and vertically polarized
radiation at five wavelengths, 4.54, 2.8, 1.66, 1.36 and 0.81 cm, with resolutions
Electric field vector parallel to the surface at the scan spot and perpendicular to the beam.
Electric field vector perpendicular to both the above horizontal vector and to the beam.
140 Chapter 4 Meteorological Satellite Instrumentation
ranging from 20 km (at 0.81 cm) to 100 km (at 4.54 cm). This was accomplished
using a rotating scan mirror and six Dicke radiometers much like that of the MSU.
SMMR was designed to measure several meteorologically important parameters,
among them: precipitation rate, column-integrated atmospheric water vapor and
liquid water, sea surface temperature and wind speed, soil moisture, and sea ice
concentration.
Finally, the Special Sensor Microwave Imager (SSM/I) is currently flying on
some of the DMSP satellites. Similar in construction to the SMMR, the SSM/I
measures both horizontally and vertically polarized radiation at 19.35, 37.0,
and 85.5 GHz, and vertically polarized radiation only at 22.235 GHz. Measure-
ments at 85 GHz are new, and they should provide improved detection of rain
and clouds.
Most satellite-borne meteorological instruments are passive-, that is, they mea-
sure radiation provided to them by the Earth. Active instruments, like radar,
transmit their own signal and measure the properties of the returned radiation.
Because nature does not necessarily provide optimal amounts of radiation, active
instruments can be designed to be more sensitive than passive instruments to
certain parameters. Active instruments, however, require much more electrical
power to operate than passive instruments. Seasat had three active instruments,
one of which was useful for meteorology. The Seasat-A Satellite Scatterometer
(SASS) was a 2.05-cm radar that measured the backscatter coefficient from the
sea surface. The backscatter coefficient is related to surface wind speed and
direction. By observing the same patch of ocean as the satellite approached and
as it left an area, these two parameters could be estimated (Jones et ai, 1979;
see Chapter 7). Unfortunately, Seasat failed due to a short circuit after only 104
days in orbit, but not before demonstrating the feasibility and importance of
active instrumentation.
Finally, instruments flown on DMSP satellites (in addition to the SSM/I dis-
cussed above) are of interest. The primary DMSP instrument is the Operational
Linescan System (OLS), which is a two-channel (visible and ll-μπι infrared)
imager (Dickinson et ai, 1974). It differs from the AVHRR in three interesting
ways. First, it has extremely high spatial resolution of about 600 m. Second, it
also has a variable scan, which causes the distance between pixels on the same
scan line to be the same near nadir and at the ends of the scan line. This makes
images easy to display and grid. Third, the visible sensor on the OLS has a much
broader bandwidth than the AVHRR visible channel. It covers approximately
the same spectral region as AVHRR channels 1 and 2 combined. This greater
bandwidth means that more radiation from a given scene is sensed, which means
that the instrument has greater sensitivity. Some OLS visible data have been
4.4 Satellite Data Archives 141
collected at night using reflected moonlight, and city lights are detectable on
OLS images.
The data from the OLS are archived mostly in the form of positive transparen-
cies, which can be useful for studies requiring higher resolution than that available
from the NOAA satellites or for studies in which digital data on magnetic tape
are not appropriate.
DMSP satellites have also carried infrared (SSH/T) and microwave (SSM/T)
sounders.
User services
Bibliography
Ball Aerospace Systems Division (1986). SBUV/2 Final Engineering Report for Flight #1 (S/N002).
Document N o . B 6 8 0 2 - 9 1 , Boulder, C O .
Barkstrom, Β. R. (1986). ERBE data and its validation. Preprints: Second Conference on Satellite
Meteorology/Remote Sensing and Applications. American Meteorological Society, Boston, pp.
J1-J3.
Barnes, J. C , and M. D. Smallwood (1982). TIROS-N Direct Readout Services Users Guide. N O A A /
NESS, Washington, D C .
Chen, H. S. (1985). Space Remote Sensing Systems: An Introduction. Academic Press, Orlando.
Clark, J. D. (ed.) (1983). The GOES User's Guide. N O A A / N E S D I S , Washington, D C .
Dickinson, L. G., S. E. Boselly III, and W. S. Burgmann (1974). Defense Meteorological Satellite
Program (DMSP) User's Guide. Air Weather Service Tech. Rep. AWS-TR-74-250, Scott Air Force
Base, Illinois.
Gibson, J. (ed.) (1984). GOES Data Users' Guide. N O A A / N E S D I S , Washington, D C .
Halliday, D., and Resnick, R. (1962). Physics. John Wiley, N e w York.
Hayden, C. M. (1988). GOES-VAS simultaneous temperature-moisture retrieval algorithm. / . Appl.
Meteor., 2 7 , 7 0 5 - 7 3 3 .
Hayden, C. M. (1989). GOES I - M sounding products. Program: GOES I-M Operational Satellite
Conference. April 3 - 6 , Arlington, VA, pp. 1 2 6 - 1 3 0 .
Jones, W. L., P. G. Black, D . M. Boggs, E. M. Bracalente, R. A. Brown, G. Dome, J. A. Ernst, I. M.
Halberstam, J. E. Overland, S. Peteherych, W. J. Pierson, F. J. Wentz, P. M. Woiceshyn, and
M. G. Wurtele (1979). Seasat scatterometer: Results of the Gulf of Alaska workshop. Science,
204, 1 4 1 3 - 1 4 1 5 .
4.4 Satellite Data Archives 143
Kidwell, Κ. B. (1986). NOAA Polar Orbiter Data User's Guide. N O A A / N E S D I S Satellite Data
Services Division, Washington, D C .
Koenig, E.W. (1989). Characteristics of the GOES I - M imager and sounder. Program: GOES I-M
Operational Satellite Conference. N O A A , Washington, D C , pp. 1 6 8 - 1 7 5 .
Komajda, R.J., and K. McKenzie (1987). An Introduction to the GOES I-M Imager and Sounder
Instruments and the GVAR Retransmission Format. N O A A Tech. Rep. NESDIS 3 3 , Washing-
ton, D C .
Kopia, L. P. (1986). Earth radiation budget experiment scanner instrument. Rev. Geophys., 24,
400-406.
Lauritson, L. G., Nelson, G. J., and F. W. Porto (1979). Data Extraction and Calibration ofTIROS-N/
NOAA Radiometers. N O A A Tech. M e m o . NESS 107, Washington, D C .
Liou, K.-N. (1980). An Introduction to Atmospheric Radiation. Academic Press, N e w York.
Luther, M. R., J. E. Cooper, and G. R. Taylor (1986). The Earth radiation budget experiment
nonscanner instrument. Rev. Geophys., 2 4 , 3 9 1 - 3 9 9 .
Menzel, W. P., and J. F. W. Purdom (1994). Introducing G O E S - I : The first of a new generation of
Geostationary Operational Environmental Satellites. Bull. Am. Meteor. Soc, 75, 7 5 7 - 7 8 1 .
Morgan, J. (1981). Introduction to the Meteosat System. European Space Agency, SP-1041, Paris
Cedex, France.
National Aeronautics and Space Administration. Observations from the Nimbus I Meteorological
Satellite. N A S A SP-89, Washington, D C .
National Aeronautics and Space Administration (1966). Nimbus II Users' Guide. Goddard Space
Flight Center, Greenbelt, M D .
National Aeronautics and Space Administration. The Nimbus III User's Guide. Goddard Space Flight
Center, Greenbelt, M D .
National Aeronautics and Space Administration (1970). The Nimbus IV User's Guide. Goddard Space
Flight Center, Greenbelt, M D .
National Aeronautics and Space Administration (1972). The Nimbus 5 User's Guide. Goddard Space
Flight Center, Greenbelt, M D .
National Aeronautics and Space Administration (1975). The Nimbus 6 User's Guide. Goddard Space
Flight Center, Greenbelt, M D .
National Aeronautics and Space Administration (1978). The Nimbus 7 Users' Guide. Goddard Space
Flight Center, Greenbelt, M D .
National Aeronautics and Space Administration (1981). Handbook of Sensor Technical Characteris-
tics. Washington, D.C.
National Environmental Satellite, Data, and Information Service (1987). Operational VAS Mode
AAA Format, Version 2.6, N O A A / N E S D I S , Washington, D C .
Rao, P. K., S. J. Holmes, R. K. Anderson, J. S. Winston, and P. E. Lehr (eds.) (1990). Weather
Satellites: Systems, Data, and Environmental Applications. American Meteorological Society,
Boston.
Savides, J., and K. G. Reseck (1989). GOES I - M system characteristics. Program: GOES I-M
Operational Satellite Conference. N O A A , Washington, D C , pp. 2 8 5 - 3 1 4 .
Schwalb, A. (1978). The TIROS-NINOAA A-G Satellite Series. N O A A Tech. M e m o . NESS 9 5 ,
Washington, D C .
Schwalb, A. (1982). Modified Version of the TIROS-NINOAA A-G Satellite Series (NOAA E-J)—
Advanced TIROS-N (ATN). N O A A Tech. M e m o . NESS 116, Washington, D C .
Smith, G. L., R. N . Green, B. A. Wielicki, J. T. Suttles, and L. M. Avis (1986). ERBE data inversion.
Preprints: Second Conference on Satellite Meteorology IRemote Sensing and Applications. Ameri-
can Meteorological Society, Boston, pp. J 9 - J 1 3 .
Smith, W. L., Η. M. Woolf, Η. B. Howell, H.-L. Huang, and Η. E. Revercomb (1987). The simul-
taneous retrieval of atmospheric temperature and water vapor profiles—application to measure-
ments with the High spectral resolution Interferometer Sounder (HIS). In A. Deepak, H. Fleming,
and J. Theon (eds.), Advances in Remote Sensing Retrieval Methods, A. Deepak Publishing,
pp. 1 8 9 - 2 0 2 .
144 Chapter 4 Meteorological Satellite Instrumentation
volumes, each the size of this book, to show examples of all of the features that
can be identified and analyzed, and more are being discovered as we write. Over
three decades of weather satellites have produced millions of images—any one
of which would provide enough material for a complete seminar. In one chapter
we cannot hope to do justice to this broad a subject. Instead, we offer two things:
an overview of the concepts that every student of satellite meteorology should
know about image interpretation, and a bibliography of works that contain
detailed instruction on the subject.
The purpose of meteorological satellite image interpretation is to relate signifi-
cant features in the image to physical processes that are occurring, or have occur-
red, in the atmosphere. For example, we want to identify clouds. The shape or
texture of the cloud or the proximity to geographic features on the Earth's surface
can tell us much about the dynamics occurring in the atmosphere. The simple
existence of a cloud on a satellite image can lead us to a number of conclusions,
or questions, such as, what was the mechanism for formation of the cloud? The
mechanism could be mechanical lifting, it could be convection due to heating, or
145
]4£ Chapter 5 Image Interpretation
the cloud could be the residual of a cloud mass that has been advected far from
its point of origin. Image interpretation can give us some clues to help identify
the mechanism. That knowledge can in turn help us determine the present and
future hydrodynamic state of the atmosphere. The objectives of image interpreta-
tion are to (1) detect and identify features such as clouds or other obscuring or
radiating phenomena on the image and (2) determine the physical mechanisms
that produce or sustain those features.
Much of the research in image interpretation being done today involves the
development of techniques to automate the analysis of digital imagery. Over the
past few decades, the primary analysis tool has been the interpretive skill of the
analyst who manually views the satellite image and makes a subjective assessment
of the features it contains. The operational meteorological satellites that are
currently flying produce more than 1 0 bytes of image data in a year. With the
13
average imaging system able to display 10 bytes at a time, this amounts to more
6
than 10 images—far too many to analyze manually. Some progress has been
7
The term satellite imagery can be used to describe two similar, yet unique
displays: photo and digital. It is important to differentiate between the two, as
some analysis techniques work only on digital images. We refer to a hard-copy
image (such as a photograph or laser print copy) as a photo image. Satellite
imagery that is typically stored on magnetic or optical media is referred to as
digital imagery, due to the nature of the collection and archive process. Photo
images are easy for an analyst to handle, but digital images can be processed in
many more ways.
Most meteorological satellite imaging devices scan the Earth scene in a manner
analogous to television raster scans. Radiance measurements are taken along
1
each scan line at time steps which produce a series of elements along that scan
(refer to Fig. 2.16). These elements are assigned a unique location within the
image by referring to a line (or scan) number and an element (or sample) number.
The radiance measured by the satellite sensor at each element location is stored
as a single digital value. When displayed on an imaging device such as a color
or black-and-white (B/W) monitor, each point on the grid is referred to as a
picture element or pixel. Digital values are used to assign a color or gray shade
1
Some devices, such as the DMSP SSM/I scanner (Section 5.2), take samples in a helical scan pattern.
5.2 Spectral Properties 147
to pixels on the final satellite image. The resultant image is displayed either on
a monitor or as a photographic (hard-copy) image.
The resolution of a satellite instrument can be defined as the size of the smallest
element in the Earth scene that can be resolved by the instrument. It is a function
of the field of view of the sensor optics and the distance to the Earth surface.
One can also think of it as the length of the side of a box on the surface of the
Earth from which radiation is gathered during each time step of the scanning
process (see Chapter 4.). These boxes are sometimes called "footprints." For a
variety of reasons, the size of this footprint varies from one channel to the next.
For example, the visible channel on a satellite may have a resolution of 1 km,
whereas the infrared channel has a resolution of 8 km. The resolution of a satellite
instrument is normally specified separately for each channel.
The resolution of a satellite instrument is usually for the satellite subpoint (the
point on the Earth directly beneath the satellite). As the satellite scans farther
away from the subpoint, the area on the surface of the Earth that is viewed becomes
progressively larger because of the curvature of the Earth and the obliquity of
view. Geostationary meteorological satellites, because of their altitude, can view
the Earth scene out to great distances from the subpoint; thus the resolution varies
greatly with distance from the subpoint. Section 5.4 discusses this degradation as
a source of error when doing image analysis.
It is important to note that the resolution of the satellite instrument may be
quite different from the resolution of an image produced from the data. For
example, the resolution of a satellite instrument may be 1 km, but an image
constructed from the data might be displayed at a resolution of 10 km by averaging
100 (10 x 10) of the original pixel values or by selecting (sampling) every tenth
pixel and every tenth line.
By using a filter, placed in front of the satellite sensor, radiation can be measured
at specific wavelengths. The filter allows radiation within a narrow segment of
the electromagnetic spectrum, called a spectral interval, to reach the sensor. These
spectral intervals are commonly referred to as channels or bands (see Chapter 4).
Most meteorological satellites make measurements in the visible and infrared
portions of the electromagnetic spectrum. Some also have additional channels
which vary from the ultraviolet (100—400-nm band) to the microwave region
(0.15-6.0 cm—often listed as a frequency range from 200 to 5 GHz). Chapter
4 contains tables with the spectral bands of a wide range of meteorological
satellites. In the remainder of this chapter, we discuss the properties of the more
common of these spectral intervals and their application to image interpretation.
7 satellite. Visible imagery generally offers the highest spatial resolution and
provides us with a view of the Earth that closely matches our senses. Land, clouds,
and ocean are easily discernible. One obvious limitation to visible data is that
they are available only from the sunlit portion of the Earth. One exception to
the daytime-only limitation is a sensitive instrument (the OLS) on board the
Defense Meteorological Satellite Program (DMSP) satellites, which can sense
features that are illuminated at night by moonlight.
The infrared channels are most often between 1 and 30 μπι. The most common
infrared band for meteorological satellites is in the 10-12.5-μτη window, in which
the atmosphere is relatively transparent to radiation upwelling from the Earth's
surface. When the word infrared is used alone to describe an image, it is nearly
always in the 10-12.5-/xm window rather than in another portion of the electro-
magnetic spectrum.
Figure 5.2 shows the infrared image that corresponds to Fig. 5.1. Again, clouds,
land, and water are easily discerned. Because infrared radiation can be related to
5.2 Spectral Properties ]49
the temperature of the emitting body, and because the troposphere generally cools
with height, this helps us interpret the atmospheric processes occurring within
the scene. An important characteristic of the infrared channels is their ability to
provide images at night. This provides continuous coverage of cloud evolution
over a full 24-hour period.
Normally in image processing, images are displayed such that the greater the
radiance, the brighter the pixel. In satellite meteorology, however, infrared images
are normally inverted; that is, the larger the radiance from an element, the darker
the pixel. This way, clouds, which are usually colder than the surface, appear
white, and the warmer ground or ocean surface appears darker than clouds, as
in visible images.
The water-vapor channels are so named because the satellite measures radiation
in water-vapor absorption bands. Several wavelengths can be used, but the most
common is centered around 6.7 μηι. At this wavelength, most of the radiation
sensed by the satellite comes from the atmospheric layer between 300 and 600
]50 Chapter 5 Image Interpretation
hPa (Morel et al., 1978); thus it measures middle levels of the troposphere.
Figure 5.3 shows the water-vapor image taken at the same time as Figs. 5.1
and 5.2. Note that the highest clouds can be seen, but surface features cannot be
detected because this is not a window channel. Instead, dramatic swirls and eddies
of water vapor are seen where high clouds are not present. Water-vapor imagery
is commonly looped (see Section 5.3) to display motions in cloud-free regions of
the atmosphere.
One way to qualitatively interpret water-vapor imagery is to say that it approxi-
mates the relative humidity of the midtroposphere. Consider a bright area and a
dark area in a water-vapor image. In the bright area, the satellite measures less
radiance than in the dark area. Less radiance means either that the atmosphere
is colder at the same level as in the dark area or that there is more water vapor
present in the bright area so that the satellite senses a higher, and therefore colder,
level. In either case, the relative humidity is likely to be higher in bright areas
than in dark areas. Bright and dark areas may also indicate rising and sinking
motions, respectively.
Some atmospheric phenomena are more easily identified by using imagery from
more than one channel. Figure 5.4 illustrates a simple two-channel scheme for
identifying clouds, and Figs. 5.5 and 5.6 are visible and infrared image pairs to
which the scheme can be applied. Low-level stratus is often difficult to identify
on an infrared image because the cloud may have a radiative temperature that is
near that of the surface (see the area north of the hurricane in Fig. 5.5, for
example). On a visible image, however, stratus clouds appear bright in contrast
to the darker background of land or water. As another example, the detection
of thin cirrus clouds can be difficult on a visible image because they are almost
transparent to visible light. These same clouds, however, are cold and mostly
opaque to infrared radiation from the warmer surface below and thus present a
strong cold signal on the infrared image (see the northwest corner of Fig. 5.6).
Q Thin Deep
_l
Ο Cirrus Convection
Ü
Έ
No Low
< Clouds Clouds
DARK BRIGHT
VISIBLE
FIGURE 5.4. Schematic diagram of a bispectral cloud classification scheme.
]52 Chapter 5 Image Interpretation
FIGURE 5.5. DMSP visible image (top) sampled concurrently with infrared image (bottom), showing
low clouds and cirrus.
FIGURE 5.6. DMSP visible image (top) and concurrent infrared image (bottom) of the eastern North
Pacific Ocean and the western United States, showing low-level stratus and ship tracks.
154 Chapter 5 Image Interpretation
Deep convection or thick cirrus is both cold and optically thick and therefore is
bright in either a visible or an infrared image. A further example of multispectral
analysis using 3.7-/xm imagery is discussed in Section 5.5.4.4, and the use of
multispectral techniques with microwave imagery was mentioned above.
Color enhancement, as the name implies, involves the assignment of colors (or
gray shades) to each pixel of an image based on the value of the pixel. The
standard technique for color enhancement of imagery is to apply a lookup table
to assign a specific color or hue to each possible numerical value that an image
pixel can take on. A second technique involves the generation of (red-green-blue)
RGB images using multispectral data.
5.3.1.1 Lookup Tables
Since an image is composed of pixels with discrete digital values, we can assign
a color or monochromatic gray shade to each value to produce an enhanced
image. Today's 8-bit image display systems allow digital values that range from
0 to 255 at each pixel location, so we can employ up to 256 colors or shadings
to highlight certain features of the image.
Infrared imagery is well suited to the use of a color enhancement table as
shown in Plate 3. Each infrared pixel value corresponds directly to an equivalent
blackbody temperature. A lookup table, sometimes referred to as an enhancement
table, is used to assign a specific color to each possible pixel value or, more
commonly, to a range of values. Thus each color represents a temperature or
range of temperatures. A temperature scale should be affixed to such images to
document the enhancement.
Color enhancement isoiseful for locating large convective cells that reach from
just above ground level to 20 km above the surface, passing through a large range
of temperatures. When viewed from the top, the cloud mass around the cell has
a range of temperature values from a temperature near that of the surface to one
5.3 Image Enhancement Techniques 155
that may be 80 Κ cooler at the cloud top. Color enhancement is also useful for
distinguishing clouds from ground or water surfaces.
Note that enhancement need not be "color." Figure 9.12 shows a gray-scale
enhancement called the MB curve which has been used for many years to locate
convective areas on infrared satellite images.
Color-enhanced images are sometimes referred to as false color or pseudocolor
images because color has been introduce into a monochromatic image.
5.3.2.1 Looping
Looping of satellite imagery has become an indispensable technique for not
only identifying features on a satellite image but also graphically displaying the
life cycle of synoptic and mesoscale meteorological features. An image loop is
created by rapidly displaying in succession a series of images to either create a
sense of movement (a time series of images) or to compare different channels
from the same time. Looping of imagery is useful for quantifying the movement
of features such as hurricanes, frontal systems, or large thunderstorm complexes.
The loop allows an analyst to quickly identify those features that are moving by
156 Chapter 5 Image Interpretation
comparing one image to the next and gives a good overview of both large and
small-scale motions. Short-term forecasts, called nowcasts, are made by extrapo-
lating ahead in time the feature seen in loops. We cannot show readers a loop in
this book, but loops are commonly shown on television weather broadcasts.
Looping is an efficient technique for identifying the misalignment of imagery
because even slight movements of land features are easily seen in loops. Looping
can also be used to identify features that have a unique spectral signature. For
example, a loop of visible and infrared images from the same time can be used
to identify cirrus or low stratus clouds.
5.4.1 Navigation
The process of geolocating pixels is also called navigation and requires current
satellite orbit and attitude parameters plus algorithms that utilize these parameters
to calculate the latitude and longitude of each pixel or (the inverse problem)
which pixel is closest to a particular latitude-longitude point. This process is
described in detail in Chapter 2.
5.4.2 Gridding
Once navigation has been accomplished, and each pixel is assigned a latitude
and longitude, the next step is to apply a grid to aid in the visualization of where
image features are located geographically. The most common is the geopolitical
boundary grid. Another common grid is a latitude—longitude grid, which consists
of parallels and meridians placed at the proper location on the image. After
navigation has been performed, the image may also be remapped into another
coordinate system.
5.4.3 Remapping
synoptic weather charts, the images are remapped into a common projection such
as polar stereographic or Mercator. Figure 5.7 shows data from four geostationary
satellites remapped into a Mollweide projection. The remapping process is done
one pixel at a time and can be computer-intensive for large images. (A 1024 x
1024 pixel image contains over 1 million pixels.)
5.4.4 Alignment
The proper navigation or remapping of an image does not guarantee that two
images taken at different times will be aligned geographically. This is due to
perturbations in the satellite's orbit and attitude parameters. The resulting mis-
alignment is most noticeable when looping images or compositing (overlaying)
successive images in time.
Several methods have been used to correct for misalignment with varying
degrees of success. One technique relies on the manual navigation of an image
by location of known geographic landmarks. Several landmarks are first identified
on an image. Because the latitude, longitude, and elevation of the landmarks are
known, the orbit and attitude parameters used to navigate the image can be
adjusted to force the identified pixels to have the correct latitude and longitude.
The new parameters are then used to renavigate the entire image.
Another technique involves warping the image to stretch or shrink it until it
fits a predefined grid. This technique also requires specific landmarks. By using
a predefined grid such as a geopolitical boundary map, the image is warped to
fit the graphic display. Warping degrades the integrity of data values due to
sampling or averaging. Both of these techniques are less accurate over the oceanic
regions where there are few recognizable geographic features. The top portion of
FIGURE 5.7. Mollweide projection of data from four geostationary satellites: GMS 4, GOES 7, Meteosat
3 , and Meteosat 5.
5.4 Geolocation and Calibration
Fig. 5.8 shows a poorly aligned GOES 7 image. Note the large error in the
geolocation of the image with respect to the graphic. The bottom portion of Fig.
5.8 shows the result after applying a third-order polynomial warping algorithm
to the image. The image now fits the graphic that was generated by the original
navigation parameters, so the navigation can be applied to the warped image to
remap it into any other projection.
Looping of two images during the alignment process, sometimes referred to
as flickering or toggling, can readily identify geographic features that are not
aligned. When properly aligned they appear as one stationary image when toggled,
but transient cloud, smoke, or airborne sand appear to move. This alignment
process is critical for any quantitative movement analysis because the loop should
show the movement of cloud features over a fixed geographic background. If the
image is not properly aligned, the movement of clouds is biased by the amount
of shift in the position of the geographic background.
5.4.5 Merging
5.4.6 Normalization
When merging images from different satellites, it is important to note that
differences in the calibration of each individual satellite sensor can cause problems
if not taken into account. For example, on a GOES image, a digital count of 200
may represent a temperature of — 35°C, whereas the same count on a GMS image
may represent a temperature of -25°C. Further, the calibration of the two satellites
may differ. To account for these differences, one of the satellites must be normal-
ized to the other, so that a digital count value represents the same temperature
for both.
The process of normalization requires the extraction of digital values from co-
located pixels on overlapping portions of adjacent images. This is normally done
on a range of values, such as cool ocean or land points and very cold cloud
masses. After a sufficient number of collocated points have been measured, the
pairs of brightness temperatures derived from the digital values are fed into a
regression scheme to produce a set of normalization coefficients. These coefficients
are then applied to one of the images to bring it in line with the other. Because
Ί£0 Chapter 5 Image Interpretation
FIGURE 5.8. A poorly gridded GOES 7 image (top) and the same image after remapping to account
for navigation error (bottom).
5.4 Geolocation and Calibration
they under-fly all geostationary satellites, NOAA polar orbiting satellites have
been used to normalize all of the geostationary satellites that are used in the
International Satellite Cloud Climatology Project (ISCCP) global cloud product
(see Chapter 8). Image normalization should be done before doing any quantitative
analysis of merged imagery.
5.4.7 Errors
There are a number of factors that limit our ability to perform the analysis of
satellite imagery. These factors include viewing angle biases, satellite sensor/lens/
filter/mirror imperfections, ground station equipment and processing, sensor lag,
signal contamination/attenuation, navigation/location errors, and the difficulty
in distinguishing land/water/cloud. The severity of a source of error varies with
each satellite system, and in some cases, a specific error may not have an adverse
affect on the image analysis (e.g., a location error with a magnitude of 2 km has
little effect on analysis of data that is sampled at 25-km resolution). Algorithms
have been developed to account for several of these errors; however, the analyst
should be aware of their presence when making quantitative assessments from
satellite imagery.
5.4.7.1 Attenuation
Attenuation is a decrease in the amount of radiation reaching the satellite due
to absorption or scattering by the intervening medium. The result is that clouds
appear cooler and subsequently higher in the atmosphere. This error is most
pronounced in the infrared bands where the primary absorbers are water vapor,
carbon dioxide, and ozone. Errors in temperature can be on the order of 10 Κ
too cold, which is approximately a 1.5-km error in height. Attenuation increases
with distance from the satellite subpoint and is the cause of the cooler temperatures
at the extreme edges of a full-disk geostationary satellite image (referred to as
limb darkening).
5.4.7.3 Contamination
Contamination is caused by radiation reaching the sensor from beneath a cloud.
This problem is most significant with thin clouds, such as cirrus. The additional
radiation makes the cloud appear warmer and, therefore, lower in the atmosphere.
This error is difficult to account for and, similar to background contrast errors,
can most often be detected by sensing in multiple channels.
162 Chapter 5 Image Interpretation
5.4.7.4 Displacement
Displacement occurs as a result of the viewing-angle geometry and the projec-
tion of the image data onto a 2D plane for display as an image or photograph.
Figure 5.9 shows how the top of a cloud is displaced to a point away from the
satellite subpoint. This error, discussed by Warren (1977) and Weiss (1978), can
be corrected if the height of the cloud, the satellite position, and the coordinates
of the cloud are known. As one would expect, displacement errors become more
significant further from the satellite subpoint. Data from high latitudes collected
by the GOES satellites are often difficult to use because of this error.
5.4.7.5 Foreshortening
Foreshortening is due to the effect of the Earth's curvature on the resolution
of an image. As an instrument scans from the subsatellite point toward the limb
of the Earth, pixels cover larger geographic areas and, with most instruments,
are spaced further apart. Figure 5.17 demonstrates this error for a NOAA polar-
orbiting satellite. For a geostationary satellite, foreshortening reduces the useful-
ness of the data beyond about 60° latitude from the subsatellite point. However,
some remapping techniques have been developed to ameliorate this problem. The
Operational Linescan System (OLS) on board the DMSP satellites has a special
(nonlinear) scan mechanism that causes all pixels to be the same distance apart
in the cross-track direction (compare Figs. 5.11 and 5.17). It does not change the
increase in size of the pixels toward the limb, however.
Markus, 1982):
dU'M
= l(L-L ) (5.1)
dt
M
Apparent
Cloud Position
Satellite —~
Line - of - Sight
where τ is called the time constant, and is about equal to 6 //,s for the VISSR or
VAS instruments according to Adler and Markus (1982). Since the VAS samples
an IR line every 8 /xs, a rough approximation is that the output of the sensor is
75% of the radiance from the current pixel, 20% of the previous pixel, and 5%
of the pixel before that one.
In addition, the IR sensors on the GOES satellites are over sampled; that is,
the east—west distance between pixels is less than the IFOV. Therefore, the
instrument senses a cloud, for example, before the center of the pixel actually
reaches the cloud and after the center of the pixel has left the cloud. These two
effects together result in a slight smoothing of the IR image, which slightly lessens
peaks and increases valleys in the IR data along a scan line. These effects are
important only when examining features which are of the same scale as the IFOV
of the instrument.
FIGURE 5.10. Infrared image produced by splicing images from DMSP and Meteosat 3. (Note: Missing
scan lines on DMSP image are included to aid in locating the polar orbiter data.)
5.5.1 Clouds
Clouds are the single most informative feature on the satellite image, and in
some cases the most difficult to analyze. The presence of clouds is often the only
clue we have to a significant meteorological process occurring in the atmosphere.
5.5 Atmospheric and Surface Phenomena 165
Both the character and the extent of the clouds are important, as well as the
specific location of the cloud mass. In some cases the absence of clouds is more
significant than the occurrence of clouds. We look at some of the characteristics
of clouds and examine how we can detect them. We also examine some of the
properties of clouds that occasionally make them difficult to detect.
Clouds indicate the presence of moisture and some type of cooling mechanism.
The most common cause of cooling is upward vertical motion (convection) that
causes adiabatic expansion. Less common, but important for cloud formation,
are radiative cooling, and the advection of warm, moist air over a cool surface
or cold air over a warm, moist surface. The presence of clouds indicates that one
of these mechanisms is, or has recently been, occurring. The type of cloud and
its height present further clues about the process that is responsible for it.
We examine the characteristics of clouds in standard height categories. The
commonly accepted layers are low, middle, and high clouds, with some clouds,
such as cumulonimbus, spanning all three layers. In midlatitudes low clouds are
typically found below 3 km, middle clouds between 3 and 7 km, and high clouds
between 7 and 12 km, with cumulonimbus occasionally extending beyond 20
km. The placement of clouds in the three layers is normally done using infrared
imagery to assign heights based on cloud top temperature, but some cloud types
are best observed in visible imagery. We also identify some clouds that require
the use of both the visible and infrared imagery for proper identification, and we
look at the use of other channels, such as the water vapor absorption bands.
infrared image because they are much colder than the surface (Fig. 5.6). Exceptions
are extremely cold wintertime outbreaks over high and midlatitudes, and moun-
tainous terrain where the surface reaches middle cloud altitudes. Middle clouds
can also be either stratiform (altostratus) or cumuliform (altocumulus) and are
used to infer processes that extend into the midtroposphere.
5.5.2 Storms
Storms are some of the most dramatic things captured in satellite data. In this
section we present images of several types of storms, but not by any means all
types of storms.
is quite evident because of the clear weather it has produced. Weldon (1983)
reports on the results of a large effort to identify the relationship between cloud
patterns and frontogenesis, cyclogenesis, and wind fields in general.
5.5.2.4 Lightning
Although not a storm, lightning is an indicator of thunderstorms. A special
low-light nighttime visible sensor on board the DMSP satellite can detect lightning,
even though its designers never envisioned that it would do so. The visible sensor
on the OLS instrument was designed to be sensitive enough to detect clouds
illuminated by moonlight. Because of its sensitivity, the sensor is momentarily
saturated when a bright lightning flash occurs as it scans across a thunderstorm.
This produces a bright streak along a scan line which lasts until the sensor falls
below saturation a few pixels later. Figure 5.12 shows a dramatic example of
lightning in an intense squall line over the Gulf of Mexico. In this example, city
lights show up over the land, but the bright streaks over the water are clearly
caused by lightning from the intense line of thunderstorms that appears as bright
(cold) clouds on the corresponding infrared image.
causes upward motion and condensation to the equatorward side of the jet and
subsiding air to the poleward side. The axis of the jet stream is at the cloud edge.
Jet streaks are small wind maxima that move through the large-scale circulation
patterns. Because they are associated with localized upward vertical motion, they
produce clouds that can be used to identify their location. Jet streaks often appear
as a thickening of the jet stream cirrus in the vicinity of the maximum winds.
Transverse banding is also associated with the jet stream and is an indication
of strong winds. Transverse bands form perpendicular to the axis of the jet
stream and have been shown to indicate winds in excess of 40 m s (80 kt) _ 1
(Brandli, 1976).
5.5.3.5 Turbulence
Turbulence cannot be detected directly from satellite imagery; however, certain
cloud patterns are unmistakably related to turbulence. The most common geo-
graphic location for turbulence-induced clouds is in mountainous terrain. The
impingement of strong wind flow into a high mountain barrier frequently results
in mountain wave turbulence and, if moisture conditions are right, in the formation
of mountain wave clouds (Fig. 5.13). Wave clouds are associated with moderate
to severe turbulence. Figure 5.14 shows a typical example of mountain wave
turbulence over the Appalachians. Wave clouds normally form in bands that are
parallel to the mountain barrier that induced them. The spacing between bands
MOUNTAIN
BARRIER
FIGURE 5.13. Mountain wave cloud formation.
5.5 Atmospheric and Surface Phenomena 171
FIGURE 5.14. Visible DMSP image showing "mountain waves" that have formed downwind from the
Appalachian Mountains. Also note the characteristic dendritic signature of the fog/stratus that has
formed in the valleys, and the cloud shadows to the north and west of the cloud edges.
is linearly related to the normal component of the wind at the barrier, though
different authors give different relationships (Brandli, 1976, p. 106; Weber and
Wilderotter, 1981, p. 15).
Another form of wave cloud is the billow cloud (Ludlam, 1967; Brandli, 1976),
which is similar in appearance to the wave cloud but is a high cloud and it is not
associated with the topographic barrier. In fact, wave clouds can form when
moist air is forced to flow over a cooler (more dense) air mass. The result is a
sinusoidal wave pattern propagating downstream and forming clouds in the upper
portion of the wave. Billow clouds, because they are high clouds, show up best
on an infrared image.
suppresses clouds over the cooler water, while rising air forms clouds over the
warmer land. Coastlines are therefore regularly outlined in clouds. Figure 5.15
shows an area along the east coast of Florida where sea breeze clouds have been
pushed inland by prevailing northeasterly winds. A more localized "lake breeze"
is also evident over and downwind of Lake Okeechobee and smaller lakes.
FIGURE 5.15. Visible DMSP image showing the effect of the "sea breeze" along the east coast of Florida
and "lake effect" over Lake Okeechobee.
5.5 Atmospheric and Surface Phenomena 173
FIGURE 5.17. GOES image showing smoke from the Yellowstone fires of 1988.
Snow and ice can sometimes be identified by the fact that they conform to
geographic features. Valleys, rivers, and lakes are often free of snow, whereas
surrounding land surfaces are covered, making it easier to distinguish between
clouds and snow. This is evident on Fig. 5.11 in the upper Ohio Valley. Also
note the ice that covers a portion of Lake Michigan.
A technique using daytime 3.7-μπι data has also shown success in discriminat-
ing low cloud from snow cover (Kidder and Wu, 1984). Figure 5.18 shows three
images of the same scene made by the AVHRR. In the visible and 11 -μηι infrared
images, the clouds in Nebraska and Kansas are nearly indistinguishable from the
snow to the east. In the 3.7-/xm image, however, there is dramatic contrast between
the clouds and the snow due to reflection of sunlight by the clouds but not by
the snow.
5.5.4.5 Sea-Surface Temperatures
Sea-surface temperatures, and more importantly temperature gradients, can be
analyzed from infrared imagery. The measurement of surface temperatures, of
FIGURE 5.18. Three N O A A A V H R R channels for discrimination of cloud from snow: (top) channel
1—visible, (middle) channel 3 - 3 . 7 /xm, and (bottom) channel 4 — 1 1 μτη. [Adapted from Kidder and
Wu (1984).]
5.5 Atmospheric and Surface Phenomena yfl
FIGURE 5.20. Visible "background" image that represents the cloud-free radiance for a portion of the
northwestern United States at 1 8 0 0 UTC during the month of July.
by L + L removes the effect of varying sun angle and viewing angle. NDVI is
x 2
high when green plants cover a pixel and low when no green plants are present.
In this chapter we have been able to present only a few, mostly black-and-
white, images made with data from weather satellites. A new and inexpensive
source of satellite imagery has become available in the last few years: images
offered on the Internet via file-transfer protcol (ftp) servers such as Gopher or by
hypertext-transfer protocol (http) servers such as Mosaic. We strongly encourage
readers who have access to the Internet to explore some of the wide variety of
still and animated satellite imagery which is available.
Bibliography
Adler, R. F., and M. J. Markus (1982). Determination of thunderstorm heights and intensities from
GOES infrared data. Preprints: Ninth Conf. on Weather Forecasting and Analysis. American
Meteorological Society, Boston, pp. 2 5 0 - 2 5 7 .
Anderson, R. Κ., V. J. Oliver, E. W. Ferguson, F. E. Bittner, A. H. Smith, J. F. W. Purdom, G. R.
Farr, and J. P. Ashman (1974). Applications of Meteorological Satellite Data in Analysis and
Forecasting, ESS A Tech. Report NESC 5 1 , U.S. Dept. Commerce, Washington, D C .
Bader, M. J., G. S. Forbes, J. R. Grant, R. Β. E. Lilley, and A. J. Waters (eds.) (1995). Images in
Weather Forecasting: A Practical Guide for Interpreting Satellite and Radar Imagery. Cambridge
University Press, N e w York.
Brandli, H. W. (1976). Satellite Meteorology, AWS-TR-76-264, Hq. Air Weather Service, Scott AFB,
IL, 188 pp.
Browning, K. A. (ed.) (1982). Nowcasting, Academic Press, San Diego.
Carlson, D. L. (1982). Introduction to Weather Satellite Imagery. N O A A Tech. M e m o . N W S
SR-103, N O A A , U.S. Dept. Commerce, 4 7 pp.
Dvorak, V. F. (1984). Tropical Cyclone Intensity Analysis Using Satellite Data, N O A A Tech. Rep.
NESDIS 1 1 , N O A A , U.S. Dept. Commerce.
Fett, R. W., and W. F. Mitchell (1977). Navy Tactical Applications Guide. Volume I Techniques and
Applications of Image Analysis. Naval Environmental Prediction Research Facility, Monterey, CA.
Fett, R. W., P. E. La Violette, Μ. Nestor, J. W. Nickerson, and K. Rabe (1979). Navy Tactical
Applications Guide. Volume II Environmental Phenomena and Effects. Naval Environmental
Prediction Research Facility, Monterey, CA.
Hubert, L. F. (1974). A Computation of the Time Constant of SMS IR Sensing System. NESS Internal
M e m o . , N O A A / N E S S , Washington, D C .
Kelly, F. P., T. H. Vonder Haar, and P. W. Mielke, Jr. (1989). Imagery randomized block analysis
(IRBA) applied to the verification of cloud edge detectors, / . Atmos. Oceanic Tech., 6, 6 7 1 - 6 7 9 .
Kidder, S. Q., and H.-T. W u (1984). Dramatic contrast between low clouds and snow cover in daytime
3.7 μπι imagery. Mow. Wea. Rev., 1 1 2 , 2 3 4 5 - 2 3 4 6 .
Kidder, S. Q., and H.-T. Wu (1987). A multi-spectral study of the St. Louis area under snow-covered
conditions using N O A A — 7 A V H R R data. Remote Sens. Environ., 2 2 , 1 5 8 - 1 7 2 .
5.6 A Final Note 181
Ludlam, R. H. (1967). Characteristics of billow clouds and their relation to clear air turbulence.
Quart. J. Royal Met. Soc, 9 3 , 4 1 9 - 4 3 5 .
Morel, P., M. Desbois and G. Szewach (1978). A new insight into the troposphere with the water
vapor channel of Meteosat. Bull. Am. Meteor. Soc, 5 9 , 7 1 1 - 7 1 4 .
Parke, P. S. (ed.) (1986). Satellite Imagery Interpretation for Forecasters. Weather Service Forecasting
Handbook N o . 6, N O A A , Washington, D C .
Parmenter-Holt, F. C. (1982). Reliability of enhanced infrared (EIR) geostationary satellite data at
high latitudes, Mon. Wea. Rev., 110, 1 5 1 9 - 1 5 2 3 .
Rao, Κ. P , S. J. Holmes, R. Κ Anderson, J. S. Winston, and P. E. Lehor (eds.) (1990). Weather
Satellites: Systems, Data, And Environmental Applications. American Meteorological Society,
Boston, 5 0 3 pp.
Spencer, R. W. (1986). A satellite passive 3 7 - G H z scattering-based method for measuring oceanic
rain rates. / . Climate Appl. Meteor., 2 5 , 7 5 4 - 7 6 6 .
Spencer, R. W , Η. M. Goodman, and R. E. H o o d (1988). Precipitation retrieval over land and ocean
with the SSM/I, Part 1: identification and characteristics of the scattering signal. / . Atmos. and
Oceanic Tech., 6, 2 5 4 - 2 7 3 .
Spencer, R. W , W. S. Olson, W. Rongzhang, D. W. Martin, J. A. Weinman, and D. A. Santek (1983)
Heavy thunderstorms observed over land by the N I M B U S — 7 Scanning Multichannel Microwave
Radiometer. / . Climate Appl. Meteor., 22, 1 0 4 1 - 1 0 4 6 .
Vonder Haar, Τ. Η, D. L. Reinke, and S. Naqvi (1989). G E O R G E — 3 D : A minicomputer-based
system for 4-dimensional analysis and display of combined satellite, radar, terrain and map data.
Proceedings: Fifth Conference on Interactive Information and Processing Systems, American Mete-
orological Society, Boston, pp. 2 2 - 2 5 .
Warren, R. (1977) Displacement Error of Satellite Cloud Tops, N W S Technical Attach. N o . 7 7 - G 4 ,
SSD-CRH, 3 pp.
Weber, Ε. M., and S. Wilderotter (1981). Satellite Interpretation. Air Weather Service, 3rd Wea.
Wing, Technical N o t e - 8 1 / 0 0 1 , Offutt AFB, N E , 95 pp.
Weiss, C. E. (1978). Cloud-Location Corrections Near the Horizon of an SMS Image. Satellite
Applications Information N o t e 7 8 / 8 , N O A A / N E S D I S , U.S. Dept. Commerce, Washington, D C ,
8 pp.
Weldon, R. (1983). Synoptic Scale Cloud Systems. NESDIS Training Notes, Satellite Applications
Laboratory, N O A A , U.S. Dept. Commerce, Washington, D C , 35 pp.
Weldon, R , and S. Holmes (1984). Characteristics of Water Vapor Imagery. NESDIS Training Notes,
Satellite Applications Laboratory, N O A A , U.S. Dept. Commerce, Washington, D C , 2 pp.
6
Temperature
and Trace
Gases
There are two basic types of passive atmospheric sounding: vertical sounding,
in which the sounding instrument senses radiation coming from the atmosphere
and the Earth's surface; and limb sounding, in which only the limb of the atmo-
sphere is sensed. Both sounding methods utilize Schwarzchild's equation. Active
atmospheric sounding includes the use of lidar and radar. In this chapter we focus
183
"Jg^ Chapter 6 Temperature and Trace Gases
τ (δ ,δ )- χρ(-|δ -δ |),
λ 1 2 ε (6.2)
2 1
LotJ'" + f
T
Lk = B {T)W (h,p)dh,
x x (6.3a)
W X ( M = ^ ( Ä (6.3b)
where L is the radiance emerging from the Earth's surface, τ is the vertical
G 0
transmittance from the surface to the satellite, and r is the slant-path transmit- k
ß
tance from level h to the satellite. W {h^) is called a weighting function because
k
infrared sounder will see only down to the cloud top, in which case L and τ Q 0
should be replaced with the cloud top values L and r , respectively, and the c c
(and thus Τ itself) as a function of h. This idea was first put forward by Kaplan
(1959). Figure 6.1 shows r and W for six channels in the 15-μπι band of C 0
k k 2
TRANSMITTANCE (-dr/d(p 2 / 7
))
FIGURE 6.1. Transmittances to the satellite and weighting functions from the Vertical Temperature
Profile Radiometer on the N O A A 2 through N O A A 5 satellites. In this figure the vertical coordinate
is h = -p .
117
N o t e that the weighting functions peak near the level where transmittance is approximately
one. [Courtesy of Donald W. Hillger, N O A A / N E S D I S . ]
ture of the ground itself. Water, however, reflects as much as 60% of the micro-
1
wave radiation incident on it. In this case, the surface radiance, L , is composed Q
of two terms, the radiation emitted by the surface and the reflected sky radiance:
L = e B (T ) + ( l - e J L
0 0 x 0 s k y , (6.4)
where the water surface has been approximated as a specular reflector with
reflectance 1 - ε . The sky brightness may be determined by applying Schwarz-
2
0
child's equation:
rhj d \ (τ \1/μ
dh, (6.5)
from the Big Bang. Combining Eqs. 6.3 and 6.5 shows that the radiance measured
from a satellite is
L = ε Β ( Τ ) τ ^ + (1 - e )B (T )r^
K 0 λ 0 0 K sp3ce + fj B (T)W (h^)dh,
K x (6.6)
1
This is not the "surface" temperature as measured by a thermometer in a standard meteorologi-
cal shelter.
2
Over water, microwave radiation is polarized, and ε is a strong function of polarization (see 0
Fig. 7.9). The following equations apply, but the appropriate value of ε , depending on the polarization 0
where
one. The differences between these equations and Eqs. 6.3 arises from the fact
that reflection affords the satellite the opportunity to see the atmosphere twice,
once before reflection, and once after reflection. Figure 6.2 shows W for the x
WEIGHTING F U N C T I O N (km"')
FIGURE 6.2. Weighting functions for two scan angles for the Nimbus 6 Scanning Microwave Spectrome-
ter. [After Staelin et al (1975).]
6.1 Sounding Theory 187
(6.8)
As μ increases, the peak of the weighting function moves higher in the atmo-
sphere. For wavelengths in which the weighting function peaks in the troposphere
(which cools with height), radiance decreases as the instrument scans away from
nadir. This is called limb darkening. The complementary effect, limb brightening,
occurs when the weighting function peaks in a region of the atmosphere in which
temperature increases with height. In the typical retrieval process, an empirical
correction for limb darkening is applied to raw sounding data before retrieval;
all of the scan spots are then treated as if they were at nadir (see Section 6.3.1.1).
The original idea for sounding the atmosphere from satellites (King, 1958) em-
ployed limb darkening. King suggested that observations at a single wavelength,
but at several zenith angles, could be used to retrieve atmospheric temperature pro-
files. 3
ratio is known and independent of height. Wavelengths that are sensitive to these
gases are used to retrieve atmospheric temperature. For wavelengths sensitive to
variable trace gases, such as 0 and water vapor, we assume that the temperature
3
profile is known and use the radiances to retrieve mixing ratios (or concentrations).
Unfortunately, no wavelengths are completely free of absorption by the variable
trace gases; this contributes to errors in retrieving temperatures. Also, errors in
our knowledge of temperatures result in errors in trace-gas retrievals (see below).
The above equations are strictly applicable only at a single wavelength, but
satellite instruments sense a band of wavelengths. The equations must be integrated
over wavelength using the spectral response function of each channel as a weight.
As discussed in Section 3.4, however, spectral integration is very time-consuming.
In the microwave portion of the spectrum, where there are relatively few absorp-
tion lines, an equivalent wavelength is often employed. The equivalent wavelength
is chosen near the center frequency of a satellite channel such that the spectrally
integrated radiance sensed by the satellite is approximately the same as the radiance
at the equivalent wavelength. When a transmittance must be calculated, a line-
by-line calculation is done, but only at one wavelength, the equivalent wavelength.
In the infrared region, line-by-line calculations at even a single equivalent
wavelength are too time-consuming. A recursive method is frequently employed
3
Fleming (1980, 1 9 8 2 , 1985) suggested combining wavelength and zenith angle scanning. Sound-
ings would be retrieved by the method used to invert CAT (computer-assisted tomography; X-ray)
scans.
188 Chapter 6 Temperature and Trace Gases
(Smith, 1969; McMillin and Fleming, 1976; Fleming and McMillin, 1977; McMil-
lin et ai, 1979). The atmosphere is divided into predetermined layers, and the
transmittance of each layer, averaged over the response function of the radiometer
channel, is calculated line-by-line. In each layer, polynomial functions of tempera-
ture, humidity, and (sometimes) zenith angle are then fitted to the results. Given
estimates of temperature and humidity (and zenith angle), the polynomial coeffi-
cients can be used to quickly calculate transmittance of any layer. The transmit-
tance from a particular level to the satellite is then simply the product of the
transmittances of the layers above. This method is fast, easy, and accurate (consid-
ering the experimental uncertainties in measurements of line parameters), but it
does not let the user change the layers used in the calculations.
άτ {χ,ζ )
•dx.
λ Ύ
L (z )='_ (6.9)
k T
dx
This can be converted to a vertical integral:
L (z )=
k T Γ B (T)W (z,z )dz,
x k T
(6.10)
where
dT (x,z )\ (dr {x,z ) dx
(6.11)
k k
W (z,z )
T T
dx dx Tz
k T
and the + and - subscripts indicate values along the positive and negative χ
axes, respectively.
Limb sounding has several advantages over vertical sounding (Gille and
House, 1971):
• The weighting functions peak very sharply at the tangent height because
the instrument senses nothing below the tangent height and atmospheric
density decreases exponentially above the tangent height (Fig. 6.3).
• The background (space) is cold and uniform in contrast to vertical sound-
ing in which the background (the surface) is hot and variable.
• There is up to 60 times more emitting material along a horizontal path
than along a vertical path. This means that temperatures can be measured
to higher altitudes than with vertical sounders, and low-concentration
gases can be better detected.
6.2 Retrieval Methods 189
Π t~i ι ι ι 1 ι ι ι ι I ι ι ι ι 1 ι ι ι ι I ι ι ι ι I ι ι ι ι I ι ι ι
0 0.1 0.2 0 3
WEIGHTING FUNCTION (km*) 1
FIGURE 6.3. Weighting functions for the Nimbus 6 L R I R wide 15-μπι channel used for temperature
sounding. [After Gille et al (1980).]
The physical problem we would like to solve is this: What temperature and
trace gas concentration profiles could have produced a set of observed radiances?
This is called the inverse problem or retrieval problem. The opposite problem,
called the forward problem, is to calculate outgoing radiances given temperature
and trace-gas profiles.
Clouds cover approximately 50% of the Earth; thus about 50% of soundings
are contaminated with clouds. We defer until Chapter 8 a discussion of methods
to detect and correct for clouds in infrared sounding data. In the remainder of
this chapter we assume that the radiances are "clear-column radiances" which
have been corrected for clouds.
Given clear-column radiances, the forward problem is easy to solve, but the
retrieval problem is difficult because the solution is not unique. Even if a noise-
free radiometer that measured radiances at all wavelengths could be constructed,
a unique solution to the radiative transfer equation would not be guaranteed
(Chahine, 1970). When a finite number of wavelengths are observed and the
measurements are contaminated with noise, an infinite number of solutions are
possible. The retrieval problem becomes one of finding temperature profiles that
190 Chapter 6 Temperature and Trace Gases
satisfy the radiative transfer equation and approximate the true profiles as closely
as possible. Approaches to the retrieval problem can be classified into three general
areas: physical retrievals, statistical retrievals, and hybrid retrievals.
4
at the jth level and B (Tj ) be the resultant Planck radiance at level ; at the
;
w)
(6.12)
The iterated temperature at level ; is found using the inverse Planck function.
This scheme works because if, for example, the calculated radiance in channel;
is greater than the observed radiance, it is reasonable to adjust downward the
Planck radiance (and thus the temperature) at the level where the weighting
function for channel / peaks. Since the peak of the weighting function is the
strongest contributor to the radiance, using the ratio of the observed to the
calculated radiance to adjust the Planck radiance is also reasonable.
Smith's (1970) scheme for adjusting temperature profiles is similar to that of
Chahine, but he relaxes the requirement that the temperature be retrieved at only
This statement is an oversimplification. The reader is referred to Fritz et al. (1972), Deepak
4
(1977, 1980), Houghton et al. (1984), Deepak et al. (1986), and the current literature for information
on the many retrieval techniques.
6.2 Retrieval Methods ]^"|
estimate of, and the observed value of, the radiance in channel;. At each level,
Smith obtains / estimates of an iterated Planck radiance:
B (Tjr )
i
1)
= Bm i + [L -L^i (6.13)
where the / estimates of T ^ , obtained by the inverse Planck function, are
+ 1 )
Σ W^TJT Υ
j ^ D = /^L_ _ ( 6 > 1 4 )
Σ ν».
7=1
Since Smith's levels are not restricted to be at the peaks of the weighting functions,
he cannot iterate the temperature at a level by using a single channel. At each
level, therefore, he obtains a suggested temperature change from each channel
and lets the weighting function discriminate among them.
Smith's scheme is more flexible than that of Chahine in that it allows the user
to choose the levels at which he will retrieve temperatures (consistent, of course,
with the predetermined levels at which the transmittances are calculated). How-
ever, from / channels of information, one can calculate independent temperatures
at / levels at most. If Κ is greater than / (as is usually the case) the extra levels
are not independent.
To retrieve moisture profiles, it is assumed that the temperature profile, and
6
thus Β (Τ), is known. Smith (1970) starts with a first-guess mixing ratio profile
λ
q( p) and proceeds in a manner which is formally the same as that for temperature
retrievals. The iteration formula is
L - Ü~ n)
1 + - γ{η)
W - »> (6-15)
where is a sensitivity factor which estimates the mixing ratio change necessary
to correct for a given radiance imbalance. In equation form,
U {p)=
{n)
f <?(/>A (6.16b)
5
T w o m e y et al. (1977) introduced a modification of Chahine's (1970) scheme that also relaxes
this requirement (see Section 8.2.1).
6
An important scheme to retrieve integrated water vapor is discussed in Section 6.6.2.
]92 Chapter 6 Temperature and Trace Gases
where U {p) is the nth estimate of the integrated water vapor above pressure
{n)
level p. For each level, Smith obtains / estimates of the iterated mixing ratio. As
above, q^ is found as a weighted average of the
+1)
using the weighting
functions as weights.
7
The logic of this scheme can be demonstrated as follows. Note first that since
dTj/dU is always less than zero, the sign of Γ| is opposite that of dT/dp. Consider
Μ)
a channel whose weighting function peaks in the troposphere. Suppose that the
calculated radiance in this channel is too high compared with the observed radi-
ance. Since Tj is a negative number (because dT/dp > 0), q^ will be multiplied
n) ]
by a number greater than one, and q^ will be greater than q^K Increasing the
+1)
moisture will move up in the atmosphere the point where the optical depth equals
one, and thus the peak of the weighting function for this channel will move up.
Since tropospheric temperature decreases with height, Lj will be less than
n+1)
Obviously the first-guess profile is important. The closer the first-guess profile
is to the actual profile, the better the solution is likely to be. In the past, the
first guess has been climatological. A better choice is a nearby radiosonde, but
radiosonde releases do not often coincide with satellite overpasses, and few radio-
sondes are released over the ocean. Forecast soundings generated by numerical
models are also being used. Recently, a great deal of work has been done on
using the radiances themselves to classify the sounding into a set that has an
appropriate first-guess profile and (for the methods listed below) a retrieval matrix
(Uddstrom and Wark, 1985; Thompson et al, 1985; McMillin, 1986).
Also obvious is that whatever auxiliary information can be brought to bear
on the subject will be helpful. The skin temperature, for example, is an important
factor, particularly in moisture retrievals. Sounders usually have window channels
which are used in part to measure the skin temperature. Some analysis schemes
attempt to incorporate surface observations of air temperature and dewpoint to
improve retrievals in the lowest atmospheric layer, in which some of the largest
retrieval errors occur. More experimental are attempts to utilize radiosonde-
estimated tropopause heights in retrievals (McMillin, 1985, 1986).
None of the retrieval schemes that we discuss is clearly superior to the others.
Users who wish to implement a retrieval scheme must weigh the advantages and
disadvantages of each. Based on our experience, the principal advantages of the
physical retrieval method are
• Physical processes are clearly evident at each stage of the retrieval. 8
This has the important implication that changes for such things as elevated terrain or for a failed
8
• Except for information contained in the first guess, it does not utilize
known statistical properties of the atmosphere.
t - <> = c(i -
t α». (6.17)
Note that once C, (t), and (1) are known, retrieval of temperatures from observed
radiances is a very simple task which involves only vector subtraction, matrix
multiplication, and vector addition. This simplicity makes a statistical method
attractive for operational retrieval schemes in which numerous soundings must
be processed.
There are several ways to find C, (t), and (1). The simplest is the regression
solution. Let (t) and (1) be the average of all the vectors t and 1, respectively, in
the training data set. Let T' be the Κ x Ν matrix whose columns are the vectors
t - (t), and let L' be the / x Ν matrix whose columns are the vectors 1 - (1).
The matrix C which, in a least-squares sense, minimizes errors in Eq. 6.17 is
C = T'L' (L'L' )
T T
(6.18)
where the superscript Τ indicates that the transpose of the matrix is to be used,
and the superscript - 1 indicates the matrix inverse. This scheme has been used
extensively to retrieve satellite soundings.
One problem with the regression solution is that no "filtering" of noise from
the input temperatures or radiances is done. As a result, the C matrix can be
unstable; that is, small radiance errors can produce unacceptably large errors in
the retrieved temperatures. Smith and Woolf (1976) developed a technique to
filter noise using statistical eigenvectors. Their scheme was used operationally for
many years. More recently, Thompson (1992) employed singular value decompo-
sition to improve statistical retrievals.
194 Chapter 6 Temperature and Trace Gases
Hybrid retrieval methods are in between physical and statistical retrieval meth-
ods. They appear much like purely statistical methods, but they do not require
the large training data set. They use weighting functions like physical retrievals,
but they do not directly involve integration of the radiative transfer equation.
Hybrid methods, better known as inverse matrix methods, are reviewed by Fritz
etal (1972).
To start, the radiative transfer equation is linearized about a standard tempera-
ture profile T (h):
s
L =L
k Ks + ε ^Τ τ
0 $0 0 + W {h) j±
K T' (h)dh,
s (6.19)
where L is the radiance if T (h) were the actual profile, and T' (h) is the deviation
Xs s s
contains the errors of measurement of the radiances. The matrix A contains the
weighting functions, the quadrature weights, and the Planck sensitivity factors
BBJdT. A is calculated with a knowledge of the transmittances as in the case of
physical retrievals.
Suppose, temporarily, that we have a set of coincident radiosonde and satellite
observations. Let the uppercase letters indicate matrices which have Ν columns
for the Ν sounding pairs in the data set:
L' = A T ' + E . (6.21)
We seek a matrix C such that
TT = C L ' . (6.22)
Substituting Eq. 6.21 in Eq. 6.18, we have
C = T'(AT + E) [(AT' + E)(AT
T
+ E) ] T _ 1
. (6.23)
Expanding and using the matrix relationship (AB) T
= B A ,
T T
C = T'(T' A T T
+ E )[AT'T' A
T T T
+ AT'E T
+ ET' A T T
+ EE ]" . 7 1
(6.24)
196 Chapter 6 Temperature and Trace Gases
Now, we assume that the measurement errors are uncorrelated with temperature
deviations. Thus Τ Έ and E T ' are negligible in comparison with the other terms.
Τ T
We have, therefore,
C = S A [AS A + S ] - \
T
T
X
T
E (6.25)
where S is the temperature covariance matrix ( T ' T ' / N ) and S is the radiance
T
T
E
error covariance matrix (EE /N). This equation, called the minimum variance
T
method, has been discussed by many people, including Strand and Westwater
(1968) and Rodgers (1970, 1976). The interesting property of Eq. 6.25 is that
all of the components can be determined separately: A can be calculated from
the transmittances, S can be determined from a sample of radiosonde soundings,
x
S = σ\\,
T (6.26)
S = σΐΐ.
E (6.27)
The main advantage of the hybrid methods is that they are easier to put into
operation than the statistical or physical methods. They share with the physical
retrieval method the disadvantage of depending on a knowledge of the transmit-
tances. Most of them share with the statistical retrieval method the advantage of
including statistical knowledge of atmospheric structure.
6.3 Operational Retrievals 197
The retrieval method is only a small part of the process by which radiances
from operational satellites are converted into temperature and moisture soundings.
In this section, we discuss the processes that are used operationally to produce
soundings from NOAA and GOES satellites.
(6.29)
• An attempt is made to correct the window channels for water vapor ab-
sorption.
• For daytime data, an albedo, estimated using the 3.7-μηι channel, is used
to correct the 4.3-^m channels for reflected sunlight.
• Since the MSU has coarser spatial resolution than does the HIRS/2, the
MSU data are interpolated to the locations of the HIRS/2 scan spots.
FIGURE 6.4. Weighting functions for the TIROS Operational Vertical Sounder (TOVS). [After Smith
et al. (1979a).]
6.3 Operational Retrievals 199
Finally, the Preprocessor obtains solar zenith angles, terrain elevations, and initial
guess values of skin temperature and surface albedo. All these data are staged to
disk to await further processing.
sounding is retrieved using only MSU, SSU, and stratospheric HIRS/2 channels.
These soundings are called "cloudy" or "third path" soundings.
In 1986, a higher resolution scheme was implemented (Reale et al., 1986). The
9 x 7 box of HIRS/2 scan spots was divided into six miniboxes, each 3 x 3 scan
spots. A clear-column radiance vector is produced for each minibox. In order not
to produce six times as many final soundings, each minibox is checked to determine
quality and usefulness of the data. The quality flag may be (1) good, (2) redundant,
(3) questionable, or (4) bad. "Redundant" means that the minibox has essentially
the same radiance vector as its neighbors. In later processing, redundant, question-
able, and bad miniboxes can be discarded to produce the appropriate number
of soundings.
First, the raw VAS data are converted to brightness (equivalent blackbody)
temperatures and Earth located. As with the TOVS retrieval system, scan spots
are blocked before processing. The VAS system is designed to be flexible, so the
number of spots in a block is variable. Typically, however, a block of 11 x 11 scans
spots is used, which results in a field of retrieved soundings with approximately 75-
km resolution.
Next, the data are filtered for clouds. Since visible data are not collected during
dwell sounding, the cloud filter relies on three window infrared channels: 3.9,
11.2, and 12.7 μηι (channels 12, 7, and 8):
• If the 3.9-μιη brightness temperature is significantly different (3-5 K)
from the 11.2-μιη brightness temperature, the scan spot is assumed to be
cloudy.
• If the 11.2-μπι brightness temperature (corrected for moisture effects by
the "split-window" technique, see below) is significantly colder than the
surface air temperature (obtained from analyzed surface observations), the
scan spot is classified as cloudy.
If a majority of the scan spots appear to be clear, the clear radiances are averaged,
and a sounding is retrieved down to the surface. (During the retrieval process
itself, the sounding can be classified as cloudy if the algorithm cools the first-
guess surface temperature by more than 6 K.) If most of the scan spots are cloudy,
the cloudy spots are averaged, and cloud-top pressure is estimated by comparing
the 11.2-μιτι brightness temperature (corrected for moisture effects) with the first-
guess temperature profile. A sounding is retrieved down to the cloud top, and
values below cloud top are estimated by interpolating between the cloud-top value
and the surface analysis. These cloudy soundings, however, are of questionable ac-
curacy.
The retrieval algorithm is a hybrid scheme described by Smith et al. (1986).
First-guess temperature and moisture profiles are selected, usually from a model
forecast such as the Nested Grid Model (NGM) or the Global Spectral Model.
Surface temperature and humidity are obtained from analyzed fields of the corre-
sponding hourly observations. The first-guess estimate of the skin temperature is
obtained by regression from the 11.2, 12.7, and (for noncloudy scan spots) 3.9-
μιη brightness temperatures. Temperatures above the 100-hPa level are obtained
from the latest analyses or from climatology. Using the first-guess profiles, the
weighting functions and the brightness temperatures in the VAS channels are
calculated. The temperature and moisture profiles are calculated by making a
single correction to the first-guess profiles, which is why the scheme is known as
the "simultaneous" or "one-step" method.
with which to compare. Radiosonde soundings are the most obvious candidates,
but two problems occur:
• Radiosonde soundings are not error-free; an rms (root mean square) error
of 1 Κ is often quoted.
• Satellite soundings are fundamentally different from radiosonde
soundings.
This latter point is illustrated by example. Figure 6.5 shows two VAS soundings
compared with radiosonde soundings. Clearly, the radiosonde soundings exhibit
more vertical structure than the satellite soundings. The satellite sounding has
trouble where the lapse rate changes abruptly with height, such as at surface
inversions or near the tropopause. This is particularly true of the moisture sound-
ing. An examination of the weighting functions (Fig. 6.6) reveals the cause of this
difficulty. Current satellite sounders integrate deep layers of the atmosphere.
Vertical smoothing of the sounding is unavoidable. Interferometric sounders cur-
rently under development (see Section 4.3.1) offer improved vertical resolution
100
Λ 2 0 0
Ε
300
III
Ε 4 0 0
g> 5 0 0
ω 6 0 0
ο. 7 0 0
Θ00
900
1000
-70 - 5 0 - 3 0 -10 0 10 30
TEMPERATURE e
C
TEMPERATURE *C
FIGURE 6.5. T w o VAS soundings compared with radiosonde soundings. [After Smith et al. (1981).]
6.3 Operational Retrievals
203
250 250
300 300
SUMMER 19 7 9 -
FALL 19 7 9 -
WINTER 19 8 0 -
SPRING 19 8 0 ·
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
200
250
300
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 1.0 1.5 2.0 2.5 3.0 3.5 4.0
200
250
300
SUMMER
FALL
WINTER
SPRING •
FIGURE 6.7. The rms differences in layer-mean virtual temperature between satellite and radiosonde
soundings. [From data in Gruber and Watkins (1982).] Layer-mean values are plotted at the midpoint
of the layer. The upper left corner of each plot indicates the latitude band, and the upper right corner
indicates the retrieval type (clear or cloudy). Shown for comparison are climatological values of
temperature standard deviation; seasonal values are shown in the midlatitudes, and annual values
are shown in the tropics.
6.3 Operational Retrievals
205
SUMMER 1 979 -
FALL 1 979 -
WINTER 1 980 -
SPRING 1 980 -
-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0
BIAS (K)
1 50
200
250
t"
300
-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0
FIGURE 6.8. Biases (satellite minus radiosonde) in layer-mean virtual temperatures. [From data in
Gruber and Watkins (1982).]
Satellite soundings are seldom used by themselves; they are used in addition
to radiosonde soundings, surface observations, and perhaps other data. How does
one blend data from these various sources to yield an accurate picture of the
atmosphere? This is known as the data assimilation problem. The previous section
and the work discussed below suggest the following guidelines. First, it is clear
that biases in satellite soundings (and in conventional data) should be removed
as part of the assimilation process. Second, the quality or error characteristics of
the data should be taken into account during analysis. This can be accomplished
by weighting each datum inversely in proportion to its error variance. A third
guideline arises from a consideration of atmospheric dynamics. The mass and
flow fields of the atmosphere are usually closely in balance. If the mass field is
modified by the insertion of a satellite sounding, the flow field probably should
be modified to maintain balance. Hay den (1973) suggested that this can be approx-
imated by modifying the wind field by the difference of the geostrophic wind
before and after insertion of a temperature from a satellite sounding. Finally,
most satellite soundings are not made at synoptic times (0000 and 1200 UTC)
when radiosondes are launched and when most analyses are done. Corrections
for the asynoptic nature of satellite soundings probably should be attempted.
Much of the work in data assimilation has been done in connection with
model impact studies. One hopes, of course, that satellite soundings will improve
numerical weather forecasts. Another reason for these studies, however, is that
comparing numerical forecasts made with (SAT) and without (NOSAT) satellite
data to observations is one way to determine whether inclusion of satellite data
has improved the initial analysis. Presumably a better initial analysis will result
in a better forecast. Ohring (1979) reviewed early satellite data assimilation experi-
ments. He concluded that satellite soundings had a small positive impact in
the data-rich Northern Hemisphere and large positive impact in the data-sparse
Southern Hemisphere.
Beginning in the late 1970s there were several model impact studies. Ghil et
al. (1979) used a global, primitive-equation, general circulation model (Somerville
et al., 1974) at NASA's Goddard Laboratory for Atmospheric Sciences (GLAS).
The GLAS model had a 4° latitude x 5° longitude horizontal grid and nine sigma
levels in the vertical. The purpose of their experiments was to test data assimilation
schemes. They used time-continuous assimilation. In most numerical models, all
data within a window (several hours either side of the time when the model is
to be initialized) are analyzed together as if all observations were made at the
same time. In time-continuous assimilation, the model is initialized at a time, say
12 h, before the desired initial analysis. The model is run for the intervening
period, and new observations are assimilated at the times when they are made.
6.3 Operational Retrievals
This process, also called dynamic initialization, alleviates the problem of satellite
observations not being made at synoptic times. Ghil etal. assimilated observations
at each model time step (10 min).
At each time step, one is faced with the problem of how to use observations
to change the current model fields. Ghil et al. tested three methods. In the direct
insertion method, they essentially replaced the model temperature with the average
of the temperature observations within one model grid length. In the successive
correction method, they did a Cressman-type (1959) analysis of the difference
between the observations and temperatures interpolated from model grid points.
The weights used depended only on the distance between the observation and a
grid point. Three passes with decreasing radii of influence were used. The statistical
assimilation method was like the successive correction method, except that the
weights were based on the error characteristics of the observations. This is similar
to the optimum interpolation method of data analysis (Gandin, 1963; Schlatter,
1975). Regardless of the data insertion method used, the winds were geostrophi-
cally modified (Hayden, 1973).
Ghil et al. used NOAA 4 and Nimbus 6 soundings during the period January
through March 1976 from the Data System Test 6 [DST—6; Desmarais et al.
(1978)]. Forecasts of 48 and 72 h were compared with analyses using S^ skill
scores (Teweles and Wobus, 1954) and rms differences. Surface pressure and 500-
hPa height fields over North America and Europe were used for verification.
Comparing SAT and NOSAT forecasts, Ghil et al. found small but statistically
significant positive impact of satellite soundings. This result was highly dependent
on the data insertion method, however. The direct insertion method resulted in
negligible impact, the successive corrections method resulted in appreciable results,
and the statistical assimilation method gave very good results. They concluded
that when the better data insertion methods were employed, satellite soundings
extended the usefulness of forecasts 8-16 h; thus a 60-h SAT forecast was about
as accurate on the average as a 48-h NOSAT forecast. They also found that
impact increased with quantity of satellite data; when both NOAA 4 and Nimbus
6 soundings were used, the forecasts were better than when data from only one
satellite were used.
Tracton et al. (1980) did an experiment that was similar to that of Ghil et al.
(1979); however, their purpose was quite different. Rather than testing different
ways to assimilate satellite data, Tracton et al. wanted to test the impact that
satellite soundings would have on an existing, operational, numerical weather
prediction model. They used the six-layer primitive equation (PE) model (Shuman
and Hovermale, 1968), which was then operational at the U.S. National Meteoro-
logical Center (NMC). The model was hemispheric and had a resolution of 381
km at 60°N. Their data assimilation scheme was that of the operational system:
data within 3 h of analysis time were analyzed together; satellite data, in the form
of temperatures at mandatory pressure levels, were treated exactly like radiosonde
data. Tracton et al. used data from DST-6 (as did Ghil et al.) and from DST-5.
They verified forecasts over North America and Europe using skill scores and
rms differences as did Ghil et al. They found that the differences between SAT
208 Chapter 6 Temperature and Trace Gases
and NOSAT forecasts were "generally small and of inconsistent sign, i.e., beneficial
in some cases and harmful in others." On the average, in the DST-6 experiments
inclusion of satellite soundings resulted in slightly improved forecasts, but in the
DST-5 experiments slightly poorer forecasts resulted.
A possible conclusion from comparing the experiments of Ghil et al. (1979)
and Tracton et al. (1980) is that if Tracton et al. had assimilated satellite soundings
in a different manner, perhaps using some of the techniques of Ghil et al., they
would have had more positive results. Unfortunately the two experiments are not
sufficient to draw this conclusion. The difficulty is that the impact which satellite
soundings have on numerical forecasts is dependent on the model. Tracton et al.
(1981) found a negative correlation between model skill and satellite data impact.
Poorer models seemed to make better use of satellite soundings. On the other
hand, Atlas et al. (1982) increased the resolution of the GLAS model, which
improved its NOSAT performance, and found increased positive impact of satellite
soundings. It seems that the only way to assess the impact of satellite soundings
on numerical forecasts made with a specific model is to do a test using that model.
This is all the more important considering the fact that both today's models and
today's satellite soundings are more sophisticated than those used in the Data
System Tests. It is interesting that the current retrieval scheme for the TOVS
soundings (the MVS method) was implemented only after two data impact studies
at NMC (Dey et al., 1989).
Thomaseil et al. (1986) tested the impact of TIROS Ν and NOAA 6 soundings
on the forecast model used by the Israel Meteorological Service. Two of their
results are interesting. First, satellite soundings were most beneficial when they
were used in a two-step analysis procedure as follows. Satellite soundings and
surface observations, but no radiosonde data, were analyzed to construct the first-
guess fields. Radiosonde soundings were then used to update the first-guess fields
to produce the initial analysis for the model. This procedure neatly separates
satellite and radiosonde data, and lets each have maximal impact where they will
do the most good: satellite data over the oceans, where there are too few radio-
sondes to produce a good analysis, and radiosonde data over the land, where
additional observations from satellites are not needed. Second, Thomaseil et al.
9
attempted to verify the forecasts of the Israeli model over the ocean as well as
over land. They found that satellite soundings improved forecasts of sea level
pressure over the ocean to the point that they were almost as good as forecasts
over land. (Forecasts over land showed little impact of satellite soundings.) Readers
are warned, however, that both the model and the analysis system investigated
by Thomasell et al. are less sophisticated than those used elsewhere; their results
cannot be directly extrapolated.
The consensus opinion (and in the absence of conclusive tests it can only be
opinion) of the impact of satellite soundings on numerical forecasts is the fol-
lowing.
A sophisticated analysis system, such as optimum interpolation, should have the same result.
6.3 Operational Retrievals 209
forecasts, VAS soundings over the northeast Pacific Ocean improved the forecast
of 500-hPa heights over the United States in four cases and very slightly degraded
the forecast in two cases. O'Lenic (1986) reported that the VAS soundings com-
pared favorably with NOAA 7 TOVS soundings. However, VAS soundings are
not routinely used by NMC in its forecast models.
and House solved this problem by utilizing a second, narrower (more opaque)
channel. For each channel, a temperature profile is retrieved by assuming a pressure
at reference level z . If the assumed pressure is incorrect, the two temperature
Q
profiles will not agree. Iterating ρ between inversions yields accurate temperature
ο
channels had 2.0-km vertical resolution and 20-km horizontal resolution at the
tangent point except the H 0 channel, which had 2.5 x 25-km resolution. LIMS
2
channels had 1.8 x 18-km resolution except the H 0 and N 0 channels, which 2 2
had 3.6 x 28-km resolution. The SAMS was like the SSU in that it utilized filters
consisting of gas-filled cells; unlike SSU, however, SAMS was a limb scanner.
SAMS had six cells containing C 0 , CO, H 0 , NO, N 0 , and C H . Concentra-
2 2 2 4
N o t e that this problem is not unique to limb sounding. Vertical soundings are often retrieved
1 0
down to 1000 hPa because the actual surface pressure is not known. However, vertical sounding
retrieval schemes may use estimates of surface pressure from numerical models, whereas this informa-
tion is not available to limb sounding retrieval schemes.
6.5 Ozone and Other Gases 211
tions of these gases and temperatures were retrieved from SAMS data. Signifi-
cantly, SAMS could make measurements above 0.1 hPa in areas which are not
in local thermodynamic equilibrium.
Although few measurements in the stratosphere are available for comparison,
a substantial effort was made to verify Nimbus 6 measurements. Gille et al.
(1984a) reported on comparisons between LIMS temperature retrievals and radio-
sonde and rocketsonde measurements. Temperatures were retrieved from 100 to
0.1 hPa. Typical rms differences of 3 Κ between LIMS retrievals and in situ
observations were found, except for the equatorial tropopause region, where they
were higher. Below 1 hPa LIMS biases were generally less than 2 K. Above 1 hPa,
LIMS temperatures were systematically cooler than rocketsonde measurements.
Russell et al. (1984a) reported that LIMS water vapor mixing ratio retrievals
were accurate to within 2 0 - 3 0 % from 50 hPa to the stratopause (1 hPa) and to
within 40% in the lower stratosphere. Figure 6.9 shows zonal mean temperature
and water vapor mixing ratios derived from LIMS data. Barnett and Corney (1984)
compared retrieved temperatures from SAMS with radiosonde and rocketsondes.
They found temperature biases of up to 2 Κ in the lower stratosphere and 1 Κ
in the upper stratosphere. The rms errors were approximately 1-2 K.
Ozone has absorption lines in all major portions of the electromagnetic spec-
trum and thus can be measured with a variety of techniques. Krueger et al. (1980)
reviewed 21 satellite experiments (prior to SBUV/2) to measure ozone. We will
discuss a representative sample of the more recent experiments.
Chapter 6 Temperature and Trace Gases
212
-90 - 60 -30 0 30 60 90
L A T I T U D E (deg)
FIGURE 6.10. Zonal mean cross section of ozone concentration derived from L I M S data for 6 January
1979. [After Gille and Russell (1984).]
6.5 Ozone and Other Gases 213
scattering from the air (p ), reflection from the surface ( p f ) , and the zenith
air s c
ozone per unit area, that is, the total ozone. BUV, SBUV, SBUV/2, and TOMS
all make measurements of £ (X). Using two closely spaced observations in the
sun
Hartley-Huggins bands, one with a higher ozone absorption coefficient than the
other, the quantities are formed:
Nj = - 1 0 0 log 10 (6.32)
Because reflectance changes much less rapidly with wavelength than gaseous
absorption, in many cases ρ can be considered nearly constant between λ and χ
Solar Radiation
Ozone
Layer
Troposphere
>//////////?///////////////// 1
Surface
FIGURE6.il. Schematic diagram illustrating the retrieval of ozone.
Chapter 6 Temperature and Trace Gases
214
zenith angles would yield total ozone. However, because of the λ dependence - 4
scattering calculations can account for the influence of the atmosphere on p, but
not for the surface reflectance. Therefore a third observation at λ outside the 3
ozone band is necessary. This observation establishes the surface reflectance p , sfc
90 0 90 180
W longitude Ε
FIGURE 6.12. Global distribution of total ozone for December 1970 derived from Nimbus 4 BUV
measurements. The units are D o b s o n units, that is the depth (in 1 0 " cm) of pure ozone at standard
3
temperature and pressure (0°C, 1013.25 hPa). T o obtain the mass of ozone per unit area (in grams
per square meter), multiply by 0 . 0 2 1 4 . [After Krueger et al. (1980).]
PLATE 1 ATS 3 full-disk image. [Courtesy of NASA.]
PLATE 6 A perspective view of a thunderstorm cloud top with height contours derived from
stereoscopic analysis. [Courtesy of A. F. Hasler, NASA/Goddard Space Flight Center.]
6.5 Ozone and Other Gases 215
shows the global distribution of total ozone for December 1970 derived from the
Nimbus 4 BUV.
Total ozone can be retrieved in the presence of clouds because most of the
ozone lies above the cloud tops. However, a change in the retrieval scheme must
be made. When the reflectance measured by the photometer was greater than 0.6,
it was assumed that clouds were present and that the cloud top was approximately
400 hPa. Separate lookup tables were calculated for cloudy retrievals.
Klenk et al. compared their retrievals with surface-based observations and
found a bias of 3.5%, which they attribute primarily to differences in ozone
absorption coefficients used in the two techniques. The standard deviation of the
difference from ground-based measurements was 1.5-3.0%, which results in an
rms difference of 3.8-4.6%.
The Total Ozone Mapping Spectrometer (TOMS) on Nimbus 7 is much like
BUV or SBUV, except that (1) it scans instead of making measurements only at
nadir, and (2) it makes measurements only of total ozone, not ozone profiles.
Daily global maps of total ozone at 50-150-km resolution are produced. Bhartia 11
et al. (1984a) report that TOMS measures less total ozone than do ground-based
instruments. The bias is - 6 % ± 2%. This is thought to be due in large part to
a difference in the absorption coefficients used. The standard deviation between
TOMS measurements and Dobson (ground-based) measurements averaged 2.6%.
This results in an rms difference of about 6.7%. Accuracies of total ozone retrieved
from Nimbus 7 SBUV measurements are similar, and these values are probably
representative of total ozone retrieved from the SBUV/2 on the NOAA satellites.
Total ozone retrievals utilize measurements in the longer-wavelength, more
transparent portion of the Hartley-Huggins band. Estimates of ozone profiles
utilize measurements at shorter, more opaque wavelengths where the surface
contribution is negligible. The basic equation can be derived as follows (London
et al., 1977). Consider a plane-parallel atmosphere as shown in Fig. 6.11. All
UV instruments which retrieve ozone profiles only view nadir; therefore, 0 = sat
the only source of UV radiation, the no-emission equations of Section 3.3.2 apply,
specifically Eq. 3.45. Assume now that all radiation reaching the satellite is sunlight
which has been scattered once and only once from a Rayleigh scatterer. The
second term on the right-hand side of Eq. 3.45 may be neglected, and the radiative
transfer equation becomes
(1 +>sun) e X
P (6.33)
where Eq. 3.60 has been used for the Rayleigh scattering phase function, and the
scattering angle ψ has been set equal to 0 because 0 = 0. Assuming that the
5 sun sat
transmittance from the surface is zero, this equation can be integrated to yield
sun
>0
ώ exp Γ - (δ - δ ) ( 1 + — ) 1 άδ .
λ 0 λ κ (6.34)
Ό L V ^sun/
1 1
T O M S also flew on the M e t e o r - 3 5 satellite.
2]£ Chapter 6 Temperature and Trace Gases
equation becomes
\ °ώ (ρ)Ψ (ρ,μ)άρ,
Ρ
κ κ (6.35a)
This is very like Eqs. 6.3, with a slightly different form for W , and with ώ (ρ)
x λ
knowledge of the solar zenith angle and ozone absorption coefficients, it should
be possible to retrieve ώ as a function of p. Now ώ is the ratio of the scattering
λ λ
between 20 and 53 km (Cunnold et al., 1989). The SAGE II profiles were found
to be more precise than the SAGE profiles, but the SAGE profiles have been
revised and Cunnold et al. state that the SAGE and SAGE II data can be combined
to form a long-term (since 1979) stratospheric ozone data set.
All of the methods for measuring ozone have their advantages and disadvan-
tages. The solar occultation method is simple, both in equipment and in retrieval
technique. However, the fact that measurements can be made only at satellite
sunrise and sunset limits the number and the spatial distribution of the measure-
ments. Backscatter ultraviolet methods are accurate and measurements have been
made over long time periods. However, they cannot be made in darkness, such
as over the winter pole. Infrared methods, made with either a downward-looking
or a limb sounder are perhaps a little less accurate than UV methods, but they
do not depend on reflected sunlight. Retrieval of total ozone from operational
sounders is easy, twice-daily global maps of total ozone can be obtained, and
measurements have been collected since 1978 (on the NOAA satellites). Limb
infrared measurements of ozone profiles have resolution similar to BUV-estimated
profiles, but can be retrieved in daylight as well as darkness. They have been
made intermittently since 1975.
Several satellite experiments have measured stratospheric gases other than
ozone. These measurements are extremely important for the understanding of the
chemistry and dynamics of the stratosphere. All are made with infrared limb
sounders or solar occultation measurements. These experiments are summarized
in Table 6.1. As an example, Fig. 6.13 shows the zonal mean cross section of
HNO? concentration derived from LIMS data.
a
Pressure-modulated radiometry, like SSU (see Section 4.1.4).
b
CO concentration measured above 100 km.
2
2]8 Chapter 6 Temperature and Trace Gases
H N 0 Z O N A L M E A N C R O S S S E C T I O N (ppbv)
3
—ι—ι—ι—ι—1—ι—ι—ι—ι—ι—I
1/6/79
LATITUDE (deg)
FIGURE 6.13. Zonal mean cross section of HN0 3 concentration derived from LIMS data for 6 January
1979. [After Gille and Russell (1984).]
1 _
^ο( ) r λ J
0
Li = Bi(Tbi) = B (T )t
1 S 1 + B^TaXI - τ,), (6.38a)
L = B (T ) = B ( T > + B (T )(1 - τ ),
2 2 B2 2 2 2 A 2 (6.38b)
where the subscripts 1 and 2 indicate wavelength, T is a brightness temperature B
• The surface emittance and thus T will also be the same in each equation.
s
T ~txp(-ßß).
K (6.39)
220 Chapter 6 Temperature and Trace Gases
in three unknowns: T , T , and U. It will turn out, however, that because the
A s
nique can be used directly to obtain T , but not 17, from satellite measurements s
Since the weighting functions peak at the surface, T , T , and T will be close s B1 B2
Β (Τ)~Β (Τ )+^(Τ-Τ ),
λ λ Α Α (6.40)
B (T) « B (T ) + [-^fj
2 (Bi(T) - B (T )).
2 A a A (6.41)
Using Eq. 6.41 to approximate B (T ) and B (T ) yields Eq. 6.38b as 2 B2 2 S
B (T )
x m = Β {Τ,)τλ 2 + B!(T )(1 - r ) . A 2 (6.42)
Finally, eliminating Β {Τ )λ κ between Eqs. 6.38a and 6.42 yields the split-win-
dow equation
Β,(Τ,) = B {T ) + τ,ΕΒ^Τβ!) -
x m ß!(T )], B2 (6.43)
where
(6.44)
τ λ - r 2
Ts = T B1 + η(Τ Β1 - T ). B2 (6.45)
Two important points must be made about the split-window equation. First,
the split-window technique is a correction technique. The difference between
observations at two wavelengths is used to correct one of the observations for
atmospheric effects to yield an improved estimate of surface radiance or tempera-
ture. Second, the factor η does not depend on the amount of absorber 17. Since
the transmittance is relatively large in atmospheric windows, Eq. 6.39 can be
approximated as
T °*l-ßß,
k (6.46)
6.6 The Split-Window Technique 221
and, therefore
The most accurate way of calculating η is not via Eq. 6.47 but by simulating
satellite-measured radiances using a variety of model atmospheres with differing
absorber amounts. If AVHRR channels 4 (11 μπή and 5 (12 μπι) are used for
the first and second wavelengths, respectively, then η ~ 3. (Note that η is positive
because channel 5 is more sensitive to water vapor than is channel 4.)
Channels in separate windows, (e.g., the 3.7- and the 10.5-12.5-μιη windows),
which are both clouded by the same gas, can be used to estimate surface tempera-
ture, but one should not expect the equation to take the exact form of the split-
window equation. In general, the surface temperature will be a linear combination
of the measured brightness temperatures plus a constant term. When channels in
different windows are used, the technique is referred to as a multiple-window tech-
nique.
In practice the split-window technique is rarely used directly. Rather, it provides
theoretical justification for regressing the surface temperature of interest against
measured or simulated brightness temperatures.
Since 1970, NESDIS has operationally produced and archived global sea surface
temperatures (SSTs), which are important for a variety of meteorological and
oceanographic purposes (McClain et ah, 1985). Until the launch of NOAA 7,
SST was based on the 11 -μτη brightness temperature. A variety of correction
schemes were used depending on what instruments were in orbit. Prior to the
launch of NOAA 2, only crude corrections based on climatological values of
precipitable water were possible. NOAA 2, launched in 1972, had the Vertical
Temperature Profile Radiometer (VTPR), an operational sounder. Beginning in
1973, ll-μπι Scanning Radiometer (SR) measurements were corrected using re-
trieved VTPR temperature and moisture profiles. Because of problems with the
retrieved soundings, after 1976 NESDIS retrieved SSTs with regression equations
using ll-μπι SR measurements and VTPR radiances. With the launch of TIROS
N, AVHRR and HIRS/2 measurements were substituted for SR and VTPR mea-
surements. All of these SST retrieval algorithms are known as global opera-
tional SST computation (GOSTCOMP). McClain (1979) has detailed these
algorithms.
With the launch of NOAA 7 in 1981, 11-, 12-, and 3.7-/xm channels became
simultaneously available. New, more accurate SST retrieval algorithms using the
split- and multiple-window techniques were derived. The equations were based
on regression between calculated radiances and SSTs, but they were corrected
based on comparisons between NOAA 7 data and SSTs measured by drifting
buoys (Strong and McClain, 1984).
222 Chapter 6 Temperature and Trace Gases
The following discussion summarizes the major aspects of the scheme. AVHRR
data are first blocked into arrays as small as 8 km on a side, near coastlines, to
25 km on a side, in the open ocean. Daytime and nighttime retrievals have separate
algorithms. An important part of the algorithm is the detection and elimination
of clouds. In the daytime this is accomplished by taking advantage of the fact
that at 0.9 μπι the sea surface is nearly black and very uniform. Blocks in which
0.9-μπι reflectance varies by more than 0.32% or in which a pixel exceeds an
empirically determined bidirectional reflectance are rejected as cloudy. At night
reflected solar radiation is not available; less accurate techniques based on thermal
infrared observations must be used. One of these tests relies on the uniformity
of the sea surface. The l l - μ π ι brightness temperature of the pixels in a block
must agree to within 0.2 K. A second test relies on the fact that over the clear
ocean, the 11-, 12,- and 3.7-μηι brightness temperatures are correlated with each
other. Regression equations, derived from cloud-free data sets, are used to predict
the 3.7-μπι brightness temperature from the ll-μπι brightness temperature and to
predict the 1 l-μπι brightness temperature from the 12-μπι brightness temperature.
The predicted and observed temperatures must closely agree. Unfortunately, low
uniform stratus clouds can pass both of these tests. A final test, which eliminates
most stratus clouds requires that the 3.7- and 11-μπι brightness temperatures
differ by less than 0.7 K.
After cloud scenes have been eliminated, the sea surface temperatures are
retrieved. In the daytime, the 3.7-μπι brightness temperatures are contaminated
by reflected sunlight. Therefore only the 11- and 12-μπι brightness temperatures
are used in a regression equation:
SST = 1.0346T + 2.5779(T - T ) - 283.21.
n n 12 (6.48)
In this equation the SST is in degrees Celsius, and the brightness temperatures
are in kelvins. At night, all three channels may be used; three independent SSTs
are calculated:
SSTj = 1.0088T . + 0.4930(T
3 7 37 - Tu) - 273.34, (6.49a)
SST = 1.0350T + 2.5789(T
2 n n - Ti ) - 283.18,
2
(6.49b)
SST = 1.0170T„ + 0.9694(T .
3 3 7 - T12) - 276.58. (6.49c)
All three SSTs must agree to within 1 K. SST , called the triple-window solution,
3
is usually chosen as the SST. However, the 3.7-μπι AVHRR channel has had
problems with electrical noise. When the noise is too great at 3.7 μπι, SST is 2
used instead, and the consistency test is not applied. Two final tests are applied
to the retrieved SSTs, either day or night. First, the SST must be between —2.0
and + 35.0°C, and second, the SST must be within 7.0 Κ of the monthly climatolog-
ical value for its location. SSTs retrieved with this scheme are called multichannel
sea surface temperatures (MCSSTs). They are archived by NESDIS on a variety
of space and time scales (McClain et al., 1985). An example is shown in Fig. 6.14.
As with any operational retrieval system, the algorithm is occasionally updated. Users of the
1 2
60E 70E 8QE 90E 100E 110E 120E 130E 140E 150E 160E 170E 180 170W160W150W140W130W120W110W100W90W 80W 70W 60W 50W 40W 30W 20W 10W 0 10E 20E 30E 40E 50E 60E
FIGURE6.14. Monthly mean sea surface temperature (°C) for March 1 9 8 2 from A V H R R data. [Courtesy of E. Paul McClain, N O A A / N E S D I S . ]
224 Chapter 6 Temperature and Trace Gases
Instrument
T*i - ? A = (Ά ~ T )t
A u (6.50a)
7B2 - T = (T, - T ) r .
A A 2 (6.50b)
i _ T
T
M - T A
— -;= —. (6.51)
1
T
2 1
B2 ~ A
— = exp - i ß i - ß i ) - ] - (6-52)
T
2
estimate precipitable water over the United States (Plate 4). The success of this
technique rests on the fact that atmospheric temperature is much less variable
spatially than water vapor. Robinson et al. (1986) have used all 12 VAS channels
plus surface observations in regression equations to retrieve precipitable water.
Mostek et al. (1986) found these estimates useful in analyzing the clear, preconvec-
tive environment of thunderstorms.
The above infrared techniques for estimating precipitable water may be used
over either land or ocean, but they may not be used in the presence of clouds. A
similar technique employing two microwave frequencies (22.235 and 31 GHz)
can be used to retrieve precipitable water over the ocean (but not over land) in
the presence of clouds (Grody, 1976). This same technique also yields estimates
of column-integrated cloud liquid water content, again over the ocean. This
technique is discussed in Chapter 8.
Chapter 6 Temperature and Trace Gases
226
Bibliography
Achtemeier, G. L., Η. T. Ochs III, S. Q. Kidder, and R. W. Scott (1986). Evaluation of a multivariate
variational assimilation of conventional and satellite data for the diagnosis of cyclone systems. In
Υ. K. Sasaki (ed.), Variational Methods in Geosciences. Elsevier, N e w York, pp. 4 9 - 5 4 .
Anding, D., and R. Kauth (1970). Estimation of sea-surface temperature from space. Remote Sens.
Environ., 1, 2 1 7 .
Atlas, R., M. Ghil, and M. Halem (1982). The effect of model resolution and satellite sounding data
on GLAS model forecasts. Mon. Wea. Rev., 110, 6 6 2 - 6 8 2 .
Atlas, R. (1982). The growth of prognostic differences between GLAS model forecasts from SAT and
N O S A T initial conditions. Mon. Wea. Rev., 110, 8 7 7 - 8 8 2 .
Baker, W. E., R. Atlas, M. Halem, and J. Susskind (1984). A case study of forecast sensitivity to data
and data analysis techniques. Mon. Wea. Rev., 1 1 2 , 1 5 4 4 - 1 5 6 1 .
Barnett, J. J., and M. Corney (1984). Temperature comparisons between the NIMBUS 7 SAMS,
rocket/radiosondes and the N O A A 6 SSU. / . Geophys. Res., 89, 5 2 9 4 - 5 3 0 2 .
Bates, J. J., and W. L. Smith (1985). Sea surface temperature: Observations from geostationary
satellites. / . Geophys. Res., 9 0 , 1 1 6 0 9 - 1 1 6 1 8 .
Bhartia, P. Κ., K. F. Klenk, C. K. Wong, and D . Gordon (1984a). Intercomparison of the NIMBUS
7 SBUV/TOMS total ozone data sets with Dobson and M 8 3 results. / . Geophys. Res., 8 9 ,
5239-5247.
Bhartia, P. Κ., K. F. Klenk, A. J. Fleig, C. G. Wellemeyer, and K. Gordon (1984b). Intercomparison
of NIMBUS 7 solar backscattered ultraviolet ozone profiles with rocket, balloon, and Umkehr
profiles./. Geophys. Res., 89, 5 2 2 7 - 5 2 3 8 .
Chahine, Μ. T. (1968). Determination of the temperature profile of an atmosphere from its outgoing
radiance. / . Optical Soc. Am., 58, 1 6 3 4 - 1 6 3 7 .
Chahine, Μ. T. (1970). Inverse problems in radiative transfer: Determination of atmospheric parame-
ters. / . Atmos. Set., 27, 9 6 0 - 9 6 7 .
Chahine, Μ. T. (1974). Remote Sounding of cloudy atmospheres. I. The single cloud layer. / . Atmos.
Sei., 3 1 , 2 3 3 - 2 4 3 .
Chahine, Μ. T. (1975). An analytical transformation for remote sensing of clear-column atmospheric
temperature profiles. / . Atmos. Sei., 3 2 , 1 9 4 6 - 1 9 5 2 .
Chang, H. C , P. H. Hwang, Τ. T. Wilheit, A. T. C. Chang, D. H. Staelin, and P. W. Rosenkranz
(1984). Monthly distribution of precipitable water from the Nimbus 7 S M M R data. / . Geophys.
Res., 89, 5 3 2 8 - 5 3 3 4 .
Chesters, D . , L. W. Uccellini, and W. Robinson (1983). Low-level water vapor fields from the VISSR
Atmospheric Sounder (VAS) "split w i n d o w " channels. / . Climate Appl. Meteor., 22, 7 2 5 - 7 4 3 .
Chu, W. P., M. P. McCormick, J. Lenoble, C. Brogniez, and P. Pruvost (1989). SAGE II inversion
algorithm. / . Geophys. Res., 9 4 , 8 3 3 9 - 8 3 5 2 .
Chu, W. P. (1986). Inversion of SAGE II measurements. Preprints: Second Conference on Satellite
Meteorology/Remote Sensing and Applications, American Meteorological Society, Boston, pp.
J49-J51.
Chu, W. P., and M. P. McCormick (1979). Inversion of stratospheric aerosol and gaseous constituents
from spacecraft solar extinction data in the 0 . 3 8 - 1 . 0 - μ η ι wavelength range. Appl. Optics, 18,
1404-1413.
Cram, J. M., and M. L. Kaplan (1985). Variational assimilation of VAS data into a mesoscale model;
assimilation method and sensitivity experiments. Mon. Wea. Rev., 1 1 3 , 4 6 7 - 4 8 4 .
Cressman, G. P. (1959). An operational objective analysis system. Mon. Wea. Rev., 85, 3 6 7 - 3 7 4 .
Cunnold, D. M., W. P. Chu, R. A. Barnes, M. P. McCormick, and R. E. Veiga (1989). Validation
of SAGE II ozone measurements. / . Geophys. Res., 9 4 , 8 4 4 7 - 8 4 6 0 .
Curran, R. J., R. R. Nelms, F. M a r i o , R. A. Bindschadler, Ε. V. Browell, J. L. Bufton, R. L. Byer,
S. Cohen, J. J. Degnan, W. B. Grant, R. Greenstone, W. S. Heaps, Β. M. Herman, A. Jalink,
D. K. Killinger, L. Korb, J. B. Laudenslager, Μ. P. McCormick, S. H. Melfi, R. T. Menzies, V.
Mohnen, J. Spinhirne, and H. J. Zwally (1987). LASA-Lidar Atmospheric Sounder and Altimeter.
Earth Observing System Instrument Panel Report, Vol. lid, NASA, Washington, D C .
6.6 The Split-Window Technique 227
Dave, J. V. (1964). Meaning of successive iteration of the auxiliary equation in the theory of radiative
transfer. Astrophys. J., 140, 1 2 9 2 - 1 3 0 3 .
Deepak, A. (ed.) (1977). Inversion Methods in Atmospheric Remote Sounding. Academic Press,
N e w York.
Deepak, A. (ed.) (1980). Remote Sensing of Atmospheres and Oceans. Academic Press, N e w York.
Deepak, Α., Η. Fleming, and M. Chahine (eds.) (1986). Advances in Remote Sensing Retrieval Methods.
A. Deepak Publishing, Hampton, VA.
Desmarais, A. J., S. Tracton, R. McPherson, and R. van Haaren (1978). The N M C report on the
Data Systems Test. N O A A / N a t i o n a l Meteorological Center, Washington, D C .
Dey, C. H., R. A. Peterson, B. A. Ballish, P. M. Caplan, L. L. Morone, H. J. Thiebaux, Η. H. White,
Η. E. Fleming, A. L. Reale, D. G. Gray, M. D. Goldberg, and J. M. Daniels (1989). An Evaluation
of NESDIS TOVS Physical Retrievals Using Data Impact Studies. N O A A Tech. M e m o . N W S
N M C 69, Washington, D C .
Fishman, J., C. E. Watson, J. C. Larsen, and J. A. Logan (1989). Distribution of tropospheric ozone
determined from satellite data. / . Geophys. Res., 9 5 , 3 5 9 9 - 3 6 1 7 .
Fleming, Η. E., and L. M. McMillin (1977). Atmospheric transmittance of an absorbing gas. 2: A
computationally fast and accurate transmittance model for slant paths at different zenith angles.
Appl. Optics, 16, 1 3 6 6 - 1 3 7 0 .
Fleming, Η. E. (1980). Application of computerized axial tomography techniques to satellite remote
sensing of the atmosphere. Extended Abstract Volume, International Radiation Symposium, Fort
Collins, C O , 1 1 - 1 6 August, pp. 8 7 - 8 9 .
Fleming, Η. E. (1982). Satellite remote sensing by the technique of computed tomography. / . Appl.
Meteor., 2 1 , 1 5 3 8 - 1 5 4 9 .
Fleming, Η. E. (1985). Temperature retrievals via satellite tomography. In Deepak, Α., Η. Ε. Fleming,
and Μ. Τ. Chahine (eds.), Advances in Remote Sensing Retrieval Methods, A. Deepak Publishing,
Hampton, VA, pp. 55—67.
Fleming, Η. Ε., M. D. Goldberg, and D. S. Crosby (1986). Minimum variance simultaneous retrieval
of temperature and water vapor from satellite radiance measurements. Preprints: Second Confer-
ence on Satellite Meteorology IRemote Sensing and Applications. American Meteorological Society,
Boston, pp. 2 0 - 2 3 .
Foster, M. R. (1961). An application of Wiener-Kolmogorov smoothing theory to matrix inversion.
/. Soc. Indust. Appl. Math., 9, 3 8 7 - 3 9 2 .
Fritz, S., D . Q. Wark, Η. E. Fleming, W. L. Smith, H. Jacobowitz, D. T. Hilleary, and J. C. Alishouse
(1972). Temperature Sounding from Satellites. N O A A Tech. Rep. NESS 5 9 , Washington, D C .
Gandin, L. S. (1963). Objective Analysis of Meteorological Fields. Gidrometeor. Izd., Leningrad.
English translation by Israel Program for Scientific Translations, Jerusalem, 1965. [NTIS
N6618047.]
Ghil, Μ., M. Halem, and R. Atlas (1979). Time-continuous assimilation of remote-sounding data
and its effect on weather forecasting. Mon. Wea. Rev., 107, 1 4 0 - 1 7 1 .
Gille, J. C , and F. B. House (1971). On the inversion of limb radiance measurements. I: Temperature
and thickness. / . Atmos. Sei., 2 8 , 1 4 2 7 - 1 4 4 2 .
Gille, J. C , P. L. Bailey, and J. M. Russell III (1980). Temperature and composition measurements
from the l.r.i.r. and l.i.m.s. experiments on Nimbus 6 and 7. Phil. Trans. Roy. Soc. London,
A296, 2 0 5 - 2 1 8 .
Gille, J. C , J. M. Russell III, P. L. Bailey, L. L. Gordley, J..H. Lienesch, W. G. Planet, F. B. House,
L. V. Lyjak, and S. A. Beck (1984a). Validation of temperature retrievals obtained by the Limb
Infrared Monitor of the Stratosphere (LIMS) experiment on Nimbus 7. / . Geophys. Res., 89,
5147-5160
Gille, J. C , J. M. Russell III, P. L. Bailey, Ε. E. Remsberg, L. L. Gordley, W. F. J. Evans, H. Fischer,
B. W. Gandrud, A. Girard, J. E. Harries, and S. A. Beck (1984b). Accuracy and precision of the
nitric acid concentrations determined by the Limb Infrared Monitor of the Stratosphere experiment
on Nimbus 7. / . Geophys. Res., 89, 5 1 7 9 - 5 1 9 0 .
Gille, J. C , and J. M. Russell III (1984). The Limb Infrared Monitor of the Stratosphere: Experiment
description, performance, and results. / . Geophys. Res., 89, 5 1 2 5 - 5 1 4 0 .
228 Chapter 6 Temperature and Trace Gases
Gordley, L. L., and J. M. Russell III (1981). Rapid inversion of limb radiance data using an emissivity
growth approximation. Appl. Optics, 2 0 , 8 0 7 - 8 1 3 .
Grody, N . C. (1976). Remote sensing of atmospheric water content from satellites using microwave
radiometry. IEEE Trans. Antennas Propag., AP-24, 1 5 5 - 1 6 2 .
Gruber, Α., and C. D. Watkins (1982). Statistical assessment of the quality of T I R O S - N and N O A A - 6
satellite soundings. Mon. Wea. Rev., 110, 8 6 7 - 8 7 6 .
Halem, Μ., E. Kalnay, W. E. Baker, and R. Atlas (1982). An assessment of the FGGE satellite
observing system during SOP-1. Bull. Am. Meteor. Soc, 6 3 , 4 0 7 - 4 2 6 .
Hanel, R. Α., Β. Schlachman, F. D. Clark, C. H. Prokesh, J. B. Taylor, W. M. Wilson, and L. Chaney
(1970). The Nimbus III Michelson interferometer. Appl. Optics, 9, 1767.
Hanel, R. Α., Β. Schlachman, D. Rodgers, and D. Vanous (1971). The Nimbus 4 Michelson interferome-
ter. Appl. Optics, 10, 1376.
Hay den, C. M. (1973). Experiments in the four-dimensional assimilation of Nimbus 4 SIRS data.
/. Appl. Meteor., 12, 4 2 5 - 4 3 6 .
Hayden, C. M. (1979). Remote soundings of temperature and moisture. In J. S. Winston (ed.),
Quantitative Meteorological Data from Satellites. World Meteorological Organization Tech. N o t e
N o . 166 ( W M O N o . 5 3 1 ) , Geneva, pp. 1 - 3 2 .
Hayden, C. M., W. L. Smith, and Η. M. Woolf (1981). Determination of moisture from N O A A
polar orbiting satellite sounding radiances. / . Appl. Meteor., 2 0 , 4 5 0 - 4 6 6 .
Hayden, C. M. (1988). G O E S - V A S simultaneous temperature-moisture retrieval algorithm./. Appl.
Meteor., 27, 7 0 5 - 7 3 3 .
Hillger, D. W., and Τ. H. Vonder Haar (1981). Retrieval and use of high resolution moisture and
stability fields from Nimbus 6 HIRS radiances in pre-convective situations. Mon. Wea. Rev.,
109, 1 7 8 8 - 1 8 0 6 .
Houghton, J. T., F. W. Taylor, and C. D . Rodgers (1984). Remote Sounding of Atmospheres. Cam-
bridge University Press, Cambridge.
Jones, R. L., and J. A. Pyle (1984). Observations of C H and N 0 by the NIMBUS 7 SAMS: A
4 2
comparison with in situ data and two-dimensional numerical model calculations. / . Geophys.
Res., 89, 5 2 6 3 - 5 2 7 9 .
Kaplan, L. D. (1959). Inference of atmospheric structure from remote radiation measurements.
/. Optical Soc. Am., 4 9 , 1 0 0 4 - 1 0 0 7 .
Kelley, G. A. M., G. A. Mills, and W. L. Smith (1978). Impact of Nimbus 6 temperature soundings
on Australian region forecasts. Bull. Am. Meteor. Soc, 59, 3 9 3 - 4 0 5 .
Kidder, S. Q., and G. L. Achtemeier (1986). Day-night variation in operationally-retrieved TOVS
temperature biases. Mon. Wea. Rev., 1 1 4 , 1 7 7 5 - 1 7 7 8 .
Kidwell, Κ. B. (1986). NOAA Polar Orhiter Data User's Guide. N O A A / N E S D I S Satellite Data
Services Division, Washington, D C .
King, J. I. F. (1958). The radiative heat transfer of planet Earth. In J. A. Van Allen (ed.), Scientific
Uses of Earth Satellites. Univ. of Michigan Press, Ann Arbor, pp. 1 3 3 - 1 3 6 .
Klenk, K. F., P. K. Bhartia, A. J. Fleig, V. G. Kaveshwar, R. D. McPeters, and P. M. Smith (1982).
Total ozone determination from the backscattered ultraviolet (BUV) experiment. / . Climate Appl.
Meteor., 2 1 , 1 6 7 2 - 1 6 8 4 .
Koehler, T. L., J. C. Derber, Β. D. Schmidt, and L. H. Horn (1983). An evaluation of soundings,
analyses and model forecasts derived from T I R O S - N and N O A A - 6 satellite data. Mon. Wea.
Rev., I l l , 5 6 2 - 7 7 1 .
Krueger, A. J., B. Guenther, A. J. Fleig, D . F. Heath, E. Hilsenrath, R. McPeters, and C. Prabhakara
(1980). Satellite ozone measurements. Phil. Trans. Roy. Soc. London, A 2 9 6 , 1 9 1 - 2 0 3 .
London, J., J. E. Frederick, and G. P. Anderson (1977). Satellite observations of the global distribution
of stratospheric ozone. / . Geophys. Res., 8 2 , 2 5 4 3 - 2 5 5 6 .
Mateer, C. L. (1977). Experiments with the inversion of N i m b u s - 4 BUV measurements to retrieve
the ozone profile. In A. Deepak (ed.), Inversion Methods in Atmospheric Remote Sounding.
Academic Press, N e w York, pp. 577—595.
McClain, E. P. (1979). Satellite derived Earth surface temperatures. I n j . S. Winston (ed.), Quantitative
Meteorological Data from Satellites. World Meteorological Organization Tech. N o t e N o . 166
( W M O N o . 531), Geneva, 6 0 - 8 6 .
6.6 The Split-Window Technique 229
Rodgers, C. D . (1970). Remote sounding of the atmospheric temperature profile in the presence of
cloud. Quart. J. Roy. Meteor. Soc, 9 6 , 654-666.
Rodgers, C. D. (1976). Retrieval of atmospheric temperature and composition from remote measure-
ments of thermal radiation. Rev. Geophys. Space Sei., 14, 6 0 9 - 6 2 4 .
Rodgers, C. D., R. L. Jones, and J. J. Barnett (1984). Retrieval of temperature and composition from
NIMBUS 7 SAMS measurements. / . Geophys. Res., 8 9 , 5 2 8 0 - 5 2 8 6 .
Russell, J. M. III, J. C. Gille, Ε. E. Remsberg, L. L. Gordley, P. L. Bailey, H. Fischer, A. Girard, S.
R. Drayson, W. F. J. Evans, and J. E. Harries (1984a). Validation of water vapor results measured
by the Limb Infrared Monitor of the Stratosphere experiment on Nimbus 7. / . Geophys. Res.,
$9,5115-5124.
Russell, J. M. III, J. C. Gille, Ε. E. Remsberg, L. L. Gordley, P. L. Bailey, S. R. Drayson, H. Fischer,
A. Girard, J. E. Harries, and W. F. J. Evans (1984b). Validation of nitrogen dioxide results
measured by the Limb Infrared Monitor of the Stratosphere (LIMS) experiment on Nimbus 7.
/. Geophys. Res., 89, 5 0 9 9 - 5 1 0 7 .
Saunders, P. M. (1967). Aerial measurements of sea surface temperature in the infrared. / . Geophys.
Res., 72, 4 1 0 9 - 4 1 1 7 .
Schlatter, T. W. (1975). Some experiments with a multivariate statistical objective analysis scheme.
Mon. Wea. Rev., 1 0 3 , 2 4 6 - 2 5 7 .
Schlatter, T. W. (1981). An assessment of operational T I R O S - N temperature retrievals over the
United States. Mon. Wea. Rev., 109, 1 1 0 - 1 1 9 .
Schneider, W. Η., P. K. Bhartia, K. F. Klenk, and C. L. Mateer (1981). An optimum statistical technique
for ozone profile retrieval from backscattered UV radiances. Preprints: Fourth Conference on
Atmospheric Radiation, AMS, Boston, pp. 3 3 - 3 7 .
Shuman, F., and J. Hovermale (1968). An operational six-layer primitive equation forecast model.
/. Appl. Meteor., 7, 5 2 5 - 5 4 7 .
Singer, S. F., and R. C. Wentworth (1957). A method for the determination of the vertical ozone
distribution from a satellite. / . Geophys. Res., 6 2 , 2 9 9 - 3 0 8 .
Smith, W. L. (1969). A polynomial representation of carbon dioxide and water vapor transmission.
ESSA Tech. Rep. NESC 47, Washington, D C .
Smith, W. L. (1970). Iterative solution of the radiative transfer equation for temperature and absorbing
gas profile of an atmosphere. Appl. Optics, 9, 1 9 9 3 - 1 9 9 9 .
Smith, W. L., Η. M. Woolf, and Η. E. Fleming (1972). Retrieval of atmospheric temperature profiles
from satellite measurements for dynamical forecasting. / . Appl. Meteor., 1 1 , 1 1 3 - 1 2 2 .
Smith, W. L., Η. M. Woolf, P. G. Abel, C. M. Hayden, M. Chalfant, and N . Grody (1974). Nimbus 5
sounder data processing system. Part I: Measurement characteristics and data reduction procedures.
NOAA Tech. Memo. NESS 57, Washington, D C .
Smith, W. L., and Η. M. Woolf (1976). The use of eigenvectors of statistical covariance matrices for
interpreting satellite sounding radiometer observations. / . Atmos. Set., 3 3 , 1 1 2 7 - 1 1 4 0 .
Smith, W. L., Η. M. Woolf, C. M. Hayden, D . Q. Wark, and L. M. McMillin (1979a). The T I R O S - N
operational vertical sounder. Bull. Am. Meteor. Soc, 6 0 , 1 1 7 7 - 1 1 8 7 .
Smith, W. L., Η. B. Howell, and Η. M. Woolf (1979b). The use of interferometric radiance measure-
ments for sounding the atmosphere. / . Atmos. Sei., 36, 566—575.
Smith, W. L., V. E. Suomi, W. P. Menzel, Η. Μ. Woolf, L. A. Sromovsky, Η. E. Revercomb, C. M.
Hayden, D. N . Erickson, and F. R. Mosher (1981). First sounding results from V A S - D . Bull.
Am. Meteor. Soc, 6 2 , 2 3 2 - 2 3 6 .
Smith, W. L. (1983). The retrieval of atmospheric profiles from VAS geostationary radiance observa-
tions. / . Atmos. Set., 4 0 , 2 0 2 5 - 2 0 3 5 .
Smith, W. L., Η. M. Woolf, C. M. Hayden, and A. J. Schreiner (1985). The simultaneous retrieval
export package. Technical Proceedings of the Second International TOVS Study Conference,
Cooperative Institute for Meteorological Satellite Studies, University of Wisconsin, Madison,
pp. 2 2 4 - 2 7 3 .
Smith, W. L., Η. M. Woolf, and A. J. Schreiner (1986). Simultaneous retrieval of surface atmospheric
parameters: A physical and analytically correct approach. In Deepak, Α., Η. Ε. Fleming, and
6.6 The Split-Window Technique
231
Γ o SPECIFY THE current state of the atmosphere and to predict its future
state, one must know the flow field in addition to the mass field. In
this chapter we discuss the ways in which winds (the flow field) can be derived
from satellite measurements. The chapter begins with cloud- and vapor-tracking
techniques, proceeds to methods for deriving winds from the soundings discussed
in the last chapter, presents some special techniques for determining winds over
thetocean, and concludes with a new Doppler wind-estimation method introduced
on the Upper Atmosphere Research Satellite.
The great advantage of geostationary satellites is that they can make frequent
images of the same area. Since the launch of ATS 1 in 1966, meteorologists have
watched the movement of clouds and attempted to use them to estimate winds.
The concept is very simple; the vector difference of the location of a cloud in two
successive images divided by the time interval between images is an estimate of
the horizontal wind at the level of the cloud. Winds estimated by this method
are called cloud-track or cloud-drift winds. Hubert (1979) reviews the techniques
233
234 Chapter 7 Winds
used in cloud tracking. The two basic methods for tracking clouds, manual track-
ing and automatic tracking, are discussed in the next two sections.
FIGURE 7.1. (a) SMS 2 visible image at 2 0 5 8 UTC 25 April 1975. (b) Low-level winds derived by
manual cloud tracking using the image in (a) and a companion image at 2 1 0 5 UTC. [After Negri
and Vonder Haar (1980).]
7.1 Cloud and Vapor Tracking
235
tracked through thin cirrus overcast. Second, in theory more cloud vectors can
be obtained manually than with automatic methods because individual clouds (or
cloud systems) are tracked rather than areas of clouds. However, a disadvantage of
manual tracking is that it is extremely tedious, which effectively limits the number
of clouds that can be manually tracked.
Figure 7.1 shows a visible SMS 2 image and the wind vectors manually produced
from it and its companion image. Note one deficiency with cloud-track winds:
If there are no clouds, no winds can be estimated. A similar problem occurs with
hurricanes and other storms; low-level winds in the most intense part of the storm
often cannot be tracked because of obscuration by thick cirrus overcast.
It is interesting that the scanning frequency is an important factor in determining
the number of clouds that can be tracked. Rodgers et al. (1979) investigated the
effects of time and space resolution on cloud tracking in hurricanes. They found
that if 3 min or 7.5 min rapid-scan GOES data were used in place of the normal
33 Η 1 1 ι 1 1 1 1 ι f—ι—
99 98 97 96 95 94 93 92 91 90
FIGURE 7.1. {continued)
236 Chapter 7 Winds
30-min data, 10 times as many clouds could be tracked. To achieve this increase,
however, it was necessary to use full resolution (1 km) visible images. Coarser
resolution images prevented sufficiently accurate cloud location.
Leese et al. (1971) as improved by Smith and Phillips (1972) and Philips et al.
(1972). In this technique, tracking of individual cloud features, which is difficult
for a computer, is abandoned in favor of a computationally intensive method in
which the average motion of an area of clouds is calculated. Figure 7.2 illustrates
this process. In the first of two images, a target array of typically 1 6 x 1 6 pixels
is selected. The problem is to locate this area in the second image, assuming that
the clouds have moved, but changed little, during the time interval between images.
A search array of typically 32 x 32 pixels, centered on the location of the target
array, is chosen in the second image. The target array can be positioned either
vertically or horizontally to any of 17 lag positions ( - 8 < m,n < +8) in the
search array. A correlation coefficient is calculated for each lag position:
2
16 16
Σ Σ [ i+m+s,j+n+s
s
~ S{m n))](T
9 u - T)
r = (7 1)
Σ Σ [S Σ Σ(^ ~
m
* Γ 16 16 _ 16 16 _ M/2> {
'
s, s - S(m,n))] 2
f)2
U=i/=i i=i/=i J
1+m+ j+n+
where T is the pixel at the ith row and jth column in the target array (1 <
/;
< 16), Sfrj is the pixel at the kth row and /th column in the search array (1 <
kj < 32), Τ is the average of pixels in the target array, and S{m,n) is the average
of the pixels in the 16 x 16 subarea of the search array at lag position (m,n):
^ 16 16
The lag position (m,n) where the correlation coefficient is maximum is assumed
to be the final position of the target array. Navigation algorithms are used to
translate the initial and final positions into a wind vector.
The cross-correlation method works well, but the following should be consid-
ered. Computations based on Eq. 7.1 are lengthy even for a fast computer. In
practice the correlation coefficient is calculated using fast Fourier transforms (see
Leese et al., 1971). Often the volume of computation is further reduced by using
a first-guess wind from a recent analysis to estimate where to begin searching in
the search array. Calculations of the correlation coefficient are performed in a
subarea of the search array centered on the first-guess position. If a suitable peak
Pattern matching techniques have been successfully implemented. See Wolf et al. (1977) and
1
/Α
• Ρ Off:
ν
1 ο ,
ΐ Λ
Γ ι Λ ΐΝι ι/ι ι
/ Ι J I I I 1 \ <
FIGURE 7.2.Top: Visible GOES—West images for 17 July 1 9 8 7 at (left) 1916 UTC and (right) 1945
UTC. The boxes show the target array in the 1 9 1 6 UTC image and the search array in the 1945
UTC image. Middle: Blown-up images of the target and search arrays. The dashed box shows the
location of best fit of the target area, and the arrow represents the wind. Bottom: The correlation
matrix. The wind is determined by the location of the maximum correlation coefficient.
238 Chapter 7 Winds
can be found, the process is terminated. If not, calculations for the remainder of
the search array are performed.
Occasionally the algorithm fails to find a single, well-defined peak of the
correlation coefficient in the search array. Sometimes all the correlation coefficients
are small, and there is no important maximum. This usually means that the
cloud feature changed substantially between images. Sometimes the correlation
coefficients are all large and there is no important maximum. This usually means
that the cloud field was nearly uniform so that many locations produce high
correlation coefficients. Finally, there may be more than one maximum. This
occurs when the cloud field is repetitive, such as stratocumulus, so that more
than one location in the search array matches the target array. Occasionally two
levels of clouds moving with different velocities are contained in the arrays. In
this case the array of correlation coefficients often has a single peak at zero lag.
Several strategies are available for dealing with these cases. Infrared data can
be used to reject arrays in which more than one cloud layer is present. Alternately,
an analyst can locate places which are suitable for tracking and reject those which
will confuse the algorithm.
The great advantage of the automatic method is that wind vectors can be
calculated by computer; it eliminates requiring a person to perform repetitive,
error-prone tasks. However, it has two disadvantages: (1) fewer wind vectors
may be produced, since only one wind vector is calculated for each target array;
and (2) those vectors which are produced may be in error. Manual editing of the
final set of wind vectors, which is desirable for manually produced wind vectors,
is essential for automatically produced vectors.
Editing of wind vectors is a complex task having both internal (computer-
done) and external (manually done) components. The internal components may
consist of comparing each vector with its neighbors, with an existing analysis, or
with an analysis of the entire set of wind vectors. If a vector deviates excessively
from the analysis, it is rejected or flagged for manual inspection. Also, if the
correlation coefficient matrix fails to find an unambiguous peak, the vector is
rejected or flagged. The external component consists of inspection by an analyst,
who looks at the plotted vectors, especially any flagged vectors, to determine
whether they appear to be meteorologically meaningful. Depending on the applica-
tions intended for the vectors, a balance must be achieved between the internal
and external editing process. In general, it is better to have the internal editor
flag rather than reject vectors because many meteorologically interesting situations
produce "anomalous" winds. An analyst can decide better than a computer which
vectors are significant. On the other hand, the analyst should not have to examine
too many vectors, else the advantage of the automatic method is lost.
7.1.3 Quality
Very accurate navigation and image registration are required to produce accu-
rate cloud-track winds. Navigation errors which do not change between images
do not degrade the accuracy of the wind calculation, but they cause the location
7.1 Cloud and Vapor Tracking 239
of the wind vector to be misplaced horizontally. Usually this type of error is not
serious. Far more serious are navigation errors which change between images.
These result in errors, sometimes large, in the wind vector itself. Fortunately,
these errors are infrequent today. When they occur, the pair of pictures with
which the winds are produced is usually abandoned. However, one of the images
can be reregistered by shifting it until landmarks are lined up in both images.
The process of retrieving winds then proceeds as above. If reregistration is used,
however, cloud vectors which are far from a landmark will be less accurate than
those which are closer because the navigation algorithms are nonlinear.
Not all clouds move passively with the wind. Clouds with large vertical extent
such as cumulonimbus clouds must be avoided. So too must orographically gener-
ated clouds, clouds associated with gravity waves, and clouds such as thunder-
storm anvils which are evaporating or growing in a way not associated with the
ambient wind. One must also be careful about clouds at frontal boundaries.
Studies have been done to determine which types of clouds are suitable for
tracking. Fujita et al. (1975) concluded that cumulus clouds less than 0.5 km in
diameter are too small to represent the wind. The clouds that best represented
the subcloud layer wind are cumulus between 0.5 and 4 km in diameter. Larger
cumulus clouds were also found to be suitable tracers if their vertical extent is
not too large. Hasler et al. (1979) used in situ aircraft observations to determine
that low-level cumulus clouds and cirrus clouds move essentially with the wind
and thus are good candidates for tracking. Few clouds can be tracked at midlevels.
A related question is, What types of clouds can be tracked with satellite imagery?
Visible images from the GOES satellites have 1-km resolution. Thus clouds a few
kilometers in diameter can be tracked. If infrared imagery is used, as it must be
at night, the 3 x 7-km resolution limits the size of what can be tracked. As
discussed in Section 5.4.7.6, a cloud that fills only a single pixel is not likely to
be well positioned. Thus, clouds or cloud systems several tens of kilometers
across—larger than the ideal cloud—must be used for tracking in the infrared.
Finally, the height assignment of the calculated wind vectors is difficult. The
motion of small cumulus clouds best represents the wind at cloud base, and the
motion of cirrus best represents the mean wind in the cloud. How does one
estimate these levels?
Several methods for using the satellite data themselves to make a height estimate
are discussed in Chapter 8. These include observations of cloud shadows, compari-
son of the equivalent blackbody temperature of the cloud top with a sounding,
and stereoscopic estimates made using two different satellites to view the same
cloud. Unfortunately, shadows and stereoscopy can be used only occasionally,
shadows because only low sun angles are useful, and stereoscopy because the
cloud must be viewed by each satellite at nearly the same time (normally
GOES-East and GOES-West are scheduled 15 min apart). IR height estimates
may be unreliable because trackable clouds have either unknown emittance (cirrus)
or unknown cloud fraction (low cumulus). Other techniques include (1) assigning
the vector to the level that deviates least from the latest analysis and (2) assigning
the vector to the level that in the past has agreed best with rawinsonde observa-
Chapter 7 Winds
240
tions. Often the later is used. Two multichannel infrared techniques for estimating
cloud height are discussed in Sections 7.1.5 and 7.1.6.
Unfortunately, there is more uncertainty in the height of the cloud than we
would like. This translates into an error in the wind because placing the correct
wind at the wrong level is just as bad as having the wrong wind at the correct
level.
All winds other than the low-level winds are derived manually by NESDIS on
the VDUC system, and manual tracking is used at the discretion of the analyst
3
to supplement the picture-pair winds at low levels. If three usable images are
available, two wind vectors are calculated, one between the first and second
images, and one between the second and third images. The two vectors are
compared. If they differ too much, they are discarded. If they are acceptable, a
vector between the first and last images is used as the wind. All vectors are
subjectively assigned to a pressure level by the analyst based on (1) the equivalent
blackbody temperature of the cloud, (2) recent analyses, and (3) experience. The
vast majority of these vectors are high-level winds, but middle- and even low-
level winds are occasionally produced manually.
All winds, both automatic and manual, are inspected by an analyst to ensure
consistency. The analyst is free to delete any vector that seems unjustified. It takes
about 3 h to produce a set of wind vectors from the two GOES satellites. Plate
5 shows one set of winds from GOES—West. The average set contains about
500 vectors, including all levels from both GOES satellites. These vectors are
transmitted over GTS as are the vectors produced by ESA, ISRO, and MSC.
NESDIS has archived GOES cloud-track winds since 1974. ESA, MSC and ISRO
also archive the winds which they produce.
7.1.4.2 Accuracy
For several years the World Meteorological Organization (WMO) Committee
on Coordination for Geostationary Meteorological Satellites (CGMS) has orga-
nized the collection of wind data sets from satellites and rawinsondes for compari-
son. Satellite winds are compared with (1) winds from other satellites (in the
region where satellite coverage overlaps) and (2) rawinsonde winds. NESDIS
performs comparisons between winds from GOES, Meteosat, and GMS and
between GOES winds and rawinsonde winds. These comparisons are similar to
those discussed in Chapter 6 between temperature and moisture soundings made
by satellites and rawinsondes. Observations that are nearly collocated in time
and space are compared. Recent comparisons between winds from different satel-
lites have yielded rms vector differences of about 6 m s" for high-level winds
1
and about 4 m s" for low-level winds. Comparison between satellite winds and
1
There are several sources of differences between satellite winds and rawinsonde
winds, among them: improper height assignment, inexact collocation in either
time or space, errors of measurement and tracking, and unrepresentative rawin-
sonde winds. Height assignment appears to be a primary source of error. One
would expect errors in height assignment in a sheared environment to increase
rms vector differences. When satellite winds are compared with rawinsonde winds
3
ESA uses automatic tracking for all of its winds (Bowen et al., 1979). Histograms of the IR and
visible pixels are used to determine whether an area is trackable, and if so, to what level to assign
the vector.
242 Chapter 7 Winds
not at the assigned level (as above) but at the level where the satellite wind differs
least from the rawinsonde wind (the level of best fit), rms vector differences
decrease by 4 0 - 5 0 % . Also, satellite—satellite differences are smaller than satel-
lite—rawinsonde differences. Since both satellites track virtually the same clouds,
height assignment is not a factor in this comparison. If height assignment can be
improved, cloud-drift wind errors will likely decrease. Sections 7.1.5 and 7.1.6
describe techniques that offer improved height assignment.
7.1.5 C 0 Slicing
2
Menzel et al. (1983) have used VAS images in the 15-μπι C 0 band to track
2
clouds. Figure 7.3 shows images in VAS channels 3, 4, and 5 (14.2, 14.0, and
13.3 μηι). Note that as one moves closer to the 15-μιη center of the C 0 band, 2
low clouds disappear. Clouds that appear only in channel 5 are low clouds. Clouds
that appear in channels 4 and 5 but not in channel 3 are middle clouds. Clouds
that appear in all channels have high cloud tops. This slicing of the atmosphere
using C 0 channels results in winds that are more likely to be assigned to the
2
Although 6.7-μηι water-vapor images have been available for many years on
the polar-orbiting Nimbus satellites, the advent in 1977 of the high-resolution (5
km) water vapor channel on Meteosat made possible the tracking of water vapor
features (Eigenwillig and Fischer, 1982). Vapor tracking can also be accomplished
using lower-resolution (14-km) VAS data (Stewart et al., 1985). The advantage
of vapor-track winds is that they may be estimated in areas that are free of clouds.
In addition, since the weighting function of the water vapor channels peaks near
the 450-hPa level (depending on moisture), the derived winds are representative
of the middle troposphere, a region in which there are few trackable clouds. Thus
vapor-track and cloud-track winds complement each other.
7.1 Cloud and Vapor Tracking 243
FIGURE 7.3. VAS images used in C 0 2 slicing. [After Menzel et al. (1983).]
244 Chapter 7 Winds
In general, the flow field is not independent of the mass field. Given temperature
soundings and an estimate of the sea-level pressure, one can calculate the height
of any pressure surface. In the extratopics, winds are related to the height field.
Three relationships come immediately to mind. The geostrophic wind is related
to the gradient of the geopotential:
V = y k x V<t>,
g (7.3)
V
^ / v »
V/2 ·
(7 4)
where R is the radius of curvature of the trajectory of an air parcel. The gradient
x
wind is a better approximation for curved flow than the geostrophic wind. Finally,
the quasi-geostrophic approximation (see Holton, 1992) may be used to estimate
the wind. The question is: How accurate are winds calculated from satellite
temperature soundings?
Hayden (1984) used surface pressure reports and station elevations to calculate
the 1000-hPa height field. He then used VAS soundings and the hypsometric
7.3 Ocean Surface Winds 245
under 10 m s" . The correlation coefficient was 0.87. At 850 hPa, the results
1
were not as good. Although the rms errors were lower than at 300 hPa, the
correlation coefficient was much lower, essentially zero on one of the days. These
results make physical sense because satellite soundings are least accurate near the
surface. On the other hand, the mean temperature of a deep layer of the tropo-
sphere is accurately retrieved from radiance data. Therefore, the winds aloft ought
to be accurately estimated by this technique, assuming that the gradient wind
equation is capable of approximating the actual wind. Hayden's results are sup-
ported by those of Lord et al. (1984), who compared winds derived from satellite
soundings with aircraft observations.
It is interesting to note that winds derived from VAS data complement cloud-
track winds because VAS winds can only be calculated in clear areas.
Unique properties of the ocean surface or of oceanic storms have allowed the
development of several techniques to measure surface winds over the ocean.
4
In the gradient wind relationship, R was approximated as the radius of curvature of the geo-
T
strophic streamline.
246 Chapter 7 Winds
Seasat was launched on 28 June 1978, but it suffered a power supply failure on
10 October 1978.
SASS had bar-shaped, dual-polarization, fan-beam antennas extending horizon-
tally at angles of 45°, 135°, 225°, and 315° from the satellite velocity vector. The
instrument was designed to measure σ° in three swaths, one left, one right, and
one centered on the satellite ground track (Fig. 7.4). The orientation of the
antennas caused the instrument to scan at a 45° angle to the ground track.
The antenna footprint was electronically divided by Doppler filters into cells
approximately 18 x 70 km. To estimate both wind direction and wind speed,
each spot of ocean had to be observed at least twice from different azimuth angles.
SASS accomplished this by observing a spot with the forward antennas, then
observing it at right angles 1-3 min later with the aft antennas.
The relationship between σ° and surface wind is complex and not completely
understood. We present only a simplified treatment; the reader is referred to
Stewart (1985) for a more thorough discussion. Wind induces waves of all wave-
lengths on the sea surface. Waves that are near the wavelength of the radar
backscatter radiation through Bragg scattering: reflected radiation from one wave
reinforces (or cancels) that from the successive wave depending on wave spacing
and viewing geometry. Since the SASS wavelength was about 2 cm, centimeter-
size capillary waves are responsible for the backscatter. Several parameters influ-
ence σ°: primarily wind speed, zenith angle, relative azimuth between the radar
beam and the wind, and radiation polarization; but also water temperature,
viscosity, the spectrum of larger waves, and surface tension. The largest backscatter
occurs in the upwind and downwind directions. Backscatter minima are in the
vicinity of the crosswind directions.
To account for the large wind shear that occurs in the boundary layer, the
wind speed U employed in these calculations is taken to be the 19.5 m neutral
SIDE VIEW
INSTRUMENT CHARACTERISTICS
CONSTANT 14.59927 GHz
DOPPLER 100 W PEAK RF POWER
LINES ELECTRONIC SCAN (15 DOPPLER CELLS)
ORTHOGONAL MEAS. (AZIMUTH)
ANTENNA 4 ANTENNAS
ILLUMINATION DUAL POLARIZATION
PATTERN 0.5°x25°ANTENNA BEAM
I TOP VIEW
ANTENNA BEAM
NUMBER
SASS SIDE SWATH NEAR-NADIR SWATH
(WITH OVERLAP) SAT TRACK
FIGURE 7.4. SASS swath geometry. [After Jones et al. (1982).]
7.3 Ocean Surface Winds 247
stability wind, which is the wind speed that would occur at a height of 19.5 m
above the surface if the atmosphere had neutral (dry adiabatic) stability. U is
calculated from in situ observations by applying a boundary-layer model which
uses the height of the anemometer (among other things) as an input parameter.
The algorithm for retrieving wind speed and direction from SASS observations
of σ° is called the SASS I model function (Schroeder et al., 1982; Jones et al.
1982). The model assumes that
σ° = GU , H
(7.5)
where G and Η are functions of zenith angle (ζ), relative azimuth angle (χ),
and polarization (P, vertical or horizontal). Tabulated values of G and Η were
determined empirically by fitting prelaunch aircraft and postlaunch SASS observa-
tions of σ° to in situ observations of wind speed and direction. Suppose that SASS
has made two measurements of σ° at 90° angles. Then we have two equations:
a? = G(£ , ,P )U <W.>.
1 X l 1
H
(7.6a)
solved by plotting U versus χ (Fig. 7.5). The two equations yield two curves.
2
FIGURE 7.5. A plot of possible wind speeds and directions given measurements of σ ° made by the
forward and aft antennas. The intersections of the t w o curves are the aliases, one of which represents
the true 19.5-m neutral stability wind. [After Jones et al. (1982).]
248 Chapter 7 Winds
a least-squares estimator is used to derive the aliases. In the narrow swath under
the satellite, the wind speed, but not the wind direction, can be retrieved.
The SASS retrievals must be dealiased using auxiliary meteorological informa-
tion such as ship wind reports or model first-guess fields. Figure 7.6 shows an
example of SASS aliases, a model first-guess wind analysis, and a dealiased analysis
produced by an objective dealiasing procedure (Baker et al., 1984). This dealiasing
procedure is complicated, and it has been suggested that future scatterometers
have two additional antennas to reduce the number of aliases.
The goal for SASS was to retrieve ocean surface winds with rms errors of 2
m s in speed and 20° in direction. Lame and Born (1982) reported wind speed
- 1
however, report that a substantial number of the retrieved SASS winds may not
have met the accuracy goals. They found several areas where the SASS I model
function was deficient, such as in neglecting viscosity and surface tension, and
they suggest that an improved model is both necessary and possible.
In addition to establishing accuracy, it is important to determine the contribu-
tion which a satellite instrument may make to meteorological analyses and fore-
casts. In September 1978, an explosively developing cyclone (dubbed the QEII
storm) off the east coast of the United States damaged the ocean liner Queen
Elizabeth II (Gyakum, 1983). Duffy and Atlas (1986) found that SASS surface
winds had little impact on forecasts of the storm. However, when upper-level
winds were made to agree with the SASS surface winds, a large positive impact
was found on forecasts. In other words, surface winds alone were insufficient to
improve forecasts. Duffy et al. (1984) found a small, inconsistent impact of SASS
winds in forecasts made with the U.S. Navy's operational numerical forecast
model. They cite incorrectly dealiased winds as a possible reason for this low
impact. Baker et al. (1984) found a small positive impact of SASS winds, but this
positive impact disappeared when satellite temperature soundings were included.
This indicates that temperature soundings and surface winds may be partially
redundant in their ability to improve weather forecasts.
170W 168W 166W 164W 162W 160W 170W 168W 166W 164W 162W 160W 170W 168W 162W 160W
FIGURE 7.6. (a) Ambiguous SASS wind vectors (aliases) at approximately 2 2 0 0 UTC 7 September 1 9 7 8 .
(b) Model first guess 1000-hPa wind vectors and streamlines for 0 0 0 0 UTC 8 September 1 9 7 8 . (c)
Objectively dealiased SASS winds. [After Baker et al. (1984).]
7.3 Ocean Surface Winds 249
Microwave radiation emitted from the sea surface is a function of wind speed for
two reasons. First, the same wind-generated waves that alter the radar backscatter
coefficient also change the surface emittance. Figure 7.7 shows the calculated
emittance at 19.4 GHz of a wind-roughened ocean surface as a function of
incidence (viewing) angle. The emittance change is strongly dependent on polariza-
FIGURE 7.7. Calculated emittance at 19.4 GHz of a wind-roughened ocean surface as a function of
incidence angle (measured from vertical): (a) horizontal polarization, (b) vertical polarization. [After
Stogryn (1967). © 1 9 6 7 IEEE.]
Chapter 7 Winds
250
tion and viewing angle. For nadir viewing, wind-induced roughness has little
effect on emittance. For nonnadir viewing, roughness increases emittance for
horizontally polarized radiation; roughness decreases emittance for vertically po-
larized radiation out to 55° and increases it thereafter. Note that the angle between
the viewing azimuth and the wind direction affects the emittance slightly. The
second reason that sea surface emittance is a function of wind speed is that at
wind speeds above about 7 m s foam begins to form on the sea surface. The
- 1
fraction of the surface covered by foam increases with wind speed. Since foam is
essentially black, emittance increases rapidly, and nearly linearly, with wind speed
above 7 m s . Wilheit (1979) has proposed a sea surface emittance model which
_1
instrument on the DMSP satellites can also estimate surface wind speed over the
ocean. Since surface emittance is only weakly dependent on wind direction, passive
microwave techniques yield only wind speed.
The strongest and most dangerous winds over the ocean are the result of
tropical cyclones. One of the chief contributions of meteorological satellites is
simply the detection of these storms. Since the 1960s, no tropical cyclone anywhere
on Earth has gone undetected. However, while shipping interests are pleased to
know about the location of tropical cyclones, they also need to know the wind
speed so that they can keep their ships at a safe distance. Not all tropical cyclones
are equally dangerous in terms of wind speed and the ocean waves which result.
There is a great deal of interest, therefore, in using satellite data to estimate
hurricane winds.
(T1-T8) are used for this purpose. The T-numbers are determined by detailed
examination of the images following a decision tree (Fig. 7.9) and by guidelines
on the expected day-to-day change in the T-number of a storm. Once-daily
images are generally used in this technique, and it is desirable for the images to
be at the same time each day to avoid confusion caused by diurnal changes in
tropical storm cloudiness. The T-number is related to the minimum sea level
pressure (central pressure), although the relationship is different in the different
oceans. A current intensity (CI) number can be derived from the T-number. The
CI number is related to the storm's intensity (maximum sustained surface wind
speed). Comparison with aircraft-observed winds shows that the Dvorak tech-
nique has an rms error of approximately 6 m s" in tropical cyclone intensity.
1
In 1984 Dvorak introduced a variant of the above technique, called the en-
hanced IR technique, which uses specially enhanced infrared images instead of
visible images. This, of course, allows intensity estimates to be made at night. A
decision tree similar to that in Fig. 7.9 is used in the analysis.
The visible and enhanced IR Dvorak techniques are subjective and require
training to be used effectively, but they yield good wind estimates. They are in
use throughout the tropical regions of the world.
Partly to lessen the subjective nature of the Dvorak technique, Dvorak (1984)
introduced a digital IR technique for estimating tropical cyclone strength. The
digital IR technique uses two parameters: the equivalent blackbody temperature
of the eye of a storm (T ), and the average equivalent blackbody temperature
eye
the more intense the storm. T-numbers can be estimated from these two parame-
ters using a table (see Dvorak, 1984).
T1 T2 T3
FIGURE 7.8. Examples of tropical cyclone patterns. The storm type (a-f) is noted at the left (types g
and h are not shown). The storm intensity as indicated by the T - n u m b e r is noted at the top. [After
Dvorak (1975).]
7.3 Ocean Surface Winds 253
T4 T5 T6
PRELIMINARY T* CF+BF
C F - C E N T R A L FEATURES
B F . O U T E R BANDING FEATURES
C L O U D LINE OR BAND F E A T U R E S D E F I N E T H E C S C
BANDING TYPE I TIGHT C U R V A - LOW CLD CSC
EYE*? TURE AT EDGE Of) NEAR D E N S E θ[
DENSE Φ BANDJ NO- *3*DENSE φ ? OR CB LINE CSC|
WIDTH: W I T H D I A M OF] (CONV® > 3 * )
DIAM OF CURV:
r ^
CURVATURE:
51* <2'/>· I
YES YES YES YES YES YES NO
I I I I I I
CF4 CF3 CF3 CF2.5 CF2 CFI CFO
* E Y E QUALIFICATIONS:
IS THERE AROUND 1. E Y E IS DEFINED ONLY A F T E R T 2 OR GREATER O B S E R -
DISTINCT EYE*? VATION WAS MADE PREVIOUSLY
IMBEDDED BY: 2 . R A G G E O E Y E : < C F 4 . 5 , S U B T R A C T 1/2 NUMBER; FOR
> C F 5 , SUBTRACT I NUMBER.
FIGURE 7.9. Decision tree used to determine T-numbers. [After Dvorak (1975).]
The MSU has better spatial resolution than did the SCAMS, and it measures
the brightness temperatures more accurately. Second, they retrieved atmospheric
temperature profiles inside the tropical cyclones. The retrieved temperature anom-
aly at 250 hPa is correlated with both the central pressure (minimum sea-level
pressure) and with intensity (maximum sustained sea-level wind speed). They
found rms errors of about 6 hPa in central pressure estimates and about 6 m s - 1
in intensity estimates. This technique has been used semioperationally at the U.S.
National Hurricane Center.
7.4 Doppler Wind Measurements 255
A difficulty with the technique is that none of the microwave sounders launched
to date have provided contiguous coverage of the Earth; gaps exist between
consecutive orbits. Furthermore, the outermost scan spots must be discarded
because the resolution becomes too poor. This widens the gaps between consecu-
tive orbits to the point that there is only about a 50% probability that a particular
storm will be viewed sufficiently well to apply these techniques. This difficulty is
one of the reasons that placing a microwave sounder on a geostationary satellite
would be extremely valuable.
The Upper Atmosphere Research Satellite (UARS) has two instruments devoted
to measuring horizontal winds, which are important because the dynamics and
chemistry of the upper atmosphere are inextricably linked. The instruments are
the High Resolution Doppler Imager (HRDI) and the Wind Imaging Interferometer
(WINDII). Both view the limb of the atmosphere and measure the Doppler shift
in emission and absorption lines in the visible and near-infrared portion of the
spectrum. The chemical species detected are molecular oxygen, atomic oxygen,
and the OH molecule. The Doppler shift determines the relative velocity between
the satellite and the volume of air being sampled. Correcting for the motion of
the satellite and the rotation of the Earth yields one component of the horizontal
wind. By observing the same location both as the satellite approaches and as it
departs from an area, one can estimate the horizontal wind vector. The instruments
scan vertically to give a vertical profile of the wind with about 4-km vertical
resolution. It is expected that winds will be measured to accuracies of between
5 and 15 m s" , depending on altitude. Since UARS was launched as this book
1
was being written, the reader is referred to the literature for results of these wind
measurements.
Bibliography
Atlas, R., E. Kalnay, W. E. Baker, J. Susskind, D. Reuter, and M. Halem (1985). Simulation studies
of the impact of future observing systems on weather prediction. Preprints: Seventh Conference
on Numerical Weather Prediction, American Meteorological Society, Boston, pp. 145—151.
Baker, W. E., R. Atlas, E. Kalnay, M. Halem, P. M. Woiceshyn, S. Peteherych, and D . Edelmann (1984).
Large-scale analysis and forecast experiments with wind data from the Seasat A scatterometer. / .
Geophys. Res., 8 9 , 4 9 2 7 - 4 9 3 6 .
Bengtsson, L., M. Kanamitsu, P. Kallberg, and S. Uppala (1982). FGGE 4-dimensional data assimilation
at ECMWF. Bull. Am. Meteor. Soc, 6 3 , 2 9 - 4 3 .
Bowen, R. Α., L. Fusco, J. Morgan, and K. O. Roska (1979). Operational Production of Cloud
Motion Vectors {Satellite Winds) from METEOSAT Image Data: Use of Data from Meteorological
Satellites. European Space Agency Report N o . 1 4 3 , 6 5 - 7 5 .
Curran, R. J., D. E. Fitzjarrald, J. W. Bilbro, V. J. Abreu, R. A. Anthes, W. E. Baker, D . A. Bowdle,
D. Burridge, G. D . Emmitt, S. E. D . Ferry, P. H. Flamant, R. Greenstone, R. M. Hardesty, K. R.
Hardy, R. M. Huffaker, M. P. McCormick, R. T. Menzies, R. M. Schotland, J. K. Sparkman, Jr.,
J. M. Vaughan, and C. Werner (1987). Laser Atmospheric Wind Sounder Instrument Panel Report.
256 Chapter 7 Winds
Earth Observing System, Vol. Ilg, National Aeronautics and Space Administration, Washing-
ton, D C .
Desbois, M., G. Seze, and G. Szejwach (1982). Automatic classification of clouds on METEOSAT
imagery: Application to high-level clouds. / . Appl. Meteor., 2 1 , 4 0 1 - 4 1 2 .
Duffy, D. G., R. Atlas, T. Rosmond, E. Barker, and R. Rosenberg (1984). The impact of Seasat
scatterometer winds on the Navy's operational model. / . Geophys. Res., 8 9 , 7 2 3 8 - 7 2 4 4 .
Duffy, D . G., and R. Atlas (1986). The impact of Seasat-A Scatterometer data on the numerical
prediction of the Queen Elizabeth II storm. / . Geophys. Res., 9 1 , 2 2 4 1 - 2 2 4 8 .
Dvorak, V. F. (1973). A technique for the analysis and forecasting of tropical cyclone intensities from
satellite pictures. N O A A Tech. M e m o . NESS 4 5 , U.S. Department of Commerce, Washington, D C .
Dvorak, V. F. (1975). Tropical cyclone intensity analysis and forecasting from satellite imagery. Mon.
Wea. Rev., 1 0 3 , 4 2 0 - 4 3 0
Dvorak, V. F. (1984). Tropical Cyclone Intensity Analysis Using Satellite Data. N O A A Tech. Rep.
NESDIS 1 1 , Washington, D C .
Eigenwillig, Ν . , and Η. Fischer (1982). Determination of midtropospheric wind vectors by tracking
pure water vapor structures in Meteosat water vapor image sequences. Bull. Am. Meteor. Soc,
63, 4 4 - 5 8 .
Endlich, R. M., and D . E. Wolf (1981). Automatic cloud tracking applied to GOES and Meteosat
observations. / . Appl. Meteor., 2 0 , 3 0 9 - 3 1 9 .
Fujita, Τ. Τ., E. W. Pearl, and W. E. Shenk (1975). Satellite-tracked cumulus velocities. / . Appl.
Meteor., 14, 4 0 7 - 4 1 3 .
Green, R., G. Hughes, C. Novak, and R. Schreitz (1975). The automatic extraction of wind estimates
from VISSR data. In C. L. Bristor (ed.), Central Processing and Analysis of Geostationary Satellite
Data. N O A A Tech. M e m o . NESS 64, Washington, D C .
Grody, N . C , C. M. Hayden, W. C. C. Shen, P. W. Rosenkranz and D . H. Staelin (1979). Typhoon
June winds estimated from scanning microwave spectrometer measurements at 55.45 GHz. / .
Geophys. Res., 84, 3 6 8 9 - 3 6 9 5 .
Gyakum, J. R. (1983). On the evolution of the QEII storm: Synoptic aspects. Mon. Wea. Rev.,
I l l , 1137-1155.
Halem, M., E. Kalnay, W. E. Baker, and R. Atlas (1982). An assessment of the FGGE satellite
observing system during S O P — 1 . Bull. Am. Meteor. Soc, 6 3 , 4 0 7 - 4 2 6 .
Hasler, A. F., W. C. Skillman, W. E. Shenk, and J. Steranka (1979). In situ aircraft verification of
the quality of satellite cloud winds over oceanic regions. / . Appl. Meteor., 18, 1 4 8 1 - 1 4 8 9 .
Hasler, A. F. (1981). Stereographic observations from geosynchronous satellites: An important new
tool for the atmospheric sciences. Bull. Am. Meteor. Soc, 6 2 , 1 9 4 - 2 1 2 .
Hayden, C. M. (1984). Estimating the Wind Field from VAS Temperature Soundings. Unpublished
manuscript. Available from the author at N O A A / N E S D I S , 1225 W. Dayton St., Madison, WI
53706.
Holton, J. R. (1992). An Introduction to Dynamic Meteorology, 3rd ed., Academic Press, San Diego.
Hubert, L. F., and L. F. Whitney, Jr. (1971). Wind estimation from geostationary-satellite pictures.
Mon. Wea. Rev., 9 9 , 665-672.
Hubert, L. F., and L. F. Whitney, Jr. (1974). Compatibility of Low-Cloud Vectors and Rawinsondes
for Synoptic Scale Analysis. N O A A Tech. Rep. NESS 7 0 , Washington, D C .
Hubert, L. F. (1979). Wind derivation from geostationary satellites. I n j . S. Winston (ed.). Quantitative
Meteorological Data from Satellites. World Meteorological Organization Tech. N o t e N o . 166
( W M O N o . 5 3 1 ) , Geneva, pp. 3 3 - 5 9 .
Hubert, L. F., and A. Thomasell, Jr. (1979). Error Characteristics of Satellite-Derived Winds. N O A A
Tech. Rep. NESS 7 9 . Washington, D C .
Jones, W. L., L. C. Schroeder, D . H. Boggs, E. M. Bracalente, R. A. Brown, G. J. D o m e , W. J. Pierson,
and F. J. Wentz (1982). The S E A S A T - A satellite scatterometer: The geophysical evaluation of
remotely sensed wind vectors over the ocean. / . Geophys. Res., 87, 3 2 9 7 - 3 3 1 7 .
Kastner, Μ., Η. Fischer, and Η.-J. Bolle (1980). Wind determination from Nimbus 5 observations
in the 6.3 μτη water vapor band. / . Appl. Meteor., 19, 4 0 9 - 4 1 8 .
7.4 Doppler Wind Measurements
257
Kidder, S. Q., W. M. Gray, and Τ. H. Vonder Haar (1978). Estimating tropical cyclone central
pressure and outer winds from satellite microwave data. Mon. Wea. Rev., 106, 1 4 5 8 - 1 4 6 4 .
Kidder, S. Q., W. M. Gray, and Τ. H. Vonder Haar (1980). Tropical cyclone outer surface winds
derived from satellite microwave sounder data. Mon. Wea. Rev., 108, 1 4 4 - 1 5 2 .
Lame, D . B., and G. H. Born (1982). Seasat measurement system evaluation: Achievements and
limitations. / . Geophys. Res., 87, 3 1 7 5 - 3 1 7 8 .
Leese, J. Α., C. S. N o v a k , and Β. B. Clark (1971). An automatic technique for obtaining cloud motion
from geostationary satellite data using cross correlation. / . Appl. Meteor., 10, 1 1 8 - 1 3 2 .
Lord, R. J., W. P. Menzel, and L. E. Pecht (1984). ACARS wind measurements: An intercomparison
with radiosonde, cloud motion, and VAS thermally derived winds. / . Atmos. Ocean. Tech.,
1, 1 3 1 - 1 3 7 .
Mack, R. Α., A. F. Hasler, and R. F. Adler (1983). Thunderstorm cloud top observations using
satellite stereoscopy. Mon. Wea. Rev., I l l , 1 9 4 9 - 1 9 6 4 .
Menzel, W. P., W. L. Smith, and T. R. Stewart (1983). Improved cloud motion-vector and altitude
assignment using VAS. / . Climate Appl. Meteor., 22, 3 7 7 - 3 8 4 .
Negri, A. J., and Τ. H. Vonder Haar (1980). Moisture convergence using satellite-derived wind fields:
A severe local storm case study. Mon. Wea. Rev., 108, 1 1 7 0 - 1 1 8 2 .
Njoku, E. G., and L. Swanson (1983). Global measurements of sea surface temperature, wind speed
and atmospheric water content from satellite microwave radiometry. Mon. Wea. Rev., I l l ,
1977-1987.
Phillips, D . R., E. A. Smith, and V. E. Suomi (1972). Comment on "An automated technique for
obtaining cloud motion from geosynchronous satellite data using cross correlation." / . Appl.
Meteor., 1 1 , 7 5 2 - 7 5 4 .
Rodgers, E. B., R. C. Gentry, W. Shenk, and V. Oliver (1979). The benefits of using short-interval
satellite images to derive winds for tropical cyclones. Mon. Wea. Rev., 107, 5 7 5 - 5 8 4 .
Rodgers, E. B., R. Mack, and A. F. Hasler (1983). A satellite stereoscopic technique to estimate
tropical cyclone intensity. Mon. Wea. Rev., I l l , 1 5 9 9 - 1 6 1 0 .
Rosenkranz, P. W., D. H. Staelin, and N . C. Grody (1978). Typhoon June (1975) viewed by a scanning
microwave spectrometer./. Geophys. Res., 8 3 , 1 8 5 7 - 1 8 6 8 .
Schroeder, L. C , D. G. Boggs, G. D o m e , I. M. Halberstam, W. L. Jones, W. J. Pierson, and F. J.
Wentz (1982). The relationship between wind vector and normalized radar cross section used to
derive S E A S A T - A satellite scatterometer winds. / . Geophys. Res., 87, 3 3 1 8 - 3 3 3 6 .
Smith, Ε. Α., and D . R. Phillips (1972). Automated cloud tracking using precisely aligned digital ATS
pictures. IEEE Trans. Computers, C-21,715-729.
Smith, E. A. (1975). The McIDAS system. IEEE Trans. Geosci., GE-13, 1 2 3 - 1 2 8 .
Stewart, R. H. (1985). Methods of Satellite Oceanography. Univ. California Press, Berkeley.
Stewart, T. R., C. M. Hayden, and W. L. Smith (1985). A note on water-vapor wind tracking using
VAS data on McIDAS. Bull. Am. Meteor. Soc, 66, 1 1 1 1 - 1 1 1 5 .
Stogryn, A. (1967). The apparent temperature of the sea at microwave frequencies. IEEE Trans.
Antennas Propag., AP-15, 2 7 8 - 2 8 6 . S u c h m a n , D., and D . W. Martin (1976). Wind sets from SMS
images: An assessment of quality for G A T E . / . Appl. Meteor., 1 5 , 1 2 6 5 - 1 2 7 8 . S z e j w a c h , G. (1982).
Determination of semi-transparent cirrus cloud temperature from infrared radiances: Application
to METEOSAT. / . Appl. Meteor., 21, 3 8 4 - 3 9 3 .
Thomasell, Α., Jr. (1979). Wind Analysis by Conditional Relaxation. N O A A Tech. Rep. NESS 77,
Washington, D C .
Velden, C. S., and W. L. Smith (1983). Monitoring tropical cyclone evolution with N O A A satellite
microwave observations. / . Climate Appl. Meteor., 22, 7 1 4 - 7 2 4 .
Velden, C. S., W. L. Smith, and M. Mayfield (1984). Applications of VAS and TOVS to tropical
cyclones. Bull. Am. Meteor. Soc, 6 5 , 1 0 5 9 - 1 0 6 7 .
Whitney, L. F. (1983). International comparison of satellite w i n d s — a n update. Adv. Space Res.,
2, 7 3 - 7 7 .
Wilheit, Τ. T. (1979). A Model for the Microwave Emissivity of the Ocean's Surface as a Function
of Wind Speed. N A S A Tech. M e m o . 8 0 2 7 8 , Goddard Space Flight Center, Greenbelt, M D .
258 Chapter 7 Winds
Wilheit, Τ. T. and A. T. C. Chang (1980). An algorithm for retrieval of ocean surface and atmosphere
parameters from the observations of the scanning multichannel microwave radiometer. Radio Sei.,
15, 5 2 5 - 5 4 4 .
Wilheit, Τ. T., J. A. Greaves, J. A. Gatlin, D. Han, M. Krupp, A. S. Milman, and E. S. Chang (1984).
Retrieval of ocean surface parameters from the scanning multifrequency microwave radiometer
(SMMR) on the Nimbus 7 satellite. IEEE Trans. Geosci. Remote Sensing, GE-22, 1 3 3 - 1 4 3 .
Woiceshyn, P. M., M. G. Wurtele, D. H. Boggs, L. F. McGoldrick, and S. Peteherych (1986). The
necessity for a new parameterization of an empirical model for wind/ocean scatterometry. / .
Geophys. Res., 9 1 , 2 2 7 3 - 2 2 8 8 .
Wolf, D. E., D. J. Hall, and R. M. Endlich (1977). Experiments in automatic cloud tracking using
SMS-GOES data. / . Appl. Meteor., 16, 1 2 1 9 - 1 2 3 0 .
8
Clouds
and
Aerosols
C LOUDS COVER ROUGHLY half the Earth, and aerosols are always pres-
ent. They constitute major factors in determining the radiation budget
of Earth, and they play crucial, if as yet not completely understood, roles in
modulating the climate. They also contaminate radiometric observations necessary
for sounding the atmosphere and remotely sensing surface parameters. Space-
based instruments are the only means by which the global distribution of clouds
and aerosols can be adequately sampled.
In the first part of the chapter, we discuss methods for determining cloud
parameters from three sources: sounders, imagers, and microwave radiometers.
In the second part of the chapter, we detail methods for retrieving stratospheric
and tropospheric aerosols.
Many cloud properties are of interest to the meteorologist. The ones most
often retrieved are cloud-top temperature, cloud-top height (or pressure), and
cloud amount (the fractional area covered by cloud). Useful also are reflectance,
259
Chapter 8 Clouds and Aerosols
260
emittance, optical depth, phase (ice or water), liquid water content, and drop size
distribution. Attempts to retrieve cloud properties can be divided into schemes
that use sounding instruments, schemes that use visible and infrared imaging
instruments, and schemes that use microwave instruments.
Most efforts to detect clouds using sounder data have been directed toward
removing cloud contamination so that accurate, clear-column radiances can be
calculated. It is also desirable, however, to retrieve the properties of the clouds.
If we assume that clouds reflect little radiation at the wavelengths used by
sounders, then the radiation sensed by a radiometer can be written as the sum
1
of three terms:
L(k) = (1 - N)L (k) dr + NsL (k) dd + N(l - s)L (k), dl (8.1)
where L(\) is the measured radiance of a scene, L (X) is the radiance from the clr
scene if clouds were not present (the clear-column radiance), L (\) is the radiance cld
that would come from the cloud if it were black, Ν is the cloud amount, ε is the
cloud emittance, and 1 - ε is the cloud transmittance. The three terms represent,
respectively, the radiation from the clear area, the radiation emitted by the cloud,
and the radiation transmitted through the cloud. Combining the first and third
terms results in
L(k) = (1 - N')L (k) ck + N'L (k),dd (8.2)
where N' = Νε is called the effective cloud amount. Equation 8.2 is the basis 1
The basis for determining clear-column radiances was first described by Smith
(1968). It applies Eq. 8.2 to two "adjacent" scan spots:
L,(X) = (1 - N[)L (k) dl + N[L (k), dd (8.3a)
L (k)
2 = (1 - N' )L (k) 2 ck + N' L (k),
2 dd (8.3b)
(A tilde indicates an observed quantity.) The scan spots do not have to be strictly
adjacent, but they have to be close enough that the same values of L (X), L (X), dr cld
and ε (λ) describe each spot. This means that (1) the clouds have basically the
same height and microphysical composition in both spots, and (2) the land surface
Chahine (1982) treats the case of channels in the 4.3-μ,ηι band in which clouds reflect some
1
solar radiation.
The reader should be aware that the work of Harshvardhan and Weinman (1982), Weinman
2
and Harshvardhan (1982), and others calls into question whether this approximation, linear in N ' ,
is adequate in a field of broken clouds.
8.1 Clouds from Sounders 261
and overlying clear atmosphere are indistinguishable in both spots. The only
difference between the scan spots is assumed to be the cloud amount N.
Strictly speaking, Eqs. 8.3 describe only a single layer of clouds. If there is
only one layer of clouds, Ν is independent of wavelength. If there are two cloud
layers, Ν will seem to decrease as the height of the peak of the weighting function
increases because channels that peak higher in the atmosphere are less sensitive
to lower clouds (McMillin and Dean, 1982). Terms can be added to Eq. 8.2 to
account for additional cloud levels (Chahine, 1977), but multiple cloud layers
will not be treated here.
Assuming only one cloud layer, L (X) can be eliminated between Eqs. 8.3 to
dd
, % LAX) - N * L ( X ) 2
!_ * Ν > (8-4)
N* =— = =ΑΛ (8.5)
N'2 L (X) -
clr L (k)
2
If N[ Φ N' , an estimate of N * plus observations Li(X) and L (k) will yield L (X).
2 2 clr
There are several ways to estimate N*. Chahine (1974) treats N* as an addi-
tional unknown to be retrieved iteratively in the same scheme used to retrieve
temperatures. This technique is useful when the sounder makes measurements in
a single band, such as the 15-/xm band.
If an infrared sounder makes measurements in both the 15- and 4.3-μπι bands,
the difference in response to clouds in the two bands can be used to estimate
N* (Chahine, 1974; McMillin and Dean, 1982). First, a regression equation is
developed, based on synthesized (not observed) radiances, to estimate the clear
radiance in a relatively transparent 4.3-μηι channel (wavelength λ ) in terms of 4 3
£ lr(^4.3) = 00 +
C Σ <*iLc\r(Kh (8.6)
(A caret over a variable indicates that it has been estimated using other variables.)
Second, measured radiances in two adjacent scan spots are used to calculate the
sensitivity parameters
SlXJ-.^'-fo''. (8.7)
^l(^4.3) ^2(^4.3)
3
N o t e that McMillin and Dean (1982) use brightness temperatures rather than radiances in this
equation. Their scheme is more accurate than the one outlined here, but this scheme contains the
essential physics and is easier to explain. Also note that 4.3-μπι radiances must be corrected for
reflected solar radiation.
2^2 Chapter 8 Clouds and Aerosols
If a single layer of clouds affects the two scan spots, Eqs. 8.3 imply that the
4
Therefore, the sensitivity parameters can be used in combination with one of the
scan spots to relate the radiance in the 4.3-μπι channel to the radiance in each
15-μπι channel:
L(X,) = £(λ,·) + S(X,)[L(X . ) - £ (λ . )].
2 4 3 2 4 3 (8.9)
In particular, the clear radiances are related:
L (X )
dr t = L (\i)
2 + S(X )[L (k )
t dr 43 - L (X . )]. 2 4 3 (8.10)
Eliminating L (X,-) between Eqs. 8.6 and 8.10 yields an estimate of the clear
dr
£ ^ 4 3) = * ° + Σ
* ' ^ ^ (8.11)
431
ι-Σ^(λ,)
Finally, N* is given by
^ c l r ( ^ 4 . 3 ) ~~ ^ 2 ( ^ 4 . 3 )
L^usm) i+ Σ =b 0 b L (X ).
t dr t (8.13)
Only clear radiances are used to develop Eq. 8.13. When the 15-μπι channels are
fy^Msm)-
contaminated by clouds, L(X ) will be less than the observed value,
£ i ( ^ m s u 2 ) and L i ( X
MSU2
N* = ^(^MSU2) ~ ^l(^MSU2) ^
^(^Msm) ^(^-MSin)
4
The cloud microphysics must be the same in the two spots so that ε^λ,-) = ε (λ,·) = ε(λ,) and 2
ει(λ . ) = ε (λ .3) =
4 3 2 4 ε(λ . ).
4 3
8.1 Clouds from Sounders 263
The cloud filter used operationally to process TOVS soundings from NOAA
satellites is described by McMillin and Dean (1982). Since no one test or technique
5
is sufficient to detect and correct for clouds in all situations, several empirically
determined tests are employed. These are briefly mentioned here. Recall that
TOVS soundings are processed in blocks of 7 x 9 or 3 x 3 HIRS/2 scan spots.
The first task is to determine whether any of the scan spots are clear. Seven tests
are applied:
1. The frozen sea test uses MSU channel 1 to determine whether the ocean
is ice covered. A cold brightness temperature (<236 K) indicates water,
not ice. If the surface is water, and the HIRS/2 channel 8 (11 μηι) bright-
ness temperature is less than 270 Κ (the temperature at which sea water
freezes), the scan spot is assumed to be cloud covered.
2. In the adjacent spot test, the HIRS/2 channel 8 brightness temperature of
a scan spot is compared with its eight neighbors. If the center spot is
more than 1.5 Κ colder than any of the surrounding spots, the spot is as-
sumed to be cloudy.
3. The albedo test is designed to detect low clouds, which are difficult to de-
tect with infrared observations. The HIRS/2 channel 20 albedo is com-
pared with an expected value for that location. If the albedo exceeds the
expected albedo by more than a threshold value, the spot is cloudy. This
test can be applied only during daylight hours, obviously, and it is not
applied if the surface temperature is thought to be below freezing so that
ice and snow will not be mistaken for clouds.
4. The window channel tests compare brightness temperatures in HIRS/2
channels 8, 18 (4.0 μηι) and 19 (3.7 μηι). If they differ by too great an
amount, the spot is cloudy. Because of reflected solar radiation, the dif-
6
ference threshold must be greater (and the test is therefore less sensitive)
during the day than at night.
5. In the interchannel regression test, HIRS/2 channel 7 (13.4 μηι) and
MSU channels 2 and 3 are estimated by regression equations from the
other channels. If the estimated and observed brightness temperatures dif-
fer by too great an amount, the spot is assumed to be cloudy.
6. The surface temperature test is designed to detect clouds by estimating a
surface temperature from the measured radiances and comparing with
yesterday's retrieved surface temperature, a shelter temperature, or the
sea surface temperature. If the estimated surface temperature is signifi-
cantly colder than the expected surface temperature, the spot is cloudy.
7. The maximum value test classifies as cloudy any scan spot whose HIRS/2
channel 8 brightness temperature is more than 4 Κ colder than the maxi-
mum channel 8 brightness temperature for the block.
5
The much simpler cloud filter used for VAS soundings is discussed in Section 6.3.2.
6
Several empirically determined thresholds are used. The reader is referred to McMillin and Dean
(1982) for the details.
264 Chapter 8 Clouds and Aerosols
These tests do a good job of detecting clouds, but they are not perfect. Clouds
over snow can go undetected, as can low clouds at night. Also, areas with very
low cloud amount or thin cirrus are always a problem. The tests can err in the
opposite direction by classifying snow-covered areas and areas behind cold fronts
as cloudy. If four or more spots in a block pass all tests, their radiances are
averaged and returned as clear-column radiances.
If enough clear scan spots are not found, the N * technique is attempted.
Estimates using both the 4.3/15-μπι method ( N ) and the microwave method H
(NM) are made. A large number of tests are performed during the calculations
to ensure that an accurate N * is produced. Among the tests are the following: 7
• An indication that multiple cloud layers are present (and therefore that
the N * technique is inappropriate) is that the clear radiances are too cold.
A test to eliminate these cases is made.
• Estimates of N * must yield reasonable values within empirically deter-
mined limits. Different tests are applied to N and to N ^ . If only one of H
N» _ 1 Γ Nh , NM
(8.15)
Methods for retrieving cloud amount and cloud-top pressure from sounder
data may be roughly divided into two related classes: radiance residual methods
7
Details may be found in McMillin and Dean (1982), page 1010.
8.1 Clouds from Sounders 265
and radiance ratio methods. Radiance residual methods are applied after a sound-
ing has been retrieved, and therefore, the clear-column radiances and profiles of
temperature and moisture are known. Several papers discuss the development of
these methods (e.g. Smith et al., 1970; Smith and Woolf, 1976; Chahine, 1975,
1982). We present a simplified version of the procedure developed by Chahine
(1975).
1. Choose a set of possible cloud-top pressures, which range from a level
above the surface to the tropopause.
2. For each p in the set, calculate L for a channel that peaks near the sur-
c cld
face: 8
L (X) = B (T )r (p ,0)
cld xB (T)^^dp.
c x c (8.16) + x
dp j
o
Clouds are treated as black because the emittance is included in the effec-
tive cloud amount.
3. For each p , calculate N ' using the just-calculated L , the value of L
c cld clr
N' = T
L
~_ ; L lr
. (8.17)
^cld ^clr
Assuming one cloud level and channels close enough together so that
ε(λ,) is constant, N is independent of channel.
r
4. For the remainder of the channels (λ,·) that sense clouds, use Eq. 8.16 to
calculate L ^ X , ) , then use Eq. 8.2 to calculate £(λ,·). This step produces
a set of radiances {L(X,)} for each p . c
5. Compare the calculated radiances with the observed radiances. The value
of p that minimizes a measure of the radiance residuals, perhaps
c
For simplicity, p has been used as the vertical coordinate, though lnp is more often used.
2^ Chapter 8 Clouds and Aerosols
If the two channels are close enough together so that ε(λ ) = ε(λ ), then Ν' will λ 2
in p and only weakly dependent on the atmospheric temperature profile (Fig. 8.1).
c
To retrieve p for a particular scan spot, one calculates the left-hand side of
c
Eq. 8.20 and finds the level at which the radiance ratio agrees with /"(λ λ ,ρ ). ΐ5 2 ς
Ν' is then calculated from Eq. 8.19. The procedure for doing these calculations
depends on what information is available. Three approaches have been utilized:
• If a sounding has just been retrieved from the radiances, L will be clr
known for both channels, and /"(λ λ ,ρ ) can be calculated from Eqs. ΐ3 2 ς
cast sounding, or even from climatology, and the values of L can be ei- clr
looo 1 1
1 1 ι ι I
0.2 0.3 0.4 0.5 0.6 0.7 0.8
f
<λ . λ 2 2 ι p ) c
FIGURE 8.1. /"(λ ,λ ,ρ ) for tropical, mid-latitude and polar temperature profiles. H I R S / 1 channels 6
1 2 ε
( \ = 13.35 μπι) and 7 ( λ = 13.65 μπι) were used. The discontinuity in the polar plot is caused by
x 2
ther calculated from the sounding or obtained from a nearby (clear) scan
spot (Smith and Piatt, 1978; Menzel et al, 1983).
• If it can be assumed that p , L ^ X j ) , and L (X ) are constant in a local
c dr 2
area, but that N' changes, then a plot of L(Xj) versus L(X ) will be a 2
The radiance ratio method is the basis of C 0 slicing, which Menzel et al.
2
for the HIRS/1 instrument (which is similar to the HIRS/2 instrument that flies
on the NOAA satellites) were calculated for 10 cloud amounts (0.1-1.0), 12
cloud-top pressures (920-100 hPa), and 2 climatological mean profiles (tropical
and midlatitude). Cloud-top pressure and effective cloud amount were retrieved
using both the radiance ratio (Smith and Piatt, 1978) and the radiance residual
(Chahine, 1975) techniques. The retrieved parameters were compared with the
parameters used to calculate the radiances.
Wielicki and Coakley first investigated errors resulting from instrument noise.
In these calculations, temperature and humidity profiles (and thus L , L , and d r d d
added to the synthesized instrument observations. Figure 8.2 shows the bias and
standard deviation in cloud amount and cloud-top pressure retrieved by the
radiance ratio method using synthesized HIRS/1 channels 6 and 7 (13.35 and
13.65-μπι) radiances. (Note these results depend strongly on the channels used.
Other channel combinations gave better results.) As expected, the method has
trouble with low cloud and with small cloud amounts because the clear-column
radiances are not very different from the observed radiances. The radiance residual
technique produces similar biases, but standard deviations are reduced aloft (by
50% at 200 hPa), due to the addition of channels that better sense high clouds.
Unfortunately, the additional channels slightly degrade retrieval accuracy at low
levels because the channels do not sense low cloud, yet the technique attempts
to minimize their radiance residuals. In other words, the additional channels add
noise, but no information, to low cloud retrievals.
Bias errors are uncomfortably large. Wielicki and Coakley found that nearly
all of the bias error can be explained by nonlinear constraints placed on the
retrievals. These constraints consist of requiring cloud amounts to be in the range
0.0-1.0 and cloud-top pressures to be between 920 hPa and the tropopause
pressure. When a value is out of range, it is reassigned to the closest point in the
range, except that when p is greater than 920 hPa, clear sky is assumed. In the
c
approaches the noise level of the instrument, either because the clouds are lower
or the cloud amount is smaller, the distribution of retrieved cloud-top pressures
268 Chapter 8 Clouds and Aerosols
STD CLOUD PRESSURE ERROR (hPa) BIAS CLOUD PRESSURE ERROR (hPa)
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
CLOUD AMOUNT CLOUD AMOUNT
200
_ 300
o_
β 400
LU
DC
Z) DC
CO => 500
CO CO
CO
w UJ
600
DC DC
θ- CL
α Q
700
=) Z>
Ο Ο
ϋ 800
900
1000 1000
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
CLOUD AMOUNT CLOUD AMOUNT
FIGURE 8.2. Biases (retrieved minus actual) and standard deviations resulting from instrument noise
in retrieved cloud amount and cloud-top pressure (hPa) using the radiance ratio method on synthesized
HIRS/1 channels 6 and 7 (13.35 and 13.64 μπι) radiances. The rms error is the square root of the
sum of the squares of the bias and standard deviation. [After Wielicki and Coakley (1981).]
broadens (i.e., the standard deviation of cloud-top pressure error increases). For
scenes with true cloud-top pressures less than 920 hPa, instrument noise causes
a fraction of the retrieved pressures to be greater than 920 hPa and thus causes
the scene to be misclassified as clear. Removal of this lower tail of the distribution
results in a bias toward cloud-top pressures that are too low. High clouds are
similarly biased toward cloud-top pressures that are too high. Instrument noise
also produces a distribution in retrieved cloud amounts. If the true cloud amount
is near zero, some of the retrievals will be less than zero and will be assigned to
8.2 Clouds from Imagers 269
zero cloud amount. When the cloud amount bias is calculated, this results in a
high bias for low cloud amounts. Similarly, too few scenes are classified as com-
pletely cloudy. Misclassification of scenes covered with low cloud as clear causes
cloud amounts in the lower troposphere to be biased too low. Wielicki and
Coakley tested modifications to the constraints, such as using a visible channel
to detect low cloud and estimating clear skies when L - L is less than twice
clr
Most of the work on cloud estimation has been done using window channels
on imaging instruments. These instruments have the advantage of much higher
spatial resolution than sounders. Perhaps the most often used technique for cloud
characterization is simple manual inspection of satellite photographs or video
images. These analyses are used daily for weather analysis and forecasting and
in the construction of nepbanalyses, which are maps showing chiefly cloud type
and amount at various atmospheric levels (Fye, 1978). Some of these techniques
were discussed in Chapter 5. In this chapter we concentrate on objective tech-
niques.
Shenk and Salomonson (1972) investigated the effects of sensor spatial resolu-
tion on determination of cloud amount using the threshold technique. Consider
an ideal scene in which only a uniform background (radiance L ) and uniform clr
cloud (radiance L ) appear. Pixels that are partly filled with cloud will have
cld
cloud in a pixel will cause it to appear 100% cloudy, which will result in an
overestimate of cloud amount. Figure 8.3a shows cloud amount retrieved from
simulated satellite data as a function of the ratio (R) of the mean cloud area to
the area of a satellite scan spot. The line labels are the true cloud amount. For
R = 100 (cloud diameter approximately 10 times the diameter of a scan spot)
and beyond, there is good agreement between the retrieved and true cloud
amounts. For clouds approximately the same size as the scan spot, however, a
scene with 40% true cloud amount appears to be completely overcast. Alternately,
if the threshold is at nearly L , only completely cloudy pixels will be classified
cld
as cloudy, partly cloudy pixels will be classified as clear, and cloud amount will
be underestimated.
~100
- 90
80%
£ 80
70%
Ο 70
2 60
*0%
50%
Ο 50
401
Ο 40
ο 30%
w 30
I
20%
20
1 0 10% TRUE C10U0 COVCH
10'
1
10° 10' 10 2
10 J
between the thresholds being counted as 5 0 % cloud covered. [After Shenk and Salomonson (1972).]
8.2 Clouds from Imagers 271
A less biased estimate of cloud amount can be obtained by setting two thresh-
olds, one at nearly L and the other at nearly L . Pixels that are between the
clr cld
two thresholds are counted as 50% cloud covered. Figure 8.3b shows the effects
of this scheme. The large positive bias in cloud amount has disappeared, and for
R greater than about 10 (cloud diameter more than VlÖ times the scan spot
diameter), reasonable cloud amounts are retrieved. Obviously, high-spatial-resolu-
tion radiometers are desirable for cloud detection.
A related problem, which has not as yet received adequate treatment in satellite
meteorology, concerns the estimation of cloud amount in a field of broken clouds.
Because broken clouds radiatively interact with each other, and because for nonna-
dir viewing the satellite sees the sides as well as the tops of the clouds, effective
cloud amount in fields of broken clouds is not necessarily proportional to the
fractional area covered with cloud (Harshvardhan and Weinman, 1982; Weinman
and Harshvardhan, 1982).
Setting the threshold is the chief difficulty in threshold techniques. The problem
is that the threshold is a function of many variables: surface type (land, ocean,
ice), surface conditions (snowcover, vegetation, soil moisture), recent weather
(which changes surface temperature and reflectance), atmospheric conditions
(haze, temperature inversions), season, time of day, and even satel-
lite-Earth-sun geometry (bidirectional reflectance, sun glint). In the past, thresh-
olds have been set manually; that is, an analyst views each image to be processed
and picks an area that appears clear. Pixels that are a few percent brighter or a
few degrees colder than the background are classified as cloud-covered. Manual
setting of the threshold is, however, incompatible with automated processing.
Today, thresholds are more likely to be set by examining the temporal or
spatial variance of observed radiances. Spatial variance methods will be discussed
below. Here we discuss the temporal variance approach.
Determination of clear-sky radiances is critical to the success of the threshold
technique because clouds are assumed to perturb the clear-sky radiances. Suppose
that images for the last several (—15) days are analyzed to select the minimum
visible albedo and maximum infrared radiance during the period. Except in 9
relatively unusual circumstances, such as regions which are cloud-covered for the
entire period, regions in which snow has fallen, or regions that experience large
day-to-day variations in surface temperature, these extremes come close to clear-
sky values. Spatial smoothing of the results will reduce noise caused by choosing
extreme values, but care must be taken in high gradient areas such as near
coastlines. Minnis and Harrison (1984) have used a sophisticated version of
this approach.
It must be noted that in some situations, it is difficult to detect clouds with
visible and 11 -μπι data using thresholding (or other) techniques. Low clouds over
ice or snow, for example, present no contrast in either visible imagery, because
9
Bidirectional reflectance and limb darkening corrections should be made to the data before
this analysis.
Chapter 8 Clouds and Aerosols
both ice and snow have similar albedos, or in infrared imagery, because both
have similar temperatures (see Chapter 5). Cirrus is generally easier to detect with
infrared than with visible data, yet cirrus can be thin enough to be warmer than
an infrared threshold. Small broken clouds can be widely scattered enough to
escape detection. Finally, multiple levels of cloud present problems for all cloud
retrieval techniques because of obscuration of the lower clouds by the upper
clouds and because some pixels will contain both levels and perhaps the surface
as well.
Except in combination with radiative transfer techniques (see below), threshold
techniques generally are limited to determining cloud amount (by calculating the
fraction of pixels that are cloudy) and to estimating the cloud-top height by
comparing the infrared equivalent blackbody temperature with a sounding.
histogram of visible and 11-μπι infrared radiances. Tall convective clouds are
cold and bright; they form a cluster in the upper right of the histogram. The sea
surface is warm and dark; it forms a cluster in the lower left corner. Other cloud
and surface types are noted in the figure. Since some pixels sample two or more
cloud or surface types (partly cloudy pixels, for example), the histogram does not
consist of isolated islands, but rather in connected islands. Unfortunately, some
histograms are much more difficult to interpret than this one in part because some
cloud or surface types are too variable to form a local maximum in the histogram.
Setting a threshold for the detection of clouds is equivalent to drawing a vertical
or horizontal line in Fig. 8.4. As expected due to small temperature differences,
it would be difficult to set an infrared threshold (draw a horizontal line) that
distinguishes low cumulus clouds from land or sea. It is interesting to note that
it would also be difficult to set a visible threshold (draw a vertical line) that
successfully distinguishes cloud from land. This is in part because the Meteosat
image from which the histogram was formed included the bright Saharan sands,
but it illustrates the care that must be taken in setting thresholds.
Histogram techniques attempt to locate clusters of pixels as opposed to drawing
vertical or horizontal lines. There are several methods for identifying clusters.
The dynamic cluster method (Desbois etal. 1982; cf. Simmer etal., 1982; Phulpin
et al., 1983) randomly chooses a maximum number of points in the histogram
(roughly 15) to serve as cluster centers. (The coordinates of each point are ( C l5
/.) Each point in the histogram is assigned to the closest cluster. Since randomly
selected clusters are likely to be misplaced, the next step results in a correction.
COLD
* TOPS OF BIG CONVECTIVE
W A R M ^
DARK VISIBLE BRIGHT
FIGURE 8.4. Two-dimensional histogram (visible and infrared) of Meteosat pixel counts. [After Desbois
et al. (1982).]
The center of gravity (using the number of observed pixels at each point as a
weight) and standard deviations (about each axis) are calculated for each cluster.
The center of gravity serves as the new cluster center. Points are again assigned
to the clusters, this time using the standard deviations as distance measures. Any
cluster that receives too few points is eliminated and its points are distributed
among the remaining clusters. The correction step is repeated until the cluster
centers and their standard deviations do not change.
After cluster centers are identified in the histogram, they must be classified.
The clusters can be identified manually, or by setting thresholds, or by comparing
with previous classifications. Once surface clusters have been identified, cloud
amount can be determined by dividing the number of pixels that belong to cloudy
clusters (nonsurface clusters) by the total number of pixels. Cloud temperature
and brightness can be estimated from the average temperature and brightness of
each cloudy cluster. Note that cloud amount defined in this manner applies to
an area, not to an individual pixel, which is considered to be completely clear (if
it belongs to a clear cluster) or completely cloudy (if it belongs to a cloudy cluster).
274 Chapter 8 Clouds and Aerosols
For some applications, it is important to know the type of cloud (or surface).
Pattern recognition techniques attempt to classify arrays of pixels in a manner
analogous to the way a person might perform such a classification.
Consider the following simple hypothetical example. Suppose that an analyst,
armed with a visible satellite image of an oceanic area, were charged with the
task of determining whether an observer on a ship would report clear, partly
cloudy, or cloudy skies. The analyst would examine a small area of the image
centered on the location of the ship. To be consistent, the size of the area should
be approximately the same as that which would be seen by the surface observer.
If the area in the image were uniform and bright, the analyst would classify it as
cloudy. If it were uniform and dark, the analyst would classify it as clear. If a
portion of it were bright and a portion dark, the analyst would classify it as
partly cloudy.
A simple pattern recognition algorithm might simulate this process as follows.
Suppose that the mean and standard deviation of the pixels in the area were
calculated. Uniform scenes, either clear or cloudy, would have a low standard
deviation. Partly cloudy scenes must have a high standard deviation. Cloudy
scenes would have high mean brightness, whereas clear (ocean) scenes would
have low mean brightness. Partly cloudy scenes would have intermediate values
of mean brightness. The algorithm could first check the standard deviation. If it
were sufficiently high, the scene would be classified as partly cloudy. For low-
standard-deviation scenes, the mean brightness would discriminate between clear
and cloudy cases. Note that in pattern recognition, measures of the spatial variabil-
ity of pixels in a scene (such as standard deviation) are referred to as texture
parameters. Spectral parameter is the term sometimes used to describe nontexture
parameters such as mean or extreme radiances or differences of these quantities
between channels.
A training data set is necessary to determine the discrimination values used in
pattern recognition techniques. In the above simple example, an analyst would
classify a set of scenes as clear, cloudy, or partly cloud; then the mean brightnesses
and standard deviations for each classification would be examined in order to
select threshold values.
Harris and Barrett (1978) used mean brightness and two texture parameters
(including standard deviation) to classify 5 x 5-pixel scenes in DMSP visible
images as (1) clear, (2) stratiform cloud, (3) cumuliform cloud, or (4) mixed cloud.
Ebert (1987) attempted the difficult task of classifying cloud and surface types
over the Arctic using four-channel AVHRR data (0.6, 0.9, 3.7, and 11 μτη). She
used eight parameters (five spectral and three texture) to classify 32 x 32-pixel
scenes into 18 cloud, surface, and mixed cloud-surface categories. The maximum
likelihood algorithm (Duda and Hart, 1973) was similar to that used in histogram
techniques in that scenes were assumed to form clusters in the eight-dimensional
parameter space. Means and variances for each cluster were calculated. Standard
deviation was used to measure the "distance" of a scene to each cluster. The
8.2 Clouds from Imagers 275
scene was assigned to the "closest" cluster. A panel of three analysts provided
the training and verification data sets. The classification scheme agreed with the
analysts more than 80% of the time.
the clouds and the surface, respectively, and are assumed known; and £ is s u n
the solar irradiance (integrated over the radiometer spectral response), which is
assumed to be known and to be the same at the surface and at cloud top. The
radiance observed in the infrared channel is similarly modeled as being emitted by
the surface (through the holes in the cloud), emitted by the cloud, and transmitted
through the cloud after being emitted by the surface:
L I R = (1 - N)L dr + NeL M + N(l - e)L , clr (8.22)
where ε is the cloud emittance (assumed known), L is the radiance the cloud cld
would have if it were black, and L is the radiance of the clear area (also assumed
d r
N = L y i s - y d E ( 8 > 2 3 a )
Weld ~ 7clr)^sun
_ L IR - (1 - N e ) L dr
L
dd " ^ - · ( 8
- 2 3 b
)
After L has been retrieved, the cloud-top temperature is calculated using the
d d
Planck function. (Of course, if ε < 1, the temperature will represent the tempera-
ture of the interior of the cloud rather than the cloud-top temperature.) The cloud-
top height may be retrieved by finding the height in an appropriate sounding
where the temperature equals the cloud-top temperature.
In practice, retrievals based on single pixels were shown to be rather noisy.
An error analysis showed that averaging cloud height and cloud amount over
approximately 50 pixels would produce useful results.
276 Chapter 8 Clouds and Aerosols
Reynolds and Vonder Haar tested the bispectral technique using Scanning
Radiometer data from the NOAA 2 satellite. Three rawinsonde stations were
chosen for analysis: Denver, Oklahoma City, and White Sands Missile Range.
Cloud amount and cloud height were calculated and averaged for each pixel
within 75 km of the station. This area approximated the area viewed by a ground-
based observer. Between 13 and 15 satellite passes were analyzed for each station.
The retrieved parameters were compared with surface observations. The cloud
amount rms error was 0.2, and the bias was - 0 . 0 5 . Cirrus caused problems in
cloud height retrieval, in part because cloud emittance was assumed to be 0.9 for
all clouds. For all cases, rms cloud height error was 4.7 km with a bias of - 4 . 2
km. When cirrus cases were eliminated, the rms cloud height error was 0.5 km
with a bias of +0.27 km.
A problem with the bispectral technique is that bidirectional reflectance is a
difficult quantity to specify a priori. It is a function of the size, shape, and spacing
of the clouds in addition to the optical depth (McKee and Cox, 1974; Reynolds
et aU 1978; Weinman and Harshvardhan, 1982).
Piatt (1983) explored extensions to the bispectral technique including reflection
between the cloud layer and surface, parameterization of the infrared cloud emit-
tance in terms of the cloud visible optical depth, and multiple cloud layers.
Another multispectral technique allows the retrieval of the temperature of
thin cirrus. Szejwach (1982) showed both theoretically and experimentally
(with an aircraft-mounted radiometer) that cirrus clouds have nearly equal
emittances near 6.4 and 11.5 μπι in spite of the large wavelength separation.
Since collocated scan spots in the two channels will contain the same cloud
amount N, they will have the same effective cloud amount N'. The radiances
in the two channels must, therefore, be linearly related (see Eqs. 8.19 and
8.20); when observed pairs of radiances are plotted (Fig. 8.5), they fall on a
straight line. The point on the line where N' = 1 is the point where the
cloud acts as a blackbody. This point can be found with the aid of the Planck
function by plotting the curve L = B [B2 (L )]
x 1
1
The temperature (T ) where
2 N
the line (L) intersects the curve (Γ) is the temperature of the cloud. This
technique is being used with Meteosat data.
A very precise method for determining cloud amount and cloud-top tempera-
ture is the spatial coherence method of Coakley and Bretherton (1982). The
technique starts with several assumptions that preclude it from being applicable
in all situations but that result in precise estimates in those cases to which it does
apply. In part the precision results from the fact that the technique is capable of
detecting cases to which it is not applicable. Clouds are assumed to exist in layers
with uniform cloud top and emittance (the original technique only dealt with a
single layer of cloud) and to exist over a uniform background. Portions of the
image must be completely clear and portions completely cloudy. The technique
is well suited to analyzing marine stratocumulus, but is applicable in other situa-
8.2 Clouds from Imagers 277
FIGURE 8.5. Meteosat water vapor radiance versus w i n d o w channel radiance over a thin cirrus cloud.
The observed radiances fall on the straight line L. The intersection of this line with the Planck curve
ε gives the cloud-top temperature. [After Szejwach (1982).]
tions as well. It is not applicable, for example, near fronts or when cirrus overlies
the scene.
The spatial coherence technique rests on the idea that portions of the image
that are either completely clear or completely cloudy will exhibit little spatial
variability, whereas partly cloudy areas will exhibit large spatial variability. Infra-
red data are analyzed over a moderately large sector (on the order of (250 km) ). 2
The image is subdivided into blocks of at least 2 x 2 pixels. For each block, the
mean and standard deviation of the radiance are calculated. When plotted (Fig.
8.6), the data form an arch pattern. The feet of the arch (where the spatial
variation of radiance in each block is low) are the blocks that are completely
clear or completely cloudy. The average of the radiances in a foot represents the
radiance of the surface or the cloud top. The width of the foot, as measured by
the standard deviation of the radiances in the foot, is a measure of the uncertainty
in the background or cloud radiance. No other technique yields these important
uncertainty estimates.
Once the clear and cloudy radiances are known, the cloud amount can be
calculated for each pixel using Eq. 8.2. Usually, however, the mean cloud amount
for the entire sector is calculated:
N=r \ L L
T
d l
, (8-24)
^ld ^-clr
where L is the mean radiance in the sector. The uncertainty in cloud amount can
also be calculated. It is composed of instrument noise and uncertainty in L and clr
27g Chapter 8 Clouds and Aerosols
CHANNEL 4
—i 1 1 1 1 1-
-τ ' r
<
>
Ld
Q 3
Ο
α:
<
α
ζ
2 2
_ι
<
Ο
Ο
L . When averaged over the sector, the instrument noise is negligibly small.
cld
ΔΝ : frNAL^l^Rl-NlAL^y" ( 8 2 5 )
1 1
This is based on the assmption that the error of a function y of several independent variables
x is given by Ay
{
:
, where Ax, represents the error in the independent variables.
8.2 Clouds from Imagers 279
CHANNEL 4
20 ι . 1 . 1 1
CHANNEL 1
<
>
ÜJ
Q
Ο
CtL
< 10
Q
Ζ
<
I—
in
ι
<
ο
Ο 5
visible radiances are affected primarily by the cloud optical depth, 1 l-μιτι radiances
are affected primarily by the cloud-top temperature, and 3.7-μηι radiances (during
daylight) are affected primarily by the cloud microphysical parameters (chiefly
drop size and drop phase). They chose six sets (models) of microphysical conditions
to approximate the range of clouds expected. All of the cloud drops were assumed
to be spherical. Three models had ice spheres, and three had liquid water spheres.
The ice models had modal drop radii of 4, 8, and 32 μηι; the water models had
modal drop radii of 4, 8 and 16 μηι. Radiative transfer calculations determined
the reflectance, transmittance, and absorptance of plane-parallel cloud layers as
tabulated functions of optical depth, zenith angle, and azimuth angle at the
three wavelengths. Armed with these numbers, the retrieval schemes proceeds
as follows.
• A histogram of visible reflectance and 11-μηι brightness temperature is
constructed within a 2.5° x 2.5° analysis area, and clusters of points are
identified. One cluster must represent the surface. Mean reflectance, 11-
μπι brightness temperature, and 3.7-μιη radiance are calculated for the
surface cluster and serve as the surface (clear) values.
• In the cloudy cluster, the cloud amount is assumed to be one, and the 3.7-
j L i m radiance is calculated from the surface parameters and the tabulated
functions. Comparing the observed and calculated 3.7-μπι radiances, one
of the six microphysical models can be selected as most representative of
8.2 Clouds from Imagers 281
the clouds in the sector. This step is somewhat subjective, but it narrows
12
The procedure has since been changed to fit a microphysical model to each cluster in the sector
1 2
(A. Arking, personal communication, 1988). This has the effect of allowing more than one type of
cloud (for example cirrus and stratocumulus) to exist in the sector.
282 Chapter 8 Clouds and Aerosols
viewing and illumination angles, (3) cloud-top height, and (4) surface reflectance.
To retrieve the optical depth for a particular pixel, the viewing and illumination
angles are obtained from navigation files, the cloud-top height is obtained from
the IR model, and the surface reflectance is obtained from the monthly mean.
The remaining variable, optical depth, is determined by interpolating between
the calculated radiances that best match the observed visible radiance.
Finally, to account for noise (of all kinds), any pixel whose retrieved optical
depth is less than 1.2 or whose retrieved cloud-top height is less than 1.4 km is
classified as clear. Requiring pixels to pass both tests results in some clouds (e.g.,
some low stratus and some cirrus) escaping detection. On the other hand, few
pixels that are classified as cloudy are actually clear.
It is important to note that this sampling is not equivalent to "throwing away" data. The goal
1 3
from 0.83 to 0.93, and the bias ranged from - 5 . 3 % to +2.7%. All of the methods,
however, had difficulties in "low contrast" situations in which cloud radiance
differed little from clear-sky radiance. Among the other conclusions were (1) that
since all techniques benefited from better assessment of the clear-sky radiance,
the ISCCP algorithm should make clear-sky radiance determination a top priority
and (2) that a threshold technique should be adopted for cloud detection because
it is the best understood, it is the easiest to interpret, and it seems to be more
sensitive to difficult-to-detect clouds than other methods.
The ISCCP algorithm, developed as a result of the intercomparison tests, pro-
ceeds in four steps. In the first step a clear-sky radiance map is constructed. This
step is the most time consuming part of the algorithm. It rests on the observation
that temporal and spatial variations of clear-sky radiances are small compared
to the variations caused by clouds. The IR radiance map is constructed as follows.
First, pixels that are easily identified as cloudy are eliminated. Subregions of
~(500 km) over ocean and ~(250 km) over land are examined. Any pixel that
2 2
is much colder (—10 K) than the warmest pixel in the subregion is labeled cloudy.
These pixels need not be considered further. Next, pixels are compared with the
pixel in the same location on the day before and the day after. If the pixels differ
by less than a small temperature (~1 K, variable with surface type), the pixel is
labeled clear. For pixels that survive this test, the 5-day mean, 5-day maximum,
15-day mean, 15-day maximum, 30-day mean, and 30-day maximum brightness
temperatures are calculated. Finally, these numbers, ranked in order of preference,
are used in the clear-sky radiance map. Several tests, which depend on surface
type, are performed to ensure that the numbers are reasonable. Over land, the
5-day mean is selected unless the tests show it to be cloud contaminated. Testing
continues until, if all else fails, the 30-day maximum brightness temperature is
used. Over ocean, the same process is used, but it starts with the 15-day mean.
A visible clear-sky reflectance map is constructed by a similar, but indepen-
dent, process.
The second step consists in applying thresholds to detect clouds. The thresholds
consist not of absolute values, as in previously mentioned techniques, but of
increments above the clear-sky values. The two thresholds are ΔΤ, an IR brightness
temperature difference, and AR, a visible reflectance difference. These increments
are designed to take into account both instrument noise and uncertainty in the
clear-sky values. Each pixel is assumed to be totally clear or totally cloudy. Any
pixel whose IR brightness temperature is more than 2ΔΤ colder than the clear-
sky brightness temperature or whose visible reflectance is more than 2AR greater
than the clear-sky reflectance is labeled cloudy. Any pixel whose IR brightness
temperature is more than ATbut less than 2ΔΤ colder than the clear-sky brightness
temperature or whose visible reflectance is more than AR but less than 2AR
greater than the clear-sky reflectance is labeled marginally cloudy. Warmer and
darker pixels are labeled clear. Information about both of these two tests is
retained so that users can examine only infrared-detected clouds, or only visible-
detected clouds, or any combination. Over land, ΔΤ and AR are 6 Κ and 0.06,
respectively; over ocean they are 3 Κ and 0.03, respectively. Cloud amounts
284 Chapter 8 Clouds and Aerosols
are determined in the forth step by counting the cloudy pixels in the 2.5° x
2.5° regions. 14
In the third step, cloud radiative properties for each cloudy or marginally
cloudy pixel are determined using the technique of Rossow (1989). By comparing
the observed visible reflectance and infrared brightness temperature with radiative
transfer calculations, the cloud-top temperature and the cloud optical depth (dur-
ing daylight) are determined. Using TOVS temperature profiles, the cloud-top
pressure is determined from the cloud-top temperature. The philosophy here is
that although all pixels are assumed to be either 100% cloud-covered or 0%
cloud-covered, variations in cloud IR temperature and visible reflectance are
explained in a radiatively consistent way as variations in cloud-top temperature
and cloud optical depth.
In the fourth step, the data are histogrammed as functions of optical depth
and cloud-top pressure for each 2.5° x 2.5° region (280-km grid cell). Data are
also histogrammed in terms of a cloud type, which is determined by the cloud-
top pressure and cloud optical thickness (Fig. 8.9). In all, 132 parameters are
recorded for each cell, including parameters that reveal how the classifications
were made. Table 8.1 lists the retrieved parameters. Monthly summaries of the
data are also constructed (Table 8.2). Data tapes of the retrieved ISCCP parameters
are available from the Satellite Data Services Division of NESDIS. Figure 8.10
shows the mean cloud amount retrieved by the ISCCP algorithm for July 1983
and January 1984. Figure 8.11 shows the mean zonal cloud amounts for the
same months.
Actually, the boxes are equal area boxes. They extend 2.5° in the latitudinal direction, but
1 4
approximately 2.5°sec© (Θ is latitude) in the longitudinal direction so that the boxes have approxi-
mately the same area as a 2.5° x 2.5° box at the equator. (See Rossow and Garder, 1984.)
8.2 Clouds from Imagers 285
50
I 1
180
ο.
Ο 560 — OR NIMBOSTRATUS — > MIDDLE
ο ALTO STRATUS
Ο
680
CUMULUS
800 —-
Γ OR STRATUS
- ^LOW
STRATO CUMULUS
Ι 000 l 1 1
1.3 3.6 9.4 23 125
CLOUD OPTICAL THICKNESS
FIGURE 8.9. ISSCP radiometric classification of cloudy pixels by the measured values of optical thickness
and cloud-top pressure. At night only cloud-top pressure is determined, so that only the low, middle,
and high cloud types are counted. [After Rossow and Schiffer (1991).]
that clouds over locations with significant surface temperature inversions can be
detected as dark regions in infrared images. For cold clouds, an attempt was made
to classify the cloud height (low, middle, high) based on a climatological lapse rate.
The ultraviolet algorithm is interesting because both ocean and land have low
ultraviolet reflectivity, whereas land can have visible reflectivities exceeding 40%.
First, a surface reflectivity was estimated from the TOMS-measured reflectivity
using a multiple Raleigh scattering model (Dave, 1964). Two wavelengths were
used, 0.36 and 0.38 μπι, neither of which is absorbed by ozone. The two reflectivi-
ties were averaged to produce a "0.37-μπι" reflectivity. The cloud fraction for a
single TOMS scan spot was modeled as linear with respect to the 0.37-μπι reflecti-
vity: 8% reflectivity corresponds to no clouds; 50% reflectivity corresponds to
100% cloud cover.
The two algorithms differ in their treatment of a single scan spot: in the infrared
algorithm each scan spot is either cloudy or cloud free; in the ultraviolet algorithm
286 Chapter 8 Clouds and Aerosols
a
Global information provided every 3 h for each 280-km grid cell. [After
Rossow and Schiffer (1991).]
8.2 Clouds from Imagers 287
a
Global, monthly average information provided at eight
equally spaced times of day. [After Rossow and Schiffer (1991).]
each spot is partly cloudy. The difference lies in the resolution of the instruments
and in the resolution of the desired product. The goal was to estimate cloudiness
for target areas about 165 km square. The THIR has 20-km resolution at nadir;
thus there are many THIR scan spots per target area. Treating each THIR spot
288 Chapter 8 Clouds and Aerosols
JULY 1983
FIGURE 8.10. Mean cloud amounts retrieved by the ISCCP algorithm for July 1983 and January 1 9 8 4 .
[Courtesy of G. Garrett Campbell, Cooperative Institute for Research In the Atmosphere, Colorado
State University.]
as cloudy or clear results in an estimate of the cloud fraction over the target area.
TOMS, on the other hand, has 50-km resolution at nadir and 150 x 200-km
resolution at the extreme scan positions. Since there were so few TOMS scan
spots per target area, it was desirable to obtain cloud fraction estimates from
each one. Fortunately, the uniformly dark background of the snow- and ice-free
surface, coupled with the nearly linear dependence of reflectivity on cloud fraction,
made single-spot cloud fractions retrievable. The Air Force 3D Nephanalysis
snow/ice cover fields were used to determine whether the ultraviolet algorithm
should be applied.
For nighttime observations, the THIR cloud data alone were used. During
8.2 Clouds from Imagers 289
100
CLOUD AMOUNT
FIGURE 8.11. Zonal mean cloud amounts retrieved by the ISCCP algorithm for July 1983 and January
1984. N o t e that the global mean cloud amount is greater than 6 0 % , indicating that the 5 0 % figure
usually cited may be too low. [Courtesy of G. Garrett Campbell, Cooperative Institute for Research
In the Atmosphere, Colorado State University.]
daylight hours, THIR cloud data and TOMS cloud data were merged in a weighted
averaging scheme. The TOMS cloud amounts were given greater weight if the
THIR estimated no cloud or low cloud because low clouds are difficult to detect
at 11.5, μπι due to the small difference between the clouds and the surface. Low
clouds, however, are good reflectors of visible and ultraviolet light. For cases in
which middle and high cloud dominated, the THIR cloud amounts were given
more weight than the TOMS cloud amounts.
Finally, the TOMS reflectivities were used to identify cirrus and deep convective
clouds. The Nimbus 7 THIR/TOMS cloud climatology is available from the
National Space Science Data Center.
FIGURE 8.12. Simplified two-dimensional diagram illustrating h o w cloud height is observed by two
geostationary satellites. When a G O E S - E a s t image is superimposed on a mapped G O E S - W e s t image,
the separation (parallax) between the location of the same cloud in the two images is proportional
to the cloud height. [After Hasler (1981).]
In the microwave portion of the spectrum, the liquid water content of clouds can
be measured. This quantity is not measurable in other portions of the spectrum.
Recall from Chapter 3 that both water vapor and cloud droplets weakly absorb
microwave radiation and that scattering of microwave radiation is negligible for
non-precipitating drops. For centimeter wavelengths, water vapor absorption peaks
at 22.235 GHz, the center of a weak water vapor rotation line (see Fig. 3.13). Figure
8.13 shows that the absorption by water droplets increases with frequency. The
absorption coefficient for each is proportional to the amount of water (liquid or
vapor) per unit volume. Over the ocean, which provides a uniform, cold background
due to the low emittance of water, atmospheric water vapor and liquid water in-
crease the brightness temperature measured by a satellite-borne radiometer. By ex-
ploiting two frequencies, one near 22.235 GHz and one in the atmospheric window
near 31 GHz, column-integrated water vapor and liquid water content (both in units
of mass per unit area) can be retrieved as follows.
Applying the Rayleigh-Jeans law (Eq. 3.11) to the integrated form of Schwarz-
child's equation (Eq. 6.6), the brightness temperature measured by a microwave
10° ρ
|0" '
4
1 1 I I I I I I I ' ' 1
0 10 20 30 40 50 60
FREQUENCY ( G H z )
FIGURE 8.13. Mass absorption coefficient of cloud drops in the microwave region. [After Westwater
(1972).]
292 Chapter 8 Clouds and Aerosols
radiometer is
Τ (λ,μ) = ε » Τ τ ^ + [1 - e > ) ] T
Β 0 s p a c e T ^ + f ΤΨ (τ,μ)
2 1
λ dr, (8.26)
Ignoring the small second term, removing T from the integral, and integrating the G
third term with the aid of Eq. 6.8 yields (cf. Grody, 1976)
Τ -Τ {1-[1-ε»]τ Η
Β 0
2
(8.27)
and water vapor content ( l / is also called the precipitable water) and ß and ß
v L w
are the mass absorption coefficients for cloud droplets and water vapor, respec-
tively, then
τ2 /μ
0 - exp ^ ( Α Α + £νϋν + δ ; θ2
τ -τ
0 Β
(8.28)
[1 - ε (μ)]Τ
0 0
Most work (e.g., Grody etai, 1980; Wilheit and Chang, 1980) has approximated
T as a constant (280 K) and assumed that ε is a constant. If this equation is applied
0 0
ln[280 - T ]: B
The coefficients can be found by using a more accurate radiative model or (for
I7 , at least) by comparing with observations. It must be noted that independent
V
measurements of liquid water with which to compare satellite estimates are essen-
tially nonexistent; thus the accuracy of the satellite estimates is not known.
The Soviet satellite Kosmos 243 carried the first microwave radiometer capable
of making such measurements (Gurvich and Demin, 1970). U.S. satellites Nimbus
5,6,7, Seasat, and two DMSP satellites have carried instruments from which liquid
water and precipitable water could be retrieved. The Indian satellites Bhaskara 1
and 2 also carried microwave-sensing instruments.
8.4 Stratospheric Aerosols 293
1000 times more abundant. The most frequently used technique is solar occulta-
tion, in which the transmittance of solar radiation is measured during spacecraft
sunrise or sunset. A great advantage of solar occultation is that it is nearly self-
calibrating; unattenuated solar radiation is measured before each sunset or after
each sunrise. Only the ratio of the attenuated to the unattenuated radiation (the
transmittance) is used in the retrieval process. Variations in solar output and in
instrument gain are eliminated.
McCormick et al. (1979) and McCormick (1983) review solar occultation
experiments. The first such experiment was performed aboard Apollo—Soyuz in
1975. The Stratospheric Aerosol Measurement (SAM) experiment consisted of a
hand-held, one-channel sun photometer with which astronauts made measure-
ments during two sunrises and two sunsets. The measurements were centered at
0.83 μπι. The results were sufficiently promising to prompt the development of
two complementary instruments to fly on unmanned satellites.
The Stratospheric Aerosol Measurement II (SAM II) experiment was launched
aboard the Nimbus 7 satellite in October 1978. It also was a one-channel sun
photometer, but its spectral range was centered at 1 μπι, where absorption due
to atmospheric gases is negligible. Only Rayleigh scattering and scattering by
aerosols are important. Since Nimbus 7 had equator crossing times near local
noon and midnight, all of the sunrises and sunsets observed by SAM II were in
polar regions, 64°-80° north latitude and 64°-80° south latitude.
The Stratospheric Aerosol and Gas Experiment I (SAGE I) was launched aboard
a dedicated Applications Explorer Mission satellite (AEM 2) in February 1979.
Its orbit had an inclination of 55°, which caused its orbital plane to precess with
respect to the sun. It made sunrise and sunset measurements between 79°N and
79°S (depending on season), and it sampled all longitudes in about a month. Thus
the SAGE I coverage complemented that of SAM II. About 15 sunrises and 15
sunsets were observed per day for the first 4 months of operation until a problem
with the power supply limited observations to sunsets only. Observations were
made until November 1981. SAGE I made measurements in four bands centered
at 0.385, 0.45, 0.60, and 1.0 μηι. All of these bands are sensitive to Rayleigh
and aerosol scattering. In addition, the 0.6-μιη band is sensitive to ozone ( 0 ) , 3
and the 0.385- and 0.45-μπι bands are sensitive to nitrogen dioxide ( N 0 ) . SAGE
2
to SAGE I except that it has seven channels centered at 1.02, 0.936, 0.600, 0.525,
0.452,0.448, and 0.385 μπι. The 0.936-μπι channel provides the ability to retrieve
stratospheric water vapor.
The basic measurement in solar occultation is transmittance (τ ) through the λ
where 8 {h) is the optical depth along the ray path. If we assume that σ varies
x ε
spatially only with z, the height above the Earth's surface, then σ (λ, ζ) can be ε
σ (λ, ζ) = 1 f ^
ε [(h + r ) - (ζ + r ) ] "
e
2
e
2 1/2
dh, (8.32)
is the extinction due to scattering by aerosols, and σ and σ are the absorption θ 3 Ν Ο ζ
due to ozone and nitrogen dioxide, respectively. Extinction profiles for each of
the components could be retrieved from σ profiles calculated using Eq. 8.32. ε
where Δχ is the length traversed by the ray whose tangent height is h in the
ί; t
σ , and σ
θ 3 · Each constituent is retrieved using measurements in the channel
Ν θ 2
0.60 μπι, c r Q at 0.45 μπι, and cr i h at 0.385 μπι. The procedure begins at
N 2 Ray eig
1.0 μπι. The optical depth due solely to aerosols (the dominant constituent at
this wavelength) is approximated by subtracting the other constituents:
where the superscript (k) notes the kth iteration. Note that the 8 (hj), being directly x
^, 6
| r o s o
' ( / 7
; )
=1. (8.36)
2j & aerosol
i
If not, then each o- (z ) is iterated using the Twomey et al. (1977) modification
aerosol ;
Αν."
^~ aerosol(%j) ^" a e r o s o l ) ΓΊ l + t f f - ! ) ^ , 1
(8.37)
where the product is taken over all tangent heights, and for a given;, Δχ is the /;
maximum of Δχ for all values of /. The process is repeated in turn at 0.6, 0.45,
/;
1 5
The T w o m e y et al. modification uses information at all levels, not just at the peak of the
weighting function. N o t e that δχ^ΙAx^ = 1 at the peak of the weighting function ( Δ χ ) but is non- ζ/
zero at other tangent heights. Since Δχ^ is sharply peaked at the tangent height, Chu and McCormick
perform the calculation only within 4 km of the tangent height.
1 6
The extinction coefficients are wavelength-dependent. To use the aerosol extinction coefficient,
for example, at wavelengths other than 1.0 /xm where it is calculated, the wavelength dependence
must be known. At the other wavelengths, however, the aerosol extinction contributes only a minor
amount to the total extinction; thus the wavelength dependence does not have to be precisely known.
Chu and McCormick use a log-normal size distribution of 7 5 % sulfuric acid droplets at a number
density of 10 c m to approximate the aerosol distribution and derive the wavelength dependence.
- 3
296 Chapter 8 Clouds and Aerosols
10-6 10-5 10-4 10-3 10"6 10"5 10"4 10'3 IQ-6 IQ-5 10"4 10"3
E X T I N C T I O N C O E F F I C I E N T , km-1
FIGURE 8.14. Extinction coefficient profiles derived from SAGE I and SAM II measurements over Alaska
in 1979. The dashed line is extinction due to atmospheric gases ( σ ) ; the solid lines are extinction
0
due to aerosols (cr ). The arrow indicates the height of the tropopause. [After Russell et al. (1984).]
A
SAGE II aerosol data have also been compared to SAM II data (Yue et al,
1989) and to other correlative data sets, in both Europe and the United States,
17
with results similar to those for SAGE I, but at four wavelengths. Figure 8.15
shows typical extinction profiles for the four SAGE II wavelengths. Attempts have
been made to use the four SAGE II wavelengths to retrieve information about
the composition and size distribution of stratospheric aerosols (Livingston and
Russell, 1989; Wang et al, 1989).
Satellite observations have been used to study the injection and dispersal of
volcanic aerosols (Kent and McCormick, 1984) and stratospheric clouds that
form in the winter near the poles (McCormick et al, 1982). In addition, although
tropospheric clouds often block measurements below the tropopause, a surprising
number of retrievals can be made in the middle to upper troposphere (Kent et
al, 1988).
The former Soviet Union experimented with another method, called the Earth's
aureole method, for measuring extinction coefficients in the stratosphere. Avaste
and Keevallik (1983) reported that a four-channel (1.35, 1.9, 2.2, and 2.7 μηι)
radiometer was flown on the Salyut 4 and Salyut 6 space stations in 1975 and
1978, respectively. The radiometer measured scattered radiance as a function of
tangent height in the Earth's limb. Assuming only single scattering, it can be
shown that the measured radiance is given by
L (h)
K = C[l - T (h)], k (8.38)
where T (h) is given by Eq. 8.31. C is a constant:
K
C = ±ώ Ε (λ)[ρ(ψ
0 5υη 51ιη ) + 2Vs«„]- (8-39)
1 7
See the special SAGE II section of the Journal of Geophysical Research, Vol. 9 4 (1989), N o . D 6 .
8.4 Stratospheric Aerosols 297
•0.525 /*m
35
- 0 . 4 5 3 μΓΤΊ
30
HI
§ 2 5
< 20
1.02 /xm
15
10
The first term of C is the radiance due to single scattering of directly illuminated
aerosols: ω is the single scatter albedo, £ (X) is the solar irradiance, and p(i// )
0 sun sun
is the scattering phase function for the angle between the sun, tangent point, and
satellite. The second term of C is the radiance due to single scattering from aerosols
illuminated by radiation reflected from the surface or clouds: A is the albedo
below the tangent point, and μ is the cosine of the solar zenith angle at the
51ιη
d l n T
^ 1 1 U L
(8.40)
dh \L -C) dh'
The extinction profile can be obtained by substituting this expression into Eq. 8.32.
The primary advantage of the Earth's aureole method over solar occultation
is that measurements are not limited to sunrise and sunset events; they can be
made nearly continuously during the sunlit portion of each orbit. They have
essentially the same horizontal resolution as limb soundings. The disadvantage
298 Chapter 8 Clouds and Aerosols
is that one must have knowledge of the single scatter albedo and the surface
or cloud albedo near the tangent point. A secondary advantage is that aureole
measurements can apparently be made at greater altitudes than can occultation
measurements. Avaste and Keevallik report making measurements of noctilucent
clouds at 80-85 km.
Over the ocean, L includes sun glint and reflectance by suspended particles
0
lengths.
^ R a y l e i g h * always present and must be subtracted from L to obtain the radiance
s
due to aerosol scattering. Kaufman (1979) has developed a simple model, based
on the two-stream approximation of Coakley and Chylek (1975), to calculate
the radiance due to Rayleigh scattering.
Assume that the satellite-observed radiance is free of radiation reflected from
the ocean and that it has been corrected for Rayleigh scattering. Assume further
that only single scattering need be considered. Then, by Eq. 3.45, the radiative
transfer equation may be written:
= -L a e r o S oi + ^£su„P(<ku„) exp ( - 6 a e r
°^~ 6 x
), (8.42)
of the entire atmosphere, and μ is the cosine of the satellite viewing angle. The
8.5 Tropospheric Aerosols
299
(8.43)
Since 6 is often much less than one, the satellite-observed radiance can be
aerosol
further approximated as
aerosol
l i
aerosol =
7Z sunP(^sun)
£ (8.44)
elements of Eq. 8.44 are known. Three of the necessary parameters are easily
obtained. £ is a function of time of year, the solar spectrum, and the spectral
s u n
response function of the satellite radiometer. The solar zenith angle is a function
of time and the location of the scan spot. The satellite viewing angle is a function
of the locations of the satellite and the scan spot. The remaining two parameters,
ώ and p(i// ), depend on the index of refraction and size distribution of the
0 sun
aerosols. Thus in contrast with solar occultation, some knowledge of the properties
of the aerosols must be known before optical depth can be retrieved.
Two basic types of aerosols have been studied using satellite data: marine
aerosols (haze) and Saharan dust. Griggs (1975) compared surface-based measure-
ments of marine aerosol optical depth (using a Volz sun photometer) with Landsat
observations at 0.55, 0.65, and 0.75 μπι and found that optical depth could be
retrieved within ± 1 0 % . Griggs (1979) found similar results with GOES 1 and
NOAA 5 Scanning Radiometer data. Fett and Isaacs (1979) and Isaacs (1980)
studied the effects of marine aerosols on DMSP visible imagery. Hindman et al.
(1984) and Durkee et al. (1986) have used AVHRR data and Nimbus 7 Coastal
Zone Color Scanner data to study marine aerosols off the California coast. These
last three studies used the marine aerosol models of Shettle and Fenn (1979).
Durkee et al. (1991) have used AVHRR channels 1 and 2 to retrieve aerosol
optical depth. The combination of two channels lessens the uncertainty caused
by assuming a single scatter albedo and a scattering phase function.
Satellite data were first used to study Saharan dust by Fraser (1976). He used
a radiative transfer model to convert Landsat radiances to vertically integrated
aerosol mass. Carlson and Wendling (1977) and Carlson (1979) were able to
map dust optical depth over the Atlantic using NOAA 3 VHRR data. Norton et
al. (1980) studied Saharan dust using GOES data. They were able to estimate the
dust optical depth to ±0.1 accuracy for optical depths up to 0.4. Recently Deuze
et al. (1988) studied Saharan aerosols over the Mediterranean Sea using Meteosat
and AVHRR data.
In January 1989, NESDIS began producing weekly maps of aerosol optical
depth based on AVHRR channel 1 data (Rao et al., 1988). The maps have a 100-
km grid and include all the world's oceans. The retrieval scheme relies on a lookup
table calculated with a radiative transfer model. The model includes Rayleigh
scattering and ozone absorption as well as aerosol scattering. The table is calcu-
Chapter 8 Clouds and Aerosols
300
lated as a function of solar zenith angle, satellite zenith angle, and the relative
azimuth angle. Given these three parameters, aerosol optical depth can be interpo-
lated on the basis of the observed radiance. To avoid sun glint, retrievals are
made only in the antisolar half of each scan line. Clouds are filtered in two ways.
First, techniques used in sea surface temperature retrievals are applied (McClain
et al., 1985). Second, the difference between the triple-window SST (Eq. 6.73c)
and the split-window SST (Eq. 6.73b) is calculated. Rao et al. have observed that
due to the reflectance of clouds at 3.7 μτη these SST estimates differ by at least
5 Κ in cloudy areas. Thus they use a 5-K threshold to detect and eliminate cloudy
scenes. Figure 8.16 shows an example of aerosol optical depth for the Atlantic
Ocean. Based on comparison with surface measurements (Griggs and Stowe,
1984), it is estimated that the one-channel scheme is capable of retrieving aerosol
optical depth within ±0.03 to ±0.05. If the global background optical depth is
roughly 0.1, the method's accuracy is in the 3 0 - 5 0 % range. In high-signal dust
outbreaks such as that shown in Fig. 8.16, however, the relative accuracy is
much better.
Aerosol measurements over land are also possible if one knows the surface
reflectance. Fräser et al. (1984) have used GOES visible data to estimate summer-
time aerosol optical depth over the eastern United States. They then related the
optical depth to the column-integrated mass density of sulfate particles.
Measurements outside the visible to near infrared portion of the spectrum also
appear to be useful. Ackerman (1989) reports that the difference between the
brightness temperatures at 3.7 and 11 μπι is sensitive to aerosol optical depth.
FIGURE 8.16. Aerosol optical depths (xlOO) for the period 1 8 - 2 5 June 1987. [Courtesy of Larry
Stowe, N O A A . ]
8.5 Tropospheric Aerosols 301
He used this information to plot aerosol optical depth over the Arabian Peninsula.
Also, Lee et al. (1986) report that the difference between brightness temperatures
at 11 and 12 μηι is sensitive to aerosol optical depth.
Bibliography
Ackerman, S. A. (1989). Using the radiative temperature difference at 3.7 μηι and 11 μηι to track
dust outbreaks. Remote Sens. Environ., 27, 1 2 9 - 1 3 3 .
Arfken, G. (1970). Mathematical Methods for Physicists, 2nd ed., Academic Press, N e w York.
Arking, A. (1964). Latitudinal distribution of cloud cover from Tiros III photographs. Science,
143, 5 6 9 - 5 7 2 .
Arking, Α., and J. D . Childs (1985). Retrieval of cloud cover parameters from multispectral satellite
images. / . Climate Appl. Meteor., 2 4 , 3 2 2 - 3 3 3 .
Avaste, Ο. Α., and S. H. Keevallik (1983). Remote monitoring of aerosols from space. Adv. Space
Res., 2, 8 7 - 9 3 .
Bunting, J. T., and R. P. d'Entremont (1982). Improved Cloud Detection Utilizing Defense Meteorolog-
ical Satellite Program Near Infrared Measurements. Air Force Geophysics Lab. Tech. Rep. AFGL-
T R - 8 2 - 0 0 2 7 , Hanscom AFB, MA.
Carlson, Τ. N . (1979). Atmospheric turbidity in Saharan dust outbreaks as determined by analyses
of satellite brightness data. Mon. Wea. Rev., 107, 3 2 2 - 3 3 5 .
Carlson, Τ. N . , and P. Wendling (1977). Reflected radiance measured by N O A A 3 V H R R as a
function of optical depth for Saharan dust. / . Appl. Meteor., 16, 1 3 6 8 - 1 3 7 1 .
Chahine, Μ. T. (1970). Inverse problems in radiative transfer: Determination of atmospheric parame-
ters. / . Atmos. Sei., 2 7 , 9 6 0 - 9 6 7 .
Chahine, Μ. T. (1974). Remote sounding of cloudy atmospheres. I. The single cloud layer. / . Atmos.
Sei., 3 1 , 2 3 3 - 2 4 3 .
Chahine, Μ. T. (1975). An analytical transformation for remote sensing of clear-column atmospheric
temperature profiles. / . Atmos. Sei., 3 2 , 1 9 4 6 - 1 9 5 2 .
Chahine, Μ. T. (1977). Remote sounding of cloudy atmospheres. II. Multiple cloud formations. / .
Atmos. Sei., 3 4 , 7 4 4 - 7 5 7 .
Chahine, Μ. T. (1982). Remote sensing of cloud parameters. / . Atmos. Sei., 3 9 , 1 5 9 - 1 7 0 .
Chu, W. P., M. P. McCormick, J. Lenoble, C. Brogniez, and P. Pruvost (1989). SAGE II inversion
algorithm. / . Geophys. Res., 9 4 , 8 3 3 9 - 8 3 5 1 .
Chu, W. P. and M. P. McCormick (1979). Inversion of stratospheric aerosol and gaseous constituents
from spacecraft solar extinction data in the 0 . 3 8 - 1 . 0 - / x m wavelength region. Appl. Optics, 18,
1404-1413.
Coakley, J. Α., Jr., and P. Chylek (1975). The two-stream approximation in radiative transfer: Including
the angle of the incident radiation. / . Atmos. Sei., 3 2 , 4 0 9 - 4 1 8 .
Coakley, J. Α., Jr. (1983). Properties of multilayered cloud systems from satellite imagery. / . Geophys.
Res., 88, 1 0 8 1 8 - 1 0 8 2 8 .
Coaklev, J. Α., Jr., and D. G. Baldwin (1984). Towards the objective analysis of clouds from satellite
imagery data. / . Climate Appl. Meteor., 2 3 , 1 0 6 5 - 1 0 9 9 .
Coakley, J. Α., Jr., and F. P. Bretherton (1982). Cloud cover from high-resolution scanner data:
Detecting and allowing for partially filled fields of view. / . Geophys. Res., 87, 4 9 1 7 - 4 9 3 2 .
Coakley, J. Α., Jr., and R. Davies (1986). The effect of cloud sides on reflected solar radiation as
deduced from satellite observations. / . Atmos. Set., 4 3 , 1 0 2 5 - 1 0 3 5 .
Coakley, J. Α., Jr., R. L. Bernstein, and P. A. Durkee (1987). Effect of ship-stack effluents on cloud
reflectivity. Science, 2 3 8 , 1 0 2 0 - 1 0 2 2 .
Cox, C. C , and W. Münk (1954). Measurements of the roughness of the sea surface from photographs
of the sun's glitter. / . Optical Soc. Am., 4 4 , 8 3 8 - 8 5 0 .
Dave, J. V. (1964). Meaning of successive iteration of the auxiliary equation in the theory of radiative
transfer. Astrophys. J., 1 4 0 , 1 2 9 2 - 1 3 0 3 .
302 Chapter 8 Clouds and Aerosols
Desbois, M., G. Seze, and G. Szejwach (1982). Automatic classification of clouds on METEOSAT
imagery: Application to high-level clouds. / . Appl. Meteor., 2 1 , 4 0 1 - 4 1 2 .
Deuze, J. L., C. Devaux, M. Herman, R. Sanier, and D. Tanre (1988). Saharan aerosols over the
south of France: Characterization derived from satellite data and ground based measurements. / .
Appl. Meteor., 27, 6 8 0 - 6 8 6 .
Duda, R. O., and P. E. Hart (1973). Pattern Classification and Scene Analysis. Wiley, N e w York.
Durkee, P. Α., D . R. Jensen, Ε. E. Hindman, and Τ. H. Vonder Haar (1986). The relationship between
marine aerosol particles and satellite-derived radiance. / . Geophys. Res., 9 1 , 4 0 6 3 - 4 0 7 2 .
Durkee, P. Α., F. Pfeil, Ε. Frost, and R. Shema (1991). Global analysis of aerosols and their effects
on climate. Atmos. Environ., 25A, 2 4 5 7 .
Ebert, E. (1987). A pattern recognition technique for distinguishing surface and cloud types in the
polar region. / . Climate Appl. Meteor., 26, 1 4 1 2 - 1 4 2 7 .
Fett, R. W., and R. G. Isaacs (1979). Concerning causes of "anomalous gray shades" in DMSP visible
imagery. / . Appl. Meteor., 18, 1 3 4 0 - 1 3 5 1 .
Fräser, R. S. (1976). Satellite measurement of mass of Sahara dust in the atmosphere. Appl. Optics,
15, 2 4 7 1 - 2 4 7 9 .
Fräser, R. S., Y. J. Kaufman, and R. L. Mahoney (1984). Satellite measurements of aerosol mass and
transport. Atmos. Environ., 18, 2 5 7 7 - 2 5 8 4 .
Fujita, Τ. T. (1963). Use of TIROS pictures for studies of the internal structure of tropical storms.
Mesometeorology Project Research Paper No. 25., The University of Chicago.
Fye, F. K. (1978). The AFGWC Automated Cloud Analysis Model. AFGWC Tech. M e m o . 7 8 - 0 0 2 ,
U.S. Air Force Global Weather Central, Offutt Air Force Base, Nebraska.
Griggs, M. (1975). Measurements of atmospheric aerosol optical thickness over water using E R T S - 1
data. / . Air Pollution Control, 25, 6 2 2 - 6 2 6 .
Griggs, M. (1979). Satellite observations of atmospheric aerosols during the E O M E T cruise. / . Atmos
Sei., 36, 6 9 5 - 6 9 8 .
Griggs, M., and L. L. Stowe (1984). Measurements of aerosol optical parameters from satellites. In
G. Fiocco (ed.), IRS '84: Current Problems in Atmospheric Radiation. A. Deepak Publishing,
Hampton, VA, pp. 4 2 - 4 5 .
Grody, N . C. (1976). Remote sensing of atmospheric water content from satellites using microwave
radiometry. IEEE Trans. Antennas Propag., AP-24, 1 5 5 - 1 6 2 .
Grody, N . C , A. Gruber, and W. C. Shen (1980). Atmospheric water content over the tropical Pacific
derived from the N i m b u s - 6 scanning microwave spectrometer. / . Appl. Meteor., 19, 9 8 6 - 9 9 6 .
Gurvich, A. S., and V. V. Demin (1970). Determination of the total moisture content in the atmosphere
from measurements on the Cosmos 2 4 3 satellite. Atmos. Ocean Phys., 6, 4 5 3 - 4 5 7 .
Harris, R., and E. C. Barrett (1978). Toward an objective nephanalysis. / . Appl. Meteor., 17,
1258-1266.
Harshvardhan (1982). The effect of brokenness on cloud-climate sensitivity. / . Atmos. Sei., 3 9 ,
1853-1861.
Harshvardhan and J. A. Weinman (1982). Infrared radiative transfer through a regular array of
cuboidal clouds. / . Atmos. Set., 3 9 , 4 3 1 - 4 3 9 .
Hasler, A. F. (1981). Stereographic observations from geosynchronous satellites: An important new
tool for the atmospheric sciences. Bull. Am. Meteor. Soc, 6 2 , 1 9 4 - 2 1 2 .
Hasler, A. F., R. Mack, and A. Negri (1983). Stereoscopic observations from meteorological satellites.
Adv. Space Res., 2, 1 0 5 - 1 1 3 .
Henderson-Sellers, A. (ed.) (1984). Satellite Sensing of a Cloudy Atmosphere: Observing the Third
Planet. Taylor 8t Francis, London.
Hindman, Ε. E., P. A. Durkee, P. C. Sinclair, and Τ. H. Vonder Haar (1984). Detection of marine
aerosol particles in coastal zones using satellite imagery. Int. ]. Remote Sensing, 5, 5 7 7 - 5 8 6 .
Isaacs, R. G. (1980). Investigation of the effect of low level maritime haze in DMSP V H R and
LF imagery. Naval Environmental Prediction Research Facility Contractor Report CR 80-06,
Monterey, CA.
Jacobowitz, H., W. L. Smith, and A.J. Drummond (1972). Satellite Measurements of Aerosol Backseat-
8.5 Tropospheric Aerosols 303
tered Radiation from the Nimbus F Earth Radiation Budget Experiment. N O A A Tech. Rep. NESS
6 0 , Washington, D C .
Kaufman, Y. J. (1979). Effect of the Earth's atmosphere on contrast for zenith observations. / .
Geophys. Res., 8 4 , 3 1 6 5 - 3 1 7 2 .
Kent, G. S., and M. P. McCormick (1984). SAGE and SAM II measurements of global stratospheric
aerosol optical depth and mass loading. / . Geophys. Res., 8 9 , 5 3 0 3 - 5 3 1 4 .
Kent, G. S., U. O. Farrukh, P. H. Wang, and A. Deepak (1988). SAGE I and SAM II measurements
of 1 μ,πι aerosol extinction in the free troposphere. / . Appl. Meteor., 27, 2 6 9 - 2 7 9 .
Koffler, R., A. G. DeCotiis, and P. K. Rao (1973). A procedure for estimating cloud amount and
height from satellite infrared radiation data. Mon. Wea. Rev., 1 0 1 , 2 4 0 - 2 4 3 .
Lee, Τ. Y., R. S. Fraser, and Y. Kaufman (1986). Satellite measurements of physical properties
of Saharan dust. Preprints: Second Conference on Satellite Meteorology/Remote Sensing and
Applications, American Meteorological Society, Boston, pp. 2 9 4 - 2 9 7 .
Livingston, J. M., and P. B. Russell (1989). Retrieval of aerosol size distribution moments from
multiwavelength particulate extinction measurements. / . Geophys. Res., 9 4 , 8 4 2 5 - 8 4 3 3 .
Mack, R. Α., A. F. Hasler, and R. F. Adler (1983). Thunderstorm cloud top observations using
satellite stereoscopy. Mon. Wea. Rev., I l l , 1 9 4 9 - 1 9 6 4 .
McClain, E. P., W. G. Pichel, and C. C. Walton (1985). Comparative performance of AVHRR-based
multichannel sea surface temperatures. / . Geophys. Res., 9 0 , 1 1 5 8 7 - 1 1 6 0 1 .
McCleese, D. J., and L. S. Wilson (1976). Cloud top heights from temperature sounding instruments.
Quart. J. Roy. Meteor. Soc, 1 0 2 , 7 8 1 - 7 9 0 .
McCormick M. P., and J. E. Lovill (eds.) (1983). Space observations of aerosols and ozone. Adv.
Space Res., 2, 2 1 3 pp.
McCormick, M. P. (1983). Aerosol measurements from Earth orbiting spacecraft. Adv. Space Res.,
2, 7 3 - 8 6 .
McCormick, M. P., Η. M. Steele, P. Hamill, W. P. Chu, and T. J. Swissler (1982). Polar stratospheric
cloud sightings by SAM II. / . Atmos. Set., 2 9 , 1 3 8 7 - 1 3 9 7 .
McCormick, M. P., P. Hamill, T. J. Pepin, W. P. Chu, T. J. Swissler, and L. R. McMaster (1979).
Satellite studies of stratospheric aerosol. Bull. Am. Meteor. Soc, 6 0 , 1 0 3 8 - 1 0 4 6 .
McKee, T., and S. K. C o x (1974). Scattering of visible radiation by finite clouds. / . Atmos. Set.,
31, 1885-1892.
McKee, T., and S. K. C o x (1976). Simulated radiance patterns for finite cubic clouds. / . Atmos. Sei.,
33, 2 0 1 4 - 2 0 2 0 .
McMillin, L. M., and C. Dean (1982). Evaluation of a new operational technique for producing clear
radiances. / . Appl. Meteor., 2 1 , 1 0 0 5 - 1 0 1 4 .
Menzel, W. P., W. L. Smith, and T. R. Stewart (1983). Improved cloud motion wind vector and
altitude assignment using VAS. / . Climate Appl. Meteor., 2 2 , 3 7 7 - 3 8 4 .
Minnis, P., and E. F. Harrison (1984). Diurnal variability of regional cloud and clear-sky radiative
parameters derived from GOES data. Part I: Analysis method. / . Climate Appl. Meteor., 2 3 ,
993-1011.
Minzner, R. Α., W. E. Shenk, J. Steranka, and R. D. Teagle (1976). Cloud heights determined
stereographically from imagery recorded simultaneously by t w o synchronous meteorological satel-
lites, S M S - 1 and S M S - 2 . EOS, 57, 5 9 3 .
Norton, C. C , F. R. Mosher, B. Hinton, D . W. Martin, D. Santek, and W. Kuhlow (1980). A model
for calculating desert aerosol turbidity over the oceans from geostationary satellite data. / . Appl.
Meteor., 19, 6 3 3 - 6 4 4 .
Phulpin, Τ., M. Derrien, and A. Brard (1983). A two-dimensional histogram procedure to analyze cloud
cover from N O A A satellite high-resolution imagery. / . Climate Appl. Meteor., 22, 1 3 3 2 - 1 3 4 5 .
Platt, C. M. R. (1979). Remote sounding of high clouds: I. Calculation of visible and infrared optical
properties from lidar and radiometer measurements. / . Appl. Meteor., 18, 1 1 3 0 - 1 1 4 3 .
Platt, C. M. R. (1983). On the bispectral method for cloud parameter determination from satellite
VISSR data: Separating broken cloud and semitransparent cloud. / . Climate Appl. Meteor., 12,
429-439.
304 Chapter 8 Clouds and Aerosols
Platt, C. M. R., and A. C. Dilley (1979). Remote sounding of high clouds: II. Emissivity of cirrostratus.
/. Appl. Meteor., 18, 1 1 4 4 - 1 1 5 0 .
Prabhakara, C , I. Wang, A. T. C. Chang, and P. Gloersen (1983). A statistical examination of
N i m b u s - 7 S M M R data and remote sensing of sea surface temperature, liquid water content in
the atmosphere and surface wind speed. / . Climate Appl. Meteor., 2 2 , 2 0 2 3 - 2 0 3 7 .
Ramsey, R. C. (1968). Study of the Remote Measurement of Ocean Color. Final Report: NASW-
1658. TRW, Redondo Beach, California.
Randall, D. Α., J. A. Coakley, Jr., C. W. Fairall, R. A. Kropfli, and D. H. Lenschow (1984). Outlook
for research on subtropical marine stratiform clouds. Bull. Am. Meteor. Soc, 6 5 , 1 2 9 0 - 1 3 0 1 .
Rao, C. R. N . , L. L. Stowe, E. P. McClain, and J. Sapper (1988). Development and application of
aerosol remote sensing with A V H R R data from the N O A A satellites. In P. V. Hobbs and M. P.
McCormick (eds.), Aerosols and Climate, A. Deepak Publishing, Hampton, VA.
Reynolds, D. W., and Τ. H. Vonder Haar (1977). A bispectral method for cloud parameter determina-
tion. Mon. Wea. Rev., 105, 4 4 6 - 4 5 7 .
Reynolds, D. W., Τ. B. McKee, and K. S. Danielson (1978). Effects of cloud size and cloud particles
on satellite-observed reflected brightness. / . Atmos. Sei., 3 5 , 1 6 0 - 1 6 4 .
Rossow, W. B., and R. A. Schiffer (1991). ISSCP cloud data products. Bull. Am. Meteor. Soc, 7 2 , 2 - 2 0 .
Rossow, W. B. (1989). Measuring cloud properties from space: A review. / . Climate, 2, 2 0 1 - 2 1 3 .
Rossow, W. B., and L. Garder (1984). Selection of a map grid for data analysis and archival. / .
Climate Appl. Meteor., 2 3 , 1 2 5 3 - 1 2 5 7 .
Rossow, W. B., F. Mosher, E. Kinsella, A. Arking, M. Desbois, E. Harrison, P. Minnis, E. Ruprecht,
G. Seze, C. Simmer, and E. Smith (1985). ISCCP cloud algorithm intercomparison. / . Climate
Appl. Meteor., 2 4 , 8 7 7 - 9 0 3 .
Rossow, W. B., L. C. Garder, and A. A. Lacis (1989). Global, seasonal cloud variations from satellite
radiance measurements Part I: Sensitivity of analysis. / . Climate, 2, 4 1 9 - 4 5 8 .
Russell, P. Β., M. P. McCormick, T. J. Swissler, J. M. Rosen, D. J. Hofmann, and L. R. McMaster
(1984). Satellite and correlative measurements of the stratospheric aerosol. Ill: Comparison of
measurements by SAM II, SAGE, dustsondes, filters, impactors and lidar. / . Atmos. Sei., 4 1 ,
1791-1800.
Russell, P. Β., M. P. McCormick, T. J. Swissler, W. P. Chu, J. M. Livingston, W. H. Fuller, J. M.
Rosen, D. J. Hofmann, L. R. McMaster D . C. Woods, and T. J. Pepin (1981). Satellite and
correlative measurements of the stratospheric aerosol. II: Comparison of measurements by SAM
II, dustsondes, and airborne lidar. / . Atmos. Sei., 3 8 , 1 2 9 5 - 1 3 0 2 .
Saunders, R. W. (1985). Monthly mean cloudiness observed from M E T E O S A T - 2 . / . Climate Appl.
Meteor., 2 4 , 1 1 4 - 1 2 7 .
Saunders, R. W., and G. E. Hunt (1983). Some radiation budget and cloud measurements derived
from Meteosat 1 data. Tellus, 35B, 1 7 7 - 1 8 8 .
Schiffer, R. Α., and W. B. Rossow (1983). The International Satellite Cloud Climatology Project
(ISCCP): The first project of the World Climate Research Programme. Bull. Am. Meteor. Soc,
6 4 , 779-784.
Shenk, W. E., and R. J. Curran (1973). A multi-spectral method for estimating cirrus cloud top
heights. / . Appl. Meteor., 12, 1 2 1 3 - 1 2 1 6 .
Shenk, W. E., and V. V. Salomonson (1972). A simulation study exploring the effects of sensor spatial
resolution on estimates of cloud cover from satellites. / . Appl. Meteor., 1 1 , 2 1 4 - 2 2 0 .
Shettle, E. P., and R. W. Fenn (1979). Models for the Aerosols of the Lower Atmosphere and the
Effects of Humidity Variations on Their Optical Properties. Air Force Geophysics Laboratory
Tech. Rep. AFGL-TR-79-0214, Hanscom Air Force Base, MA.
Simmer, C , E. Raschke, and E. Ruprecht (1982). A method for determination of cloud properties
from two-dimensional histograms. Ann. Meteor., 18, 130—132.
Smith, Ε. Α., and D . W. Reynolds (1976). Comparison of cloud top height determinations from
three independent sources: Satellite IR measurements: Satellite viewed cloud shadows: Radar.
Proceedings of the Symposium on Meteorological Observations from Space: Their contribution
to the First GARP Global Experiment. COSPAR, Boulder, CO.
8.5 Tropospheric Aerosols
305
Smith, W. L. (1968). An improved method for calculating tropospheric temperature and moisture
from satellite radiometer measurements. Mon. Wea. Rev., 9 6 , 3 8 7 - 3 9 6 .
Smith, W. L., and C. M. R. Piatt (1978). Comparisons of satellite-deduced cloud heights with indica-
tions from radiosonde and ground-based laser measurements. / . Appl. Meteor., 17, 1 7 9 6 - 1 8 0 2 .
Smith, W. L., and Η. M. Woolf (1976). The use of eigenvectors of statistical covariance matrices for
interpreting satellite sounding radiometer data. / . Atmos. Sei., 3 3 , 1 1 2 7 - 1 1 4 0 .
Smith, W. L., Η. M. Woolf, and W. J. Jacob (1970). A regression method for obtaining real-time
temperature and geopotential height profiles from satellite spectrometer measurements and its
application to Nimbus 3 "SIRS" observations. Mon. Wea. Rev., 9 8 , 5 8 2 - 6 0 3 .
Smith, W. L., Η. M. Woolf, P. G. Abel, C. M. Hayden, M. Chalfant, and N . Grody (1974). Nimbus-5
Sounder Data Processing System. N O A A Tech. M e m o . NESS 5 7 , Washington, D C .
Stowe, L. L., C. G. Wellemeyer, T. F. Eck, Η. Υ. Μ. Yeh, and the N i m b u s - 7 Cloud Data Processing
Team (1988). N i m b u s - 7 global cloud climatology. Part I: Algorithms and validation./. Climate,
1, 4 4 5 - 4 7 0 .
Susskind, J., J. Rosenfield, D. Reuter, and Μ. T. Chahine (1984). Remote sensing of weather and
climate parameters from HIRS2/MSU on T I R O S - N . / . Geophys. Res., 89, 4 6 7 7 - 4 6 9 7 .
Szejwach, G. (1982). Determination of semi-transparent cirrus cloud temperature from infrared radi-
ances: Application to METEOSAT. / . Appl. Meteor., 21, 3 8 4 - 3 9 3 .
Twomey, S., B. Herman, and R. Rabinoff (1977). An extension to the Chahine method of inverting
the radiative transfer equation. / . Atmos. Sei., 3 4 , 1 0 8 5 - 1 0 9 0 .
Wang, P.-H., M. P. McCormick, T. J. Swissler, Μ. T. Osborn, W. H. Fuller, and G. K. Yue (1989).
Inference of stratospheric aerosol composition and size distribution from SAGE II satellite measure-
ments. / . Geophys. Res., 9 4 , 8 4 3 5 - 8 4 4 6 .
Weinman, J. Α., and Harshvardhan (1982). Solar reflection from a regular array of horizontally finite
clouds. Appl. Optics, 2 1 , 2 9 4 0 - 2 9 4 4 .
Westwater, E. R. (1972). Microwave Emission from Clouds. N O A A Tech. Rep. 219-WPL 18. Boul-
der, C O .
Wielicki, Β. Α., and J. A. Coakley Jr. (1981). Cloud retrieval using infrared sounder data: Error
analysis. / . Appl. Meteor., 2 0 , 1 5 7 - 1 6 9 .
Wilheit, Τ. T. and A. T. C. Chang (1980). An algorithm for retrieval of ocean surface and atmosphere
parameters from the observations of the scanning multichannel microwave radiometer. Radio Sei.,
15, 5 2 5 - 5 4 4 .
Yue, G. Κ., M. P. McCormick, and W. P. Chu (1984). A comparative study of aerosol extinction
measurements made by the SAM II and SAGE satellite experiments. / . Geophys. Res., 8 9 ,
5321-5327.
Yue, G. Κ., M. P. McCormick, W. P. Chu, P. Wang, and Μ. T. Osborn (1989). Comparative studies
of aerosol extinction measurements made by the SAM II and SAGE II satellite experiments. / .
Geophys. Res., 9 4 , 8 4 1 2 - 8 4 2 4 .
9
Precipitation
gm τ FIRST, ONE might think that if the clouds have been located in a
f % satellite image (as in Chapter 8) precipitation estimation should be a
simple task. Unfortunately, early researchers found that only a small fraction of
clouds produce rain at any one time (Nagle and Serebreny, 1962). Much of the
research into techniques for estimating precipitation with satellite data involves
separating raining from nonraining clouds. An excellent book covering this subject
is The Use of Satellite Data in Rainfall Monitoring by E. C. Barrett and D. W.
Martin (1981).
Precipitation-estimation techniques may be divided into three categories: those
that use visible or infrared data, those that use passive microwave data, and future
techniques that will use radar. The first three sections of this chapter discuss these
techniques. The final section discusses the related topic of severe weather detection
and monitoring.
Visible and infrared techniques are grouped together because they share a
common characteristic: The radiation does not penetrate through the cloud. Visi-
307
308 Chapter 9 Precipitation
ble and infrared techniques estimate precipitation falling from the bottom of the
cloud based on radiation coming from the top and/or the side of the cloud,
depending on viewing geometry. All visible and infrared precipitation-estimation
schemes are necessarily indirect; a cloud's brightness or equivalent blackbody
temperature may be related to the rain falling from it, but the raindrops themselves
are not directly sensed.
Because of the indirect nature of the relationship between satellite-measured
visible or infrared radiance and precipitation, the techniques discussed below are
not universally applicable. Techniques developed for the tropics may not perform
well in the extratropics. Similarly, techniques developed to estimate monthly
rainfall may not be useful when estimating hourly rainfall. When selecting a visible
or infrared technique, one should carefully compare one's application with the
application for which the technique was developed. In fact, Barrett (1988) argues
that a hierarchy of techniques is necessary to estimate rainfall from satellite data
under the wide range of meteorological conditions and user needs which are
encountered in practice.
A further note is in order here, although it applies just as well to microwave
and radar techniques as it does to visible and infrared techniques. There are two
ways of verifying precipitation estimates from satellite data; one may compare
either with radar or with rain-gauge estimates. Radar probably provides the more
reasonable comparison because it samples volumes of the atmosphere which
are comparable with the size of satellite pixels. However, radar meteorologists
generally do not claim better than a factor-of-2 accuracy for their precipitation
estimates. Rain gauges make very accurate measurements, but a standard rain
gauge samples an area only about 1 0 m . Satellite estimates, on the other hand,
- 1 2
R = 2rifi, (9.1)
i
where r- is the rain rate assigned to cloud type /, and /· is the fraction of time that
t
the point is covered with (or fraction of the area covered by) cloud type /.
9.1.1.1 Barrett
Cloud indexing was pioneered by Barrett (1970), who wanted to estimate
precipitation over Australia and the "Maritime Continent." He used once-daily
nephanalyses produced by NOAA's predecessor, the Environmental Science Ser-
vice Administration (ESSA), to estimate monthly precipitation with a slightly
more complex scheme than that represented by Eq. 9.1 (Fig. 9.1). Barrett's
1
9.1.1.2 Follansbee
Whereas Barrett (1970) wanted high-spatial-resolution precipitation estimates
averaged over a month, Follansbee (1973) wanted daily precipitation estimates
averaged over a state-sized area. Using once-daily, afternoon, visible satellite
images, he visually estimated the fraction of a state covered by cumulonimbus
1
Barrett's cloud indexing equation also included a mean monthly cloudiness and the probability
of rainfall from each cloud type as well as the rain rate. Since these factors are multiplicative, however,
Eq. 9.1 still applies, but r must be interpreted as an effective rain rate.
t
Chapter 9 Precipitation
310
90° 100° 110° 120° 130° 140° 150° 160° 170° 180 (
PRECIPITATION COEFFICIENTS
FIGURE 9.2. Rain-gauge-estimated precipitation versus precipitation estimated by cloud indexing for
the area in Fig. 9.1 during the months of March-June 1966. [After Barrett (1970).]
9.1 Visible and Infrared Techniques J]]
FIGURE 9.3. Quadrant constructed for estimation of daily rainfall at Valentia Island, Ireland. The 8 0 0
hPa wind is used to construct the quadrant. [After Barrett (1973).]
hours during which precipitation fell. At any point, the 12-h precipitation is based
on the duration (D, in hours) of precipitation, on the normal monthly precipitation
(P , which helps correct for the variable precipitation rate in synoptic-scale sys-
N
FIGURE 9.5. Schematic diagram illustrating the interpolation technique of Follansbee (1976). The heavy
lines outline the raining area of an extratropical cyclone with a trailing front. The solid heavy line
is the position at midnight, the dashed heavy line outlines the position at noon. The thin lines (and
the numbers) indicate the number of hours that a point was under a raining area.
precipitation, the mean absolute difference was approximately 100% of the mean.
Although the 100% figure seems large, it is for a one-day period at a single
station; in the above schemes, precipitation was averaged over a large area or
over a long time. When averaged over a given area, the mean absolute difference
3] 4 Chapter 9 Precipitation
would fall. It must also be kept in mind that depending on one's application,
even a 100% mean absolute difference may be adequate.
50 W e
40° 30* 20* HTW
50 W
e
40· 30* 20* 10* W
FIGURE 9.6. Cumulative rainfall estimates (mm) for 80 days during GATE: (a) Kilonsky-Ramage
technique, (b) Griffith-Woodley technique. [After Garcia (1981).]
9.1 Visible and Infrared Techniques
315
9.7.7.4 Arkin
to the Kilonsky-Ramage technique, but it uses infrared GOES data. Arkin found
that for the GATE area, radar-estimated precipitation was highly correlated (corre-
lation coefficient .86) with the fraction of the area covered by pixels colder than
235 K. Other thresholds were tried but yielded lower correlations. Of course, the
correlation coefficient depends on the area and time over which the precipitation
is estimated. Richards and Arkin (1981) tested averaging areas between 0.5° x
0.5° and 2.5° x 2.5° latitude and averaging times from 1 to 24 h. They found
that the correlations increase with averaging area and with averaging time.
Arkin and Meisner (1987) call their precipitation estimate the GOES Precipita-
tion Index (GPI). They use a 235 Κ threshold and a constant rain rate of 3 mm 3
h" , which are appropriate values for estimating tropical precipitation in areas
1
2
Actually this is a threshold technique rather than a true cloud indexing technique, but it fits best
in this section.
3
This represents the slope of the regression between areally averaged rain rate and fractional area
with equivalent blackbody temperatures less than the threshold. The small intercept in this regression
is ignored.
3]£ Chapter 9 Precipitation
1 >270 <119
2 266-270 119-128
3 261-265 129-138
4 256-260 139-148
5 251-255 149-158
6 246-250 159-168
7 241-245 169-177
8 236-240 178-182
9 231-235 183-187
10 226-230 188-192
11 221-225 193-197
12 216-220 198-202
13 211-215 203-207
14 201-210 208-217
15 191-200 218-227
16 <191 >227
period after the archive's start in December 1981. Using the full-resolution GOES
4
Clouds that are bright in visible images are more likely to precipitate than
dark clouds because brightness is related to optical depth and thus to cloud
thickness. Clouds that are cold in infrared images are more likely to precipitate
than warm clouds because cold clouds have higher tops than warm clouds. There
are exceptions to these rules, however. Stratus clouds are bright, but do not rain
as much, nor as often, as cumulonimbus clouds. Cirrus clouds are cold, but do
not produce as much precipitation as some warmer clouds. Bispectral methods
Phenomena with periods between 1 and 15 days cannot be studied because of the half-month aver-
4
aging.
9.1 Visible and Infrared Techniques 317
attempt to combine these rules by saying that clouds which have the best chance
of raining are both cold and bright (Lethbridge, 1967). Lesser amounts (lower
probabilities) of precipitation can be expected from cold-but-dark clouds (cirrus)
and bright-but-warm clouds (stratus).
37 2) 10 1 2 0 3 0 0 0 0
40 32 13 1« 12 10 3 1
• • s
30 77 23 IS 24 • 7 17 4 3 S
IS 34 20 1» 12 10 11 13 7 10 4
« 42 14 1« 1« 14 7 « 0 4
1·
1 33 14 10 IS 14 11 11 7 1 0
2t 22 12 10 • • t S 9 7
1 27 20 17 7 < • 9 % S 2
2 12 24 20 « 7 • • C « 2
11 30 20 C 4 9 S C 7 4
ι
0 13 It 27 10 IS t 13 < 7 3
0 3 2S S3 34 10 10 I 10 7 S
1 0 17 47 40 1« 13 17 13 14 ·
0 2 2 10 44 2» IS 1« 9 IS 13
0 0 4 10 27 34 20 30 13 34 2«
0 0 2 3 11 9 24 30 35 33
•
0 0 0 0 0 0 2 2 4 2< 31
0 0 0 0 0 0 0 0 0 0 0
VIS VIS
FIGURE 9.7. Two-dimensional histograms of GOES data for (a) nonraining and (b) raining pixels as determined by radar. The data are for GATE area on 5
September 1974 at 1300 UTC. (Infrared counts increase downward, infrared radiance increases upward.) [After Lovejoy and Austin (1979a).]
9.1 Visible and Infrared Techniques 319
vis
FIGURE 9.8.P] ecipitation probabilities derived from Fig. 9.7. The contour separates raining from
nonraining pi xels based on an optimum probability. [After Lovejoy and Austin (1979a).]
a time when only 2.5% of the area was covered with rain; thus the minimum
success rate was 97.5%. Another way to measure success is to ask the question,
of pixels that are classified as raining, what fraction are not raining? This measure
is called the false-alarm ratio (FAR). The FAR averaged 3 5 % for GATE and
about 4 5 % for Montreal. Lovejoy and Austin (1979a) offer other measures of
success that are probably better than the two mentioned here but are also less
intuitive; the reader is referred to their paper.
Lovejoy and Austin studied only one stratiform rain case. Their technique
worked less well in the stratiform case than in the convective cases. The major
difficulty was that the histogram for the raining pixels was very similar to that
for the nonraining pixels, with the raining pixels only slightly colder and brighter
that the nonraining pixels and with considerable overlap in the distributions. In
other words, visible and infrared data contain little information that can be used
to distinguish raining from nonraining stratiform clouds.
Lovejoy and Austin compared their bispectral technique to monospectral
threshold techniques. The bispectral technique always performed better than either
visible or infrared thresholds.
320 Chapter 9 Precipitation
Ζ
3
Δ
Raining Clouds
Ο
0
Non-raining Clouds
ϋ
Ο • Clear Skies (land or water)
UJ UJ
Q_ tr
UJ 2
2
-71 °C 216
-27°C 168 /
( Δ
\
-15°C 144
CLUSTER 1 ^ - " ο """δ^χ ^
-3°C 120
ο °ο ° Λ ° ° 7 C L U S T E R
")
+9°C 96 / ο ο %ο / ^
' ο ο , ^
y '
+21°C 72 Ο Ο Ο ΟΟ
0
I WATERο S
LAND
+33°C 48
+45°C 24
+57° 0 1 1 » 1 1 1 1 ι ΐ—
6 12 18 24 30 36 42 48 54 60
VISIBLE COUNT
FIGURE 9.9. Pixel clusters in visible-infrared space. [After Tsonis and Isaac (1985).]
9.1 Visible and Infrared Techniques 321
zenith angle. This approximates the count value which would be measured if the
sun were directly overhead.
where A is the cloud area, dAldt is the time rate of change of the cloud area,
and a and a are empirically determined coefficients. Because a is positive, this
0 1 x
equation ensures that the rain rate will be larger in the growing stage than in the
decaying stage of the cloud.
Using half-hourly GOES data in the GATE area, Stout et al. adopted a threshold
count of 172 to define clouds in the visible channel, and 160 (250 K) in the
infrared The satellite areas were regressed against radar-estimated rain rates.
Table 9.2 gives the regression coefficients a and a which best fit radar and
0 x
satellite data. Note that the resulting rain rate is in units of m s" , which is called
3 1
the volumetric rain rate. It is the volume of water falling from the cloud per
unit time.
399 Chapter 9 Precipitation
τ ι ι η 1 1 1 1 1 1 1 Γ
VISIBLE AREA K M (DIGITAL L E V E L 172)
2
30,000
VOLUMETRIC RAINFALL RATE M S " ' 3
20,000h
10,000
—I I I L.
1100 1200 1300 1400 1500 1600
GMT
FIGURE 9.10. The evolution of a typical thunderstorm. Cloud area as observed in visible and infrared
GOES images and radar-estimated volumetric rain rate are plotted against time. [After Stout et
al. (1979).]
Comparing rainfall from a single thunderstorm for a 30-min period, the stan-
dard error of estimate was 76% of the mean rain rate for the infrared estimate
and 62% for the visible estimate. These are very good estimates in comparison
with some of those above. The improvement comes about because (1) Stout et
al. were dealing with a single cloud type in a single area, and (2) their rain-rate
equation better approximates the relationship between satellite-observed cloud
area and precipitation. Comparing hourly rainfall over the GATE area (10 km ) 5 2
estimated by radar and by the Stout et al. technique produced a .84 correlation
coefficient and a 0.25-mm standard error of estimate.
Band a0 (m s- )
1
ax (m)
resulting clouds are treated as new clouds. Figure 9.11a (Griffith et al., 1978)
shows the empirical curves used to determine the radar echo area from the satellite-
estimated area of the cloud (A ). The echo area (A ) is estimated as a fraction of
c e
the maximum cloud area depending on the ratio AJA and the sign of the time
m
rate of change of A . A itself determines which curve to use in Fig. 9.11a. The
c m
rain rate is estimated using Fig. 9.11b (Woodley et al., 1980; Griffith et al., 1980)
and a knowledge of the ratio of the echo area to the maximum echo area. The
rain volume falling from the cloud is then the product of (1) the rain rate, (2)
the echo area, (3) the time interval between successive satellite images, and (4)
an empirical factor (Griffith et al., 1978) that starts at 1.00 and increases to a
maximum of 3.24 essentially as the mean temperature of the cloud top decreases.
Finally the total rain volume is apportioned within the cloud; one half of the rain
falls uniformly below the coldest 10% of the cloud top, the remaining half falls
below the next warmest 40% of the cloud top. No rain falls in the warmest half
of the cloud.
The Griffith-Woodley technique has been applied to many areas of the globe:
Florida, the U.S. High Plains, Venezuela, and the tropical Atlantic. In each case
the scheme was modified slightly, either in the way precipitation is apportioned
below the cloud or in the adjustment factor. In GATE, for example, the rain was
uniformly distributed below the cloud (Woodley et al., 1980). In the U.S. High
Plains, the precipitation was adjusted by a factor derived from the ratio of High
Plains to Florida rain-gauge data or derived from the ratio of cloud model precipi-
tation in the High Plains and in Florida (Griffith et al., 1981). The basic (Florida)
relationships between cloud area and precipitation embodied in Fig. 9.11, how-
ever, have remained unchanged.
The accuracy of the Griffith-Woodley technique is perhaps best portrayed by
Griffith (1987a,b). She used GOES data to calculate precipitation over a 3.6 x
10 -km area in the central United States during August 1979. Statistics were
6 2
accumulated for hourly and daily time periods and for the entire area as well as
for specific rain-gauge locations (points). A portion of her results are shown in
Table 9.3. Like other techniques, the Griffith-Woodley technique performs best
for longer time periods and larger areas. The technique could not be considered
useful for estimating how much rain falls in a single rain gauge, except, perhaps,
0.2 0.4 Ο.β 0.8 1.0 0.8 0.6 0.4 0.2
INCREASING V A m DECREASING
0.00
0.2 0.4 Ο.β 0.8 1.0 0.8 0.6 0.4 0.2
INCREASING V A m DECREASING
1 1 1
A m > 1 0 , 0 0 0 Km2
1.00
0.75
Η 0.50 ui
Η 0.25
Area Point
for periods much longer than a day. Daily rainfall over a large area, however, is
well represented by Griffith-Woodley, and hourly precipitation is acceptable
(depending, of course, on one's requirements).
Recently, Negri et al. (1984) and Griffith (1987a,b) have critically examined
the Griffith-Woodley technique. The weakest part of the technique seems to
be the process of tracking clouds throughout their lifetimes before assigning
precipitation. While convective clouds do go through a life cycle, it is often not
observable with visible or infrared satellite data because the cirrus anvils of
neighboring clouds interact. In an area such as southern Florida, although there
can be dozens of active thunderstorms on a summer afternoon, late in the day
there may be only one cloud, as defined by the 253-K isotherm. This explains in
part why most visible and infrared precipitation schemes tend to underestimate
convective precipitation early in the day and to overestimate it later. Negri et al.
(1984) and Griffith (1987a,b) showed independently that a simplified precipitation
scheme, which does not track clouds (each one is treated as if it existed in only
one image), performs no worse than the full Griffith-Woodley scheme, and it
takes only 10% of the computer time.
FIGURE 9.11. (a) Echo area/cloud area relationships for infrared data in the Griffith-Woodley technique,
normalized by maximum cloud area. Curves are the actual Griffith-Woodley approximations to
empirical data. [Adapted from Griffith et al. (1978) by Negri et al. (1984).] (b) The Griffith-Woodley
relationship between rain rate and echo area. The singular point at 2 0 . 7 mm h" is for clouds at their
1
maximum echo area. [Adapted from Griffith et al. (1980) by Negri et al. (1984).]
Chapter 9 Precipitation
326
The first property has been confirmed with radar data. Doneaud et al. (1984)
and Lopez et al. (1983) have shown that the volume of rain that falls from a
convective cloud through its lifetime is very nearly proportional to the area-time
integral (ATI):
(9.7)
where A{t) is the raining area of a cloud (defined by a reflectivity threshold; say,
20 dBZ) at time t, and t is the cloud lifetime. This means that the total rain
T
9.14.1 Gruber
The earliest such attempt was by Gruber (1973a), who noted that Kuo's (1965)
parameterization of convection could be used to relate fractional cloud cover to
rain rate. In extremely simplified form, the scheme is as follows. Consider a grid
box into which moisture is flowing at the rate I. If Q is the amount of moisture
necessary to completely saturate the box, then the rate at which the box is filling
with cloud could be estimated as HQ. However, convective clouds do not live
forever. Suppose that all of the clouds which form are convective and have identical
lifetimes t . Then the rate at which clouds are dying is approximately c/t , where
L L
c is the fractional cloud cover. If one makes the further assumption that the rate
of influx of moisture / is exactly the same as the rain rate R, then one arrives at
the equilibrium relationship R = Qc/t . Gruber developed his scheme using Florida
L
Area Cumulus Experiment (FACE) data from June and July 1970. He calculated
Q from a nearby sounding, c from infrared satellite data, and R from radar data.
The resulting t correlated well with observed cloud lifetimes and averaged 30
L
min. He then tested this scheme on a squall line that moved through Illinois and
Indiana on 3 July 1970 (Gruber, 1973b). Assuming t = 30 min, the mean R
L
mm h . ]
9.1 Visible and Infrared Techniques
327
9.1.4.2 Wylie
Another use of cloud models is in adjusting calibration coefficients. Most of the
above techniques were developed in a particular location. The changes necessary to
apply them elsewhere are not obvious. Wylie (1979) attempted to use the one-
dimensional cloud model of Simpson and Wiggert (1969) to adjust the precipita-
tion estimates. He used the method of Stout et al. (1979) to estimate precipitation
in GATE and around Montreal. Wylie adjusted the satellite rain estimates for the
six cases around Montreal by the ratio of precipitation estimates made using the
cloud model. Substantially improved results were obtained in five of the six cases.
Griffith et al. (1981) used the Wylie technique to apply the Florida-calibrated
Griffith-Woodley technique to the U.S. High Plains.
Second, infrared satellite data are analyzed. Local minima in the IR tempera-
tures are found and screened to eliminate thin, nonprecipitating cirrus. The remain-
ing minima are assumed to be convective elements protruding from the top of
cirrus anvils. Around each convective element the modal temperature in an area
approximately 80 km on a side is calculated. This temperature is assumed to
represent the anvil temperature. The average of all anvil temperatures is used as
a threshold for stratiform precipitation.
Third, precipitation is assigned to the convective elements. The rain rate and
raining area are determined from the cloud-top temperature using the output of
the ID cloud model. To map the rain, the calculated rain rate is assigned to
pixels in a spiral fashion, starting at the center of the temperature minimum, and
continuing until the raining area is reached.
Finally, to every point that is colder than the stratiform threshold and that
did not receive any convective precipitation, a 2-mm-h stratiform rain rate -1
is assigned.
The CST is still experimental, however when tested with Florida data, it outper-
formed all other visible/infrared techniques. In particular, because it assigns con-
vective precipitation only to temperature minima, the CST showed much less
5
This step is more difficult than it sounds because I D cloud models require significant experience
to be applied meaningfully.
328 Chapter 9 Precipitation
tendency to overestimate precipitation late in the day when cirrus debris is wide-
spread but convection has decreased. It remains to be seen whether convective
elements embedded in the anvil, and thus undetectable in infrared data, contribute
substantially to convective precipitation. Perhaps this question could be answered
with microwave imagery.
9.1.4.4 Scofield and Oliver
The final visible-infrared scheme to be discussed is that developed by Scofield
and Oliver (1977) for the purpose of operationally estimating convective precipita-
tion, particularly heavy precipitation. The operational scheme is explained by
Scofield (1987); however, the scheme is continually being improved. The Sco-
field-Oliver technique is in use around the clock at the Synoptic Analysis Branch
of NESDIS, which issues statements on the location and intensity of heavy precipi-
tation. The technique can be and is used in many other locations that have access
to geostationary images.
Barrett and Martin (1981) classified Scofield-Oliver as a life-history technique.
Although it does follow the life history of clouds, we believe that its success is
due to the fact that it is firmly based on a conceptual cloud model; thus we discuss
it in the cloud model section.
The Scofield-Oliver technique is quite different from other visible—infrared
techniques in that it is not automated (though parts of it could be). It relies
on the judgment of a satellite meteorologist to locate precipitation-associated
signatures in satellite images and to assign an appropriate rain rate to points
affected by these features. The technique is thus subjective, but it can take into
account phenomena that automated techniques cannot.
The Scofield-Oliver technique is designed to use (primarily) enhanced, half-
hourly GOES infrared images (enhanced using the MB curve; see Chapter 5),
which are widely available in forecast offices (Fig. 9.12). A decision tree is used
to assign a rain rate to a particular point at the time of the satellite image. A
simplified version of this decision tree is presented in Fig. 9.13 and discussed here.
The first step is for the analyst to determine whether the cloud is convective;
the technique only applies to convective clouds.
If the analyst is using a video display device rather than hard-copy photographs,
the second step should be to determine the likely temperature of the top of the
cirrus anvil. This can be estimated from a sounding by calculating the equilibrium
level (the temperature at which the moist adiabat through the lifting condensation
level crosses the environmental temperature sounding). If the equilibrium level is
significantly warmer than -62°C, the MB enhancement curve should be adjusted
to make the boundary between black and repeat gray fall on the equilibrium-
level temperature.
The main loop starts with the analyst outlining the area of active convection.
Clues to the location of this area include (Scofield, 1987):
• The active area is near the portion of the anvil with the tightest gradient
of IR temperature or near the center of an anvil which has a uniform IR
temperature gradient.
9.1 Visible and Infrared Techniques 329
to decaying clouds, to those that are warming, and to those decreasing in area.
The initial rain estimate is augmented by 8 mm per half hour in the area of
any overshooting tops (most easily located in visible imagery), by 13 mm per half
hour in the region of cloud mergers, and by 5-13 mm per half hour if the
environment is likely to be saturated (upwind edge of storm stationary for an
hour or more). The base rain rate is the rate after this step.
Two multiplicative correction factors are applied to the base rain rate. The
first is called the speed-of-storm factor. Rapidly moving thunderstorms produce
less rain at a point than more slowly moving storms because they stay over a
point for less time. The speed-of-storm factor is between 1.00 (for stationary
storms) and 0.25 for rapidly moving storms. The second multiplicative correction
is the moisture-correction factor. It is numerically equal to the precipitable water
(in inches) times the mean surface-500 hPa relative humidity (0-1). This factor
takes into account the fact that storms in moister environments have higher rain
rates. The final half-hourly rain rate is the base rain rate times the speed-of-storm
factor times the moisture-correction factor.
If the cloud remains active, the analyst goes to the next half hour period and
repeats the main loop.
The gross features of heavy convective precipitation are almost always well
described by the Scofield—Oliver technique, which makes it quite useful for flash
flood forecasting (Fig. 9.14). Because of extreme spatial variations of convective
precipitation (Huff, 1967; Vogel, 1981), comparison with rain-gauge observations
330 Chapter 9 Precipitation
b
WHITEH
GRAYH
/
BLACK —τ -1 τ \ I I I ι ι ι ι ιτ ι ι ι ιι
50 40 30 20 10 0 10 -20 - 3 0 - 5 0 -70 - 9 0 -110
T E M P °C
FIGURE 9.12. (a) Enhanced infrared imagery (MB curve), 0 7 3 0 UTC 13 August 1 9 8 2 . (b) Digital
enhancement curve (MB), (c) Radar summary chart, 0 7 3 5 UTC 13 August 1 9 8 2 . [After Scofield
(1987).]
will always be noisy. However, for storm total precipitation, Scofield (1987) found
that the Scofield-Oliver technique agrees with rain gauges within about 30%.
A precipitation-estimation technique similar to that of Scofield and Oliver has
9.2 Passive Microwave Techniques 331
been developed for use with tropical cyclones (Spayd and Scofield, 1984). A
related technique has been developed for extratropical cyclones (Scofield and
Spayd, 1984). Both of these are in operational use by NESDIS.
Martin and Howland (1986) have developed a precipitation-estimation scheme
called "grid history," which can be viewed as a highly streamlined, semiautomated
Scofield-Oliver-type technique. A grid is placed over the area of interest. At each
grid point, IR data are examined. If the brightness temperature is warmer than
a threshold, the automatic decision is no rain. If the brightness temperature is
colder than a second threshold, heavy rain is automatically assigned. At intermedi-
ate temperatures, an analyst, who views the image (and the accompanying visible
image) is asked to assign a rain category (light, moderate, or heavy). By using
empirically determined rainfall rate coefficients and summing the rainfall through
a day, the daily rainfall can be determined within about a factor of 2.
YES
^CONTINUE
FIGURE 9.13. Simplified decision tree for the Scofield-Oliver technique. (Consult the Synoptic Analysis
Branch of NESDIS for the currently operational technique.)
Q CONTINUE J
INCREASE ^ C / c e ^ V
RAINRATE Y E S _ ^ X % ^ ,
H
R
N " G \
BY 8 mm IN ™ N 6
>
AREAS O F \ . T 0
o
P S
/
COLO T O P S ^νΛχ^
^ NO
INCREASE \w
AREA O F \ · /
MERGER
1 ^ NO
INCREASE \v
RAI Ν R A T E ^SATURAT Ε θ \ .
OF COLDER >sEN VI RON Μ E N Τ /
T O P S BY X . ?
5-13 mm ^^κ^τ
I
CALCULATE
SPEEO OF
S T O R M FACTOR
(S)
CALCULATE
MOISTURE
CORRECTION
FACTOR ( M C F )
HALF-HOURLY
RAINFALL (R)
R=BR χ Sx MCF
f LOOP \
STOP ^
224 Chapter 9 Precipitation
FIGURE 9.14. (a) Twenty-four-hour rain-gauge-observed rainfall ending at 1200 UTC 28 December
1 9 8 3 . (b) Scofield-Oliver estimated rainfall between 1900 UTC 2 7 December and 1000 UTC 28
December 1 9 8 3 . [After Scofield (1987).]
lengths that have been used to measure precipitation (Fig. 9.15). The following
important properties can be seen:
• Ice essentially does not absorb microwave radiation; it only scatters.
• Liquid drops both absorb and scatter, but absorption dominates.
Assuming a Marshall-Palmer raindrop-size distribution. The reader should be aware that the
6
10 20 30 40 50 60
-τ 1 1 Γ-?—ι 1
FIGURE 9.15. Mie volume scattering coefficients (top), volume absorption coefficients (middle), and
single scattering albedos (bottom) for a Marshall—Palmer precipitation size distribution of water and
ice spheres at three frequencies (GHz). (Note that Spencer et al. use k to symbolize volume scattering
coefficient, whereas we use cr.) [After Spencer et al. (1989).]
336 Chapter 9 Precipitation
• Scattering and absorption both increase with frequency and with rain rate.
However, scattering by ice increases much more rapidly with frequency
than scattering by liquid.
Two general conclusions can be drawn. First, the microwave spectrum can be
divided roughly into three parts. Below about 22 GHz, absorption is the primary
mechanism affecting the transfer of microwave radiation. Scattering does occur,
but it is of secondary importance. Above about 60 GHz, scattering dominates
absorption. Between 22 and 60 GHz, both scattering and absorption are import-
ant. Second, at different frequencies, microwave radiometers observe different
parts of the rain structure. Below 22 GHz, any ice above the rain is nearly
transparent; microwave radiometers respond directly to the rain layer. Above 60
GHz, however, ice scattering is the dominant process; microwave radiometers
sense only the ice and cannot see the rain below. Thus precipitation estimates
made at higher frequencies are necessarily more indirect than those made at
lower frequencies.
It must also be noted that cloud droplets, water vapor, and oxygen all absorb
(but do not scatter) microwave radiation and thus have the potential to confuse
precipitation estimates based on absorption.
If a model of precipitation and its atmospheric environment are assumed,
the radiative transfer equation may be used to calculate microwave brightness
temperature as a function of rainfall rate. While such calculations depend strongly
on the assumptions made about the atmosphere (e.g., on the amount of cloud
water and on the structure of the rain layer), they show the general behavior of
microwave brightness temperatures in the presence of rain. One set of calculations,
for a nadir-vie wing instrument, is shown in Fig. 9.16.
The behavior of the brightness temperature—rain-rate curves can be under-
stood qualitatively by examining the radiative transfer equation. Using the Rayle-
igh-Jeans approximation and assuming for simplicity that we are looking straight
down (cf. Eq. 3.25),
^ = σ (Τ - Τ ) + σ,((Τ' ) - Τ ),
2 B Β Β (9.8)
where (T' ) is the directionally weighted mean brightness temperature (cf. Eq. 3.23)
B
^ = σ,(Τ-Τ ). Β (9.10)
Let's assume that Τ is nearly constant in the rain layer (T = T ) and note that A
the volume absorption coefficient is nearly zero except in the rain layer. Integrating
this simplified equation gives
T B ~ T b s t + T a ( 1 - t ) , (9.11)
9.2 Passive Microwave Techniques 337
280
OCEAN
LAND
260
ι\\ '
\Α / ^
ι \ \ 1
2*0 ^ ^ ^ - ^ 1 8 GHz
Ι
220 J1
\l / \
Ν.
\
1
\ >
200
TB (Κ)
180 || \ ^ S ^ G H z
160
140
120
> v 8 5 . 6 GHz
100
80 I I 1 1 1 1**·
10 20 30 40 50 60
RAIN RATE (mm/h)
FIGURE 9.16. Brightness temperature versus rain rate for three frequencies. [After Spencer et al. (1989).
where τ is the transmittance of the rain layer, and T is the brightness temperature BS
of the surface, that is the sum of surface emission and reflected sky brightness tem-
perature:
T BS - s T + (1 - )[T (1 - τ) + T
s E A space r]. (9.12)
Combining:
T - ετΤ + (1 - e ) r T
B 5
2
space + [1 - ετ - (1 - ε ) τ ] Τ . 2
Α (9.13)
Ignoring the T space term, Eq. 9.13 can be written
1 + ε ( £ - 1 ] τ - (1 - ε ) τ 2
(9.14)
1
A
where D is the depth of the rain layer. These equations best represent the low
frequencies (<22 GHz) where ice above the rain layer is nearly transparent.
Examination of Eq. 9.14 shows that for τ = 1 (no rain), T ~ ε Τ . As τ B δ
approaches zero, the rain layer will obscure the surface while becoming increas-
ingly visible itself. For water surfaces, T increases dramatically with rain rate
B
because ε is small (Fig. 9.17), thus the water background is cold (εΤ « 150 δ
Κ). Raining areas are easily detected because they contrast well with the cold
background. Over dry land, where ε ~ 0.9, the change in T with increasing
0 B
rain rate is small (or absent) and not useful for rainfall estimation.
a
0.65
0.60
ÜJO.55
Ο
ζ
<
h- 0.50
UJ
0.45
0 . 4 00 5 10 15 20 25 30 35 40
WATER T E M P E R A T U R E (°C)
0.9
0.8
0.7
Ö0.6
Ζ
<
1-0.5
Uj0.4
0.3
0.2
0.1
0 10 20 30 40 50 60 70 80 90
ANGLE O F INCIDENCE ( d e g r e e s )
FIGURE 9.17. (a) Nadir emittance of a smooth ocean surface as a function of sea surface temperature,
(b) Emittance of a smooth ocean surface as a function of zenith angle. [After Kidder (1979).]
9.2 Passive Microwave Techniques 339
rain rate, the depth of the rain layer must be known. Second, cloud droplets and
water vapor contribute an unknown amount to σ . 3
Increasing the rain rate or the microwave frequency causes scattering to become
more important. The second term in Eq. 9.8 reveals the effect that scattering has
on brightness temperatures (see Section 3.3). Because the scattering phase function
has the property that
(9.16)
(T' ) is simply a weighted average of the brightness temperature from all directions.
B
The scattering phase function divided by 4 π is the weight. If (T' ) is greater than
B
over both land and ocean brightness temperature decreases with increasing rain
rate (above a threshold rain rate where scattering becomes significant). Ice hydro-
meteors are particularly good at lowering brightness temperature because they
scatter but have little compensating emission. At the higher frequencies, which
are more ice-sensitive, brightness temperature decreases more rapidly with increas-
ing rain rate than at the lower frequencies.
Attempts to use microwave radiometry to estimate precipitation can be divided
into two categories: absorption schemes and scattering schemes. Schemes that
are absorption-based have the longest history and will be discussed first.
* 250 U
cc
D
OC
ui
QL
8
CC
OD
1000
R A I N F A L L R A T E (mm/hr)
I I I I I I I I I ' ' I "' T* I lift
* 250
cc
D
<
CC
UI
200
8
UI
Ζ
I-
I
Ο
£
150 h -
J I I IIII I
0.1 1 10 100 1000
R A I N F A L L R A T E (mm/hr)
FIGURE 9.18. (a) Calculated 19.35-GHz brightness temperature as a function of rain rate for freezing
levels of 1 - 5 km. (b) Brightness temperature as a function of rain rate: Nimbus 5 ESMR vs. W S R - 5 7
radar (dots) and inferred from ground-based measurements of brightness temperature and direct
measurements of rain rate (crosses). The solid line is the calculated brightness temperature for a 4-
km freezing level. The dashed lines represent departure of 1 mm h" or a factor of 2 in rain rate
1
(whichever is greater) from the calculated curve. [After Wilheit et al. (1977).]
9.2 Passive Microwave Techniques 341
FIGURE 9.19. Precipitation frequencies (percentage of days with rain) over the tropical oceans for the
period December 1 9 7 2 to February 1973 determined from Nimbus 5 ESMR measurements. [After
Kidder and Vonder Haar (1977).]
The Wilheit et al. technique, and those related to it, rely on the increase of
brightness temperature with rain rate over the ocean. They cannot determine
rain rates greater than a saturation rain rate, which decreases with increasing
microwave frequency. For this reason, lower frequencies are preferred for oceanic
precipitation estimation.
Several groups attempted to use microwave observations to map precipitation.
Kidder and Vonder Haar (1977), for example, used brightness temperature thresh-
olds to discriminate raining from nonraining pixels. They calculated precipitation
frequencies over the tropical oceans (Fig. 9.19). Rao et al. (1976) processed
all of the ESMR-5 data to produce an atlas of instantaneous rain rates over
the oceans.
The Nimbus 6 satellite carried a second ESMR (ESMR-6) that had several
important changes. First, it scanned in a cone in front of the spacecraft rather
than in a plane through nadir. The advantage of this configuration is that the
ground resolution and the viewing angle are the same for all scan spots, which
lessens problems with angle-dependent parameters such as surface emittance.
Second, it made measurements in two polarizations: horizontal and vertical 7
2451
239 248 257 267 276 286
ΤΗ (k)
FIGURE 9.20. Vertically polarized versus horizontally polarized E S M R - 6 brightness temperatures over
the southeastern United States. [After Rodgers et al. (1979).]
radiometer, of course, averages the brightness temperature over the footprint, but
because the brightness temperature is highly nonlinear in rain rate (Fig. 9.18b),
the average brightness temperature does not produce a good estimate of the mean
9.2 Passive Microwave Techniques 343
rain rate in the footprint. In fact, the rain rate estimated by using the radiometer-
measured brightness temperature will always underestimate the footprint-mean
rain rate. This underestimation can be quite serious (Lovejoy and Austin, 1980).
Again, a multiple wavelength approach may offer some relief because of different
sensitivities to rain at different wavelengths, but to make substantial improve-
ments, larger antennas are required.
9.2.2 Scattering
(9.17a)
(9.17b)
but does not scatter; it has the same effect on all polarizations. The solid line in
Fig. 9.21 shows the result of plotting T versus T for all values of τ (0 < τ <
B B
1). Although Eqs. 9.15 are nonlinear, T versus T is very nearly a straight line.
B B
through this point parallel to the no-scattering line. The parallel direction is
8
dT$ Τ -ε Τ
λ
Η
$
8
The slope of this line is = T _ ^ T·.
Β
244 Chapter 9 Precipitation
VERTICAL T B
FIGURE9.21. Polarization diagram illustrating the scattering technique of Spencer (1986). [After Spencer
et al (1989).]
chosen because changing the amount of absorbing material in the path, such as
cloud droplets and water vapor, will cause the observed point to move parallel
to the no-scattering line. The PCT is the temperature at the point that the line
intersects the no-polarization line (see Fig. 9.21). Spencer et al (1989) have begun
studies of 85.5 GHz data from the SSM/I instrument on DMSP. They calculated
PCT = 1.818 T\ - 0.818 Tg, and found that the 255-K isotherm corresponds
well with the outline of precipitating regions as determined by radar. Work
continues on ways to relate PCT to rain rate.
Scattering-based methods to estimate precipitation are too new for a realistic
assessment of accuracy, but to us they appear promising.
Soil moisture is another parameter that may be estimated using passive micro-
wave observations. As mentioned above, dry soil has a high emittance; water
surfaces have low emittances. If one adds water to the soil, the emittance falls
and becomes polarized. Grody (1983) has used channels 1 and 2 (50.30 and
53.74 GHz) of the Microwave Sounding Unit to retrieve surface emittance. In
essence, channel 2, which peaks in the lower troposphere, is used to correct the
surface-sensing channel 1 for atmospheric effects. Emittance is retrieved as a linear
combination of the two brightness temperatures. A more complicated scheme has
been investigated by Jones (1989), who combined SSM/I data, GOES data, and
soundings to retrieve soil emittance.
With a knowledge of "normal" emittance at a particular location (which
depends on soil type and vegetation), microwave observations can be used to
detect changes in emittance and therefore of soil moisture. Since soil moisture is
changed by precipitation, these emittance changes may serve as an antecedent
precipitation index (API), which is related to the precipitation that has fallen since
the last pass of the satellite. If this technique works, it will help solve one of the
major problems of microwave sensors to date: they only fly on polar-orbiting
satellites; thus observations of instantaneous rain rate are available a maximum
of four times per day, assuming two polar-orbiting satellites.
9.3 RADAR
P = 450 kW.) The pulse travels outward as a spherical shell of thickness ckt
t
(1200 m for WSR-57s), where c is the speed of light (Fig. 9.22a). The power is
not radiated in all directions, however, but in a very narrow cone (Fig. 9.22b).
For a circular paraboloid antenna, the beam width (i.e., the full width at half
maximum of the main lobe) is given by
(9.18)
where θ is the beam width in radians, λ is the wavelength of the transmitted signal,
and d is the antenna diameter (10 cm and 3.36 m, respectively, for WSR-57s). For
a perfect circular paraboloid antenna, a « 1.02, and for a practical antenna,
a « 1.26 (Battan, 1973, p. 165). The beam width of a WSR-57 is thus about 2°.
For the sake of argument, suppose that all of the radar's power is transmitted
uniformly into solid angle fl = 2π[1 - cos(0/2)]. The transmitted radiance is then
t
(9.19)
where Λ is the area of the antenna. Most weather radars are operated in so-called
nonattenuating wavelengths where absorption by atmospheric gases is negligible
and where absorption and scattering out of the transmitted beam by clouds and
precipitation are small. Thus the transmitted radiance is very nearly the radiance
received by the sampled volume (shaded in Fig. 9.22a).
The sampled volume is that portion of the spherical shell from which radiation
is being received at a particular time. If the pulse is transmitted from time t =
- | Δ ί to time t = + £Δί, then, due to the finite pulse length, at time t radiation
will be returned to the antenna from ranges between r = \c{t — \Lt) and
FIGURE 9.22. (a) Schematic diagram of the volume sampled by the radar beam, (b) Schematic
section of a radar beam from a parabolic antenna. [After Battan, L. J., Radar Observation (
Atmosphere. Univ. Chicago Press, © 1973 by the University of Chicago.]
9.3 Radar 347
r = \c{t + \At). That is, at time t the sampled volume is centered at range
r = \ct, and has depth Ar = \cAt. Because of the necessity of the two-way trip
radiation must make from the antenna to the location of interest and back to the
antenna, the sampled volume moves at half the speed and (to conserve energy)
has half the depth of the transmitted pulse. The sampled volume for a WSR-57
has a depth of about 600 m and, at a range of 100 km, has a diameter of about
3.8 km.
Assuming that the sampled volume is completely filled with particles with
volume scattering coefficient σ and scattering phase function ρ(ψ ), the radiance
δ 5
L = Ar^j
r L'p^dü' - Ar^L p(180°)i\, t (9.20)
where Ω = Air is the solid angle of the antenna subtended at the sampled
Γ
1
P = L An = !^f<rsp(l80°),
r c t (9.21)
which is one form of the radar equation. [See Battan (1973) for other forms and
for corrections for nonuniform radiance in the beam.] Note that the returned
power from a single drop would be proportional to r" , but that the returned 4
the volume sampled—and thus the number of drops scattering radiation back to
the radar—increases as r . 2
The product σ p(180°) is the backscattering cross section per unit volume. It
δ
is numerically equal to
Χ = ρ(180°) = Σ ^ β / > ( 1 8 0 ° ) ,
Β σ8 5 (9.22)
vol ^
where Q is the scattering efficiency, D is the drop diameter, and the sum is taken
s
over all drops in a unit volume. For drops in the Rayleigh size range p(180°) =
1.5, and (from Eq. 3.52),
Qs = F(5P)VL, 2
(9.23)
where Κ = (m - 1)1 [m + 2) is a weak function of temperature and wavelength,
2 1
but a strong function of phase (ice or water) Thus, for Rayleigh drops, we obtain
X =nW Eö
B
2 6
(9-24)
Λ
vol
and
(9.25)
9
Radar transmits highly polarized radiation. Since we are only considering scattering in the
backward direction, however, this nonpolarized radiative transfer equation applies.
Chapter 9 Precipitation
348
(9.27)
The reference power P is often taken to be that power which would be returned
ref
If one has an estimate of the drop size distribution and knowledge of the
velocity at which drops of a particular diameter fall, one can use the radar
reflectivity factor to calculate the precipitation rate. These relationships often take
the form
Ζ = aR , h
(9.28)
which is known as a Z-R relationship. The rainfall rate R is measured in units
of millimeters per hour (mm h" ). Z-R relationships are somewhat variable;
1
Battan (1973) states that fairly typical relationships are as follows: for stratiform
rain, Ζ = 200R - ; for orographic rain, Ζ = 3 1 R ; for thunderstorm rain,
1 6 1,71
rain drops because |iC| is much less for ice than for water. Snowflakes are also
2
not spherical, and both the fall speed and the "flake" size distribution is poorly
known. Empirical observations, however, indicate that an appropriate Z-R rela-
tionship for snow may be Ζ = 2000R (Battan, 1973). 2
σ + σ ) can be derived from Ζ with the aid of an assumed drop size distribution.
8 Λ
If measurements are made at two wavelengths, both the radar reflectivity factor
and the extinction coefficient can be retrieved. Rain rate can be estimated from
a as well as from Z; R is nearly proportional to σ .
e ε
ν = ^ = |λΔν, (9.29)
where Δν is the frequency shift. If the wind speed is 50 m s" , the frequency shift
1
on a 10-cm (3-GHz) radar is only 1000 Hz. Yet even these small shifts can be
accurately measured to yield radial drop speed and, thus, an estimate of the radial
wind speed. NEXRAD (next-generation radar), which is replacing the WSR-57
radars, is a Doppler radar.
Finally, the phase of the returned radiation can be measured. By comparing
the phases of horizontally and vertically polarized measurements, information on
drop shape can be obtained. In particular, since large drops associated with higher
rainfall rates tend to flatten as they fall, the difference in phases between the two
polarizations can be used to estimate rain rate. 10
R. J. Doviak and D . S. Zrnic, National Severe Storms Laboratory, Norman, Oklahoma, personal
1 0
communication, 1 9 8 9 .
350 Chapter 9 Precipitation
• The rapid motion of satellites means that the radar antenna may not be
looking in the right direction when the signal returns from the target pre-
cipitation.
• Narrow beam widths are required for two reasons. First, the long distance
from the target requires narrow beam widths to achieve acceptable
ground resolution. Second, unlike ground-based radar in which the beam
rises gently in the atmosphere, every beam from a satellite-borne radar in-
tercepts the surface. Narrow beam widths are required to reduce surface
clutter for range bins near the surface (Fig. 9.23). A narrow beam can be
accomplished only by means of a large antenna or operating at higher (at-
tenuating) frequencies or both.
• A low-orbiting satellite will see most points on Earth a maximum of twice
per day. Since precipitation changes on a time scale considerably shorter
than 12 h, satellite radar observations are limited either to instantaneous
or to climatological time scales.
The Tropical Rainfall Measuring Mission (TRMM), planned to fly in the late-
1990s, will carry the first space-borne, precipitation-sensing radar. It has been
designed to solve some of the above problems. TRMM is described in Chapter 11.
FIGURE 9.23. Comparison of wide and narrow beam widths. Notice that the wide beam has fewer
"good" range bins and more partially filled and surface-contaminated range bins than does the
narrow beam.
9.4 Severe Thunderstorms 351
and to process more water vapor than less intense storms. Most of the work on
severe thunderstorm detection is based on these properties.
The large vertical mass flux in a severe thunderstorm causes large divergence
in the outflow (anvil) region, which is related to the areal growth rate of the anvil:
1 dA
Ä I = v
' -
y
(· > 9 30
where V · V is the divergence of the wind velocity. The areal growth rate is
approximated by counting the number of pixels, N, colder than a temperature
threshold:
Α dt Ν dt v 1
In a modification of the work of Sikdar et al. (1970) and Sikdar (1972), Adler
and Fenn (1979a) chose not to examine the entire anvil area, which often contains
many thunderstorms in various stages of development or decay. Rather, they used
very cold GOES brightness temperature thresholds (<226 K) that apply to storm
tops which penetrate the cirrus anvil. In fact, they chose the temperature for
which 6 < Ν < 30. This area is large enough to overcome some of the sensor
resolution and response problems, but small enough to represent a single storm.
Figure 9.24 shows a thunderstorm growth rate diagram for a storm that occurred
in Nebraska and Iowa on 6 May 1975. The ordinate is the logarithm of the
number of pixels colder than each temperature threshold; the slope of each line
is (l/N)dN/dt. Severe weather reports tend to occur during or just after periods
of rapid growth. Adler and Fenn found that on this day a (l/N)dN/dt threshold
of 3.5 x 10" s , separated severe from nonsevere thunderstorms with a probabil-
3 - 1
1 1
An equally important question from this viewpoint is, H o w can satellite data be used to detect
flash flooding? This question is answered by the precipitation-estimation techniques discussed above,
some of which were designed in large part to detect flash floods.
352 Chapter 9 Precipitation
Colder storms are taller and thus more likely to be severe. Adler and Fenn
(1979a) also examined the minimum brightness temperature achieved by a storm
(T ). They found that on their case study day a T
min threshold of 212.8 Κ
min
separated severe from nonsevere storms with a POD of 0.93 and a FAR of 0.50.
Combining T and (l/N)dN/dt in a linear discriminant analysis, they found
min
that severe storms could be separated from nonsevere storms with a POD of 0.73
and a FAR of 0.31. In related work, Adler and Fenn (1981) examined 11 tornadic
storms and found that all were associated with rapid decreases in T (Fig. 9.25). min
min). In the other three cases (Fig. 9.25f—h), the tornado occurred during the
decrease. In three observations (Fig. 9.25i—k), Doppler radar showed that the
rapid T decrease was associated with the formation of a mesocyclone.
min
Lead time, or the time between classification of the storm as severe and the
first report of the occurrence of severe weather, is important for severe weather
warnings. Adler and Fenn (1979a) found that the median lead time was 24 min
using the T threshold, 27.5 min using the (l/N)dN/dt threshold, and 7 min
min
using the combined threshold. Thus these indices appear to be useful for severe
weather detection and warnings if rapid-scan (~5 min interval) satellite data
are available.
1
(a) MAY 6. 1975 *«~,(b) MAY 6. 1975 (C) MAY 6. 1975
CLOUO 5 yf , CLOUD 3b CLOUD 22
1
* 212 "MAGNET. NE. / ' OMAHA. NE WINSIDE. NE
CO
F4 F3
ff 214
^ 216
Ζ
5 218 I 1 I I I I I I I 1 1 ( 1 1 1 1 1
I 1 • * 1 l i t 1 1
-40 -20 -60 -40 -20 σ -40 -20 0 -40 -20 0
TIME (min; TIME (MIN) TIME (MIN) TIME (MIN)
MAY 6. 1975 (e) APRIL 24. 1975 MARCH 20. 1976 MARCH 20. 1976
CLOUD 38 CLOUO 18 CLOUD 7b CLOUD 23
* 208
SAUNDERS CO.,NE NEOSHO. MO. HOYLETON. IL. SADOROUS, I L
CD F1 F4 206 F2 F4
CD
206
5 208
D 20β
Ζ
5 210
I 1 I 1 I 1 I 1 I I 1 1 I I • 1 • 1 •
-40-20 0 -/20 0 *20 -40 -20 0 -40-20 0
T I M E (MIN) T I M E (MIN) T I M E (MIN) T I M E (MIN)
MAY 20. 1977 MAY 20. 1977 MAY 20. 1977
CLOUD 44 CLOUD 49 (J) CLOUD 56 (k)
JALTUS. OKJ [FORTCOB8. OK. f DEL CITY. OK.
F3 F2 F2
CD
ω
Ζ
2
I I
J 1 I 1 I 1 I I L
-40-20 0
-60 -40 -20 0 »20 -60 -40 -20 0 »20
TIME (MIN)
TIME (MIN) TIME (MIN)
FIGURE 9.25. Minimum SMS/GOES-measured blackbody temperature as a function of time for 11 tornadic storms. The times of the first
report of each tornado are indicated by dashed vertical lines. The times of the formation of mesocyclones in the three cases where
Doppler radar observations were available (i, j , k) are indicated by d a s h - d o t vertical lines. [After Adler and Fenn (1981).]
354 Chapter 9 Precipitation
gested that the minimum brightness temperature be compared with the tropopause
temperature as a way to normalize storms occurring in different environments.
Adler et al. (1985) found that indeed T - T m i n is a good indicator of severe
t r o p
thunderstorms. However, they also found that the modal brightness temperature
of the cirrus anvil, which is obtained from the satellite data rather from rawinsonde
data, is a slightly better indicator of severity. Using 5 Κ as a threshold ( T at
min
least 5 Κ colder than T ) Adler et al. found a POD of 0.79 and a FAR of 0.37.
mode
dt). They estimated the vertical velocity of the cloud top using an environmental
lapse rate:
BB
dt
w = (9.32)
In a small sample of storms Adler and Fenn found that when w exceeded 4 m
s" , the storm was likely to be severe (POD = 0.67, FAR = 0.14).
1
Index (TI). Using a threshold TI to estimate severity, they found that severe storms
could be discriminated from nonsevere storms with a POD of 0.84 and a FAR
of 0.39. The median lead time was 14 min for hail or tornadoes, and 25 min for
tornadoes only. Severe thunderstorms can display a cold V pattern (Negri, 1982;
McCann, 1983; Heymsfield and Blackmer, 1988). Adler et al. found that when
the V pattern occurs, severe weather is likely to be imminent, but many severe
storms do not display the pattern.
The GOES IR sensor is not ideally suited for severe thunderstorm detection.
The brightness temperature of cloud elements smaller than the approximately 7-
km-diameter scan spots cannot be adequately measured; for cold tops, too warm
a temperature will be measured. A second problem is the response time of the
GOES IR sensor (see Section 5.4.7.6), which causes small cloud elements to be
shifted down scan and to appear warmer than they are. Warm biases cause the
heights of the clouds to be underestimated. A solution to this problem is to make
stereoscopic height measurements using the 1-km-resolution GOES visible data.
Mack et al. (1983) used 3-min-interval stereoscopic cloud height measurements
to study thunderstorms. As expected, the stereoscopic heights were higher than
the IR heights by 1.4-6.8 km, with the larger value occurring early in the cloud's
life, and the smaller value occurring later. Interestingly, stereo ascent rates com-
pared well with IR ascent rates. Mack et al. estimate that the stereoscopic ascent
rates were accurate within about 1.3 m s" . Stereoscopic measurements would
1
clouds to yield winds, but they added surface mixing ratio observations. They
calculated the low-level moisture convergence in the prestorm environment and
found that centers of moisture convergence generally indicated the areas where
severe storms later formed. The maximum moisture convergence was 2.2 x 10" 3
g k g s" .
- 1 1
Bibliography
Adler, R. F., and Ε. B. Rodgers (1977). Satellite-observed latent heat release in a tropical cyclone.
Mon. Wea. Rev., 1 0 5 , 9 5 6 - 9 6 3 .
Adler, R. F., and D . D . Fenn (1979a). Thunderstorm intensity as determined from satellite data. / .
Appl. Meteor., 18, 5 0 2 - 5 1 7 .
Adler, R. F., and D . D . Fenn (1979b). Thunderstorm vertical velocities estimated from satellite data.
/. Atmos. Set., 3 6 , 1 7 4 7 - 1 7 5 4 .
Adler, R. F., and D . D . Fenn (1981). Satellite-observed cloud-top height changes in tornadic thunder-
storms. / . Appl. Meteor., 2 0 , 1 3 6 9 - 1 3 7 5 .
Adler, R. F., and R. A. Mack (1984). Thunderstorm cloud height-rainfall rate relations for use with
satellite rainfall estimation techniques. / . Climate Appl. Meteor., 2 3 , 2 8 0 - 2 9 6 .
Adler, R. F., Μ. J. Markus, and D. D . Fenn (1985). Detection of severe Midwest thunderstorms using
geosynchronous satellite data. Mon. Wea. Rev., 1 1 3 , 7 6 9 - 7 8 1 .
Adler, R. F., and A. J. Negri (1988). A satellite infrared technique to estimate tropical convective
and stratiform rainfall. / . Appl. Meteor., 27, 3 0 - 5 1 .
Adler, R. F., H.-Y. M. Yeh, N . Prasad, W.-K. Tao, and J. Simpson (1991). Microwave simulations
of a tropical rainfall system with a three-dimensional cloud model. / . Appl. Meteor., 3 0 , 9 2 4 - 9 5 3 .
Allison, L., E. Rodgers, T. Wilheit, and R. Fett (1974). Tropical cyclone rainfall as measured by the
Nimbus 5 electrically scanning microwave radiometer. Bull. Am. Meteor. Soc., 5 5 , 1 0 7 4 - 1 0 8 9 .
Arkin, P. A. (1979). The relationship between fractional coverage of high cloud and rainfall accumula-
tions during GATE over the B-scale array. Mon. Wea. Rev., 107, 1 3 8 2 - 1 3 8 7 .
356 Chapter 9 Precipitation
Arkin, P. Α., and B. Meisner (1987). The relationship between large-scale convective rainfall and cold
cloud over the Western Hemisphere during 1 9 8 2 - 1 9 8 4 . Mon. Wea. Rev., 115, 5 1 - 7 4 .
Atlas, D . (1982). Adaptively pointing spaceborne radar for precipitation measurements. / . Appl.
Meteor., 2 1 , 4 2 9 - 4 3 1 .
Augustine, J. Α., C. G. Griffith, W. L. Woodley, and J. G. Meitin (1981). Insights into errors of SMS-
inferred GATE convective rainfall. / . Appl. Meteor., 2 0 , 5 0 9 - 5 2 0 .
Bandeen, W. R., and I. Katz (1975). Active microwave sensing of the atmosphere. In R. E. Matthews
(ed.), Active Microwave Workshop Report, N A S A SP-376, Washington, D C , pp. 2 8 7 - 3 6 7 .
Barrett, E. C. (1970). The estimation of monthly rainfall from satellite data. Mon. Wea. Rev., 9 8 ,
322-327.
Barrett, E. C. (1973). Forecasting daily rainfall from satellite data. Mon. Wea. Rev., 1 0 1 , 2 1 5 - 2 2 2 .
Barrett, E. C. (1980). The use of satellite imagery in operational rainfall monitoring in developing
countries. In V. V. Salomonson and P. D . Bhavsar (eds.), The Contribution of Space Observations
to Water Resource Management. Pergamon Press, Oxford and N e w York, pp. 1 6 3 - 1 7 8 .
Barrett E. C , and D . W. Martin (1981). The Use of Satellite Data in Rainfall Monitoring. Academic
Press, N e w York, 3 4 0 pp.
Barrett, E. C. (1988). Hierarichal approaches to operational satellite rainfall monitoring. International
Workshop on Satellite Techniques for Estimating Precipitation, USAID, Washington, D C .
Battan, L. J. (1973). Radar Observation of the Atmosphere. Univ. Chicago Press, Chicago, 3 2 4 pp.
Bellon, Α., and G. L. Austin (1986). On the relative accuracy of satellite and rain-gauge rainfall
measurements over middle latitudes during daylight hours. / . Climate Appl. Meteor., 25,
1712-1724.
Bellon, Α., S. Lovejoy, and G. L. Austin (1980). Combining satellite and radar data for the short
range forecasting of precipitation. Mon. Wea. Rev., 1 0 8 , 1 5 5 4 - 1 5 6 6 .
Biondini, R. (1976). Cloud motion and rainfall statistics. / . Appl. Meteor., 15, 1 1 0 0 - 1 1 1 1 .
Blackmer, R. H., Jr., and S. M. Serebreny (1968). Analysis of maritime precipitation using radar data
and satellite cloud photographs. / . Appl. Meteor., 7, 1 2 2 - 1 3 1 .
Bristor, C. L., and M. A. Ruzecki (1960). TIROS 1 photographs of the Midwest storm of April 1,
1 9 6 0 . Mon. Wea. Rev., 88, 3 1 5 - 3 2 6 .
Browning, K. A. (1979). The FRONTIERS plan: A strategy for using radar and satellite imagery for
very-short-range precipitation forecasting. Meteorol. Mag., 108, 1 6 1 - 1 8 4 .
Cherna, E., A. Bellon, G. L. Austin, and A. Kilambi (1985). An objective technique for the delineation
and extrapolation of thunderstorms from GOES data. / . Geophys. Res., 9 0 , 6 2 0 3 - 6 2 1 0 .
Chesters, D., L. W. Uccellini, and W. D. Robinson (1983). Low-level water vapor fields from
the VISSR Atmospheric Sounder (VAS) "split w i n d o w " channels. / . Climate Appl. Meteor.,
22, 7 2 5 - 7 4 3 .
Del Beato, R., and S. L. Barrell (1985). Rain estimation in extratropical cyclones using GMS imagery.
Mon. Wea. Rev., 1 1 3 , 7 4 7 - 7 5 5 .
Dennis, A. S. (1963). Rainfall Determination by Meteorological Satellite Radar. Final Report to
NASA, Stanford Research Institute, Menlo Park, CA, 105 pp.
Dittberner, G. J., and Τ. H. Vonder Haar (1973). Large scale precipitation estimates using satellite
data; application to the Indian M o n s o o n . Arch. Met. Geoph. Biokl. Ser. B, 2 1 , 3 1 7 - 3 3 4 .
Doneaud, Α. Α., J. R. Miller, Jr., L. R. Johnson, Τ. H. Vonder Haar, and P. Laybe (1987). The
area-time integral technique to estimate convective rain volumes over areas applied to satellite
d a t a - a preliminary investigation. / . Climate Appl. Meteor., 26, 1 5 6 - 1 6 9 .
Doneaud, Α. Α., S. Ionescu-Niscov, D . L. Priegnitz, and P. L. Smith (1984). The area-time integral
as an indicator for convective rain volumes. / . Climate Appl. Meteor., 2 3 , 555-561.
Doviak, R. J., and D . S. Zrnic (1984). Doppler Radar and Weather Observations. Academic Press, Or-
lando.
Follansbee, W. A. (1973). Estimation of Average Daily Rainfall from Satellite Cloud Photographs.
N O A A Tech. M e m o . NESS 4 4 , Washington, D C , 3 9 pp.
Follansbee, W. Α., and V. J. Oliver (1975). A Comparison of Infrared Imagery and Video Pictures
in the Estimation of Daily Rainfall from Satellite Data. N O A A Tech. M e m o . NESS 6 2 , Washington,
D C , 14 pp.
9.4 Severe Thunderstorms 357
Follansbee, W. A. (1976). Estimation of Daily Precipitation over China and the USSR Using Satellite
Imagery. N O A A Tech. M e m o . NESS 8 1 , Washington, D C , 3 0 pp.
Fujita, Μ., K. Okamoto, H. Masuko, and N . Fugono (1985a). Quantitative measurements of path-
integrated rain rate by an airborne microwave radiometer over the ocean. / . Atmos. Ocean. Tech.,
2, 2 8 5 - 2 9 2 .
Fujita, Μ., K. Okamoto, S. Yoshikado, and K. Nakamura (1985b). Inference of rain rate profile and
path-integrated rain rate by an airborne microwave rain scatterometer. Radio Sei., 2 0 , 6 3 1 - 6 4 2 .
Garcia, O. (1981). A comparison of t w o satellite rainfall estimates for GATE. / . Appl. Meteor.,
20, 4 3 0 - 4 3 8 .
Griffith, C. G., and W. L. Woodley (1973). O n the variation with height of the top brightness of
precipitating convective clouds. / . Appl. Meteor., 12, 1 0 8 6 - 1 0 8 9 .
Griffith, C. G., W. L. Woodley, S. Browner, J. Teijeiro, M. Maier, D. W. Martin, J. Stout, and D .
N . Sikdar (1976). Rainfall Estimation from Geosynchronous Satellite Imagery during Daylight
Hours. N O A A Tech. Rep. ERL 3 5 6 - W M P O 7, Boulder, C O , 106 pp.
Griffith, C. G., W. L. Woodley, P. G. Grube, D . W. Martin, J. Stout, and D . N . Sikdar (1978). Rain
estimation from geosynchronous satellite imagery-visible and infrared studies. Mon. Wea. Rev.,
106, 1 1 5 3 - 1 1 7 1 .
Griffith, C. G., W. L. Woodley, J. S. Griffin, and S. C. Stromatt (1980). Satellite-Derived Precipitation
Atlas for the GARP Atlantic Tropical Experiment. Division of Public Documents, U.S. Government
Printing Office, Washington, D C , 2 8 4 pp.
Griffith, C. G., J. A. Augustine, and W. L. Woodley (1981). Satellite rain estimation in the U.S. High
Plains. / . Appl. Meteor., 2 0 , 5 3 - 5 6 .
Griffith, C. G. (1987a). The Estimation from Satellite Imagery of Summertime Rainfall over Varied
Space and Time Scales. N O A A Tech. M e m o . ERL ESG-25, Boulder, C O , 102 pp.
Griffith, C. G. (1987b). Comparisons of gauge and satellite rain estimates for the central United States
during August 1 9 7 9 . / . Geophys. Res., 9 2 , 9 5 5 1 - 9 5 6 6 .
Grody, N . C. (1983). Severe storm observations using the Microwave Sounding U n i t . / . Climate Appl.
Meteor., 2 2 , 6 0 9 - 6 2 5 .
Gruber, A. (1973a). An Examination of Tropical Cloud Clusters Using Simultaneously Observed
Brightness and High Resolution Infrared Data from Satellites. N O A A Tech. M e m o . NESS 5 0 ,
Washington, D C , 2 2 pp.
Gruber, A. (1973b). Estimating rainfall in regions of active convection./. Appl. Meteor., 1 2 , 1 1 0 - 1 1 8 .
Hakkarinen, I. M., and R. F. Adler (1988). Observations of convective precipitation at 9 2 and 183
GHz: Aircraft results. Meteor. Atmos. Phys., 1 6 4 - 1 8 2 .
Heymsfield, G. M., and R. H. Blackmer, Jr. (1988). Satellite-observed characteristics of Midwest
severe thunderstorm anvils. Mon. Wea. Rev., 116, 2 2 0 0 - 2 2 2 4 .
Hillger, D . W., and Τ. H. Vonder Haar (1981). Retrieval and use of high-resolution moisture and
stability fields from Nimbus 6 HIRS radiances in pre-convective situations. Mon. Wea. Rev.,
109, 1 7 8 8 - 1 8 0 6 .
Huff, F. A. (1967). Rainfall gradients in warm season rainfall. / . Appl. Meteor., 6, 4 3 5 - 4 3 7 .
Jones, A. S. (1989). Microwave Remote Sensing of Cloud Liquid Water and Surface Emittance over
Land Regions. M.S. thesis, Dept. Atmospheric Science, Colorado State University, Ft. Collins,
145 pp.
Kidder, S. Q., and Τ. H. Vonder Haar (1977). Seasonal oceanic precipitation frequencies from Nimbus
5 microwave data. / . Geophys. Res., 8 2 , 2 0 8 3 - 2 0 8 6 .
Kidder, S. Q. (1979). Determination of Tropical Cyclone Surface Pressure and Winds from Satellite
Microwave Data. Atmospheric Science Paper N o . 3 0 7 , Colorado State University, Fort Collins,
87 pp.
Kilonsky, B. J., and C. S. Ramage (1976). A technique for estimating tropical open-ocean rainfall
from satellite observations. / . Appl. Meteor., 15, 9 7 2 - 9 7 5 .
King, P. (1984). Test results from RAINS AT. Nowcasting II: Mesoscale Observations and Very-
Short-Range Weather Forecasting. Proceedings of 2nd International Symposium on Nowcasting,
ESA SP-208, ESA Scientific &c Technical Publications Branch, Noordwijk, Netherlands, pp.
139-144.
358 Chapter 9 Precipitation
Kuo, H. L. (1965). On formation and intensification of tropical cyclones through latent heat release
by cumulus convection. / . Atmos. Sei., 2 2 , 4 0 - 6 3 .
Lethbridge, Μ. (1967). Precipitation probability and satellite radiation data. Mon. Wea. Rev., 9 5 ,
487-490.
Lopez, R. E., J. Thomas, D . O. Blanchard, and R. L. Holle (1983). Estimation of rainfall over an
extended region using only measurements of the area covered by radar echoes. Preprints: 21st
Conf. Radar Meteor., American Meteorological Society, Boston, pp. 6 8 1 - 6 8 6 .
Lovejoy, S., and G. L. Austin (1979a). The delineation of rain areas from visible and IR satellite data
for GATE and mid-latitudes. Atmosphere-Ocean, 17, 7 7 - 9 2 .
Lovejoy, S., and G. L. Austin (1979b). The sources of error in rain amount estimating schemes for
GOES visible and IR satellite data. Mon. Wea. Rev., 107, 1 0 4 8 - 1 0 5 4 .
Lovejoy, S., and G. L. Austin (1980). The estimation of rain from satellite-borne microwave radiome-
ters. Q. / . Roy. Meteor. Soc, 106, 2 5 5 - 2 7 6 .
Mack, R. Α., A. F. Hasler, and R. F. Adler (1983). Thunderstorm cloud top observations using
satellite stereoscopy. Mon. Wea. Rev., I l l , 1 9 4 9 - 1 9 6 4 .
Maddox, R. Α., and Τ. H. Vonder Haar (1988). The use of satellite observations. Chapter 12 of E.
Kessler (ed.), Instruments and Techniques for Thunderstorm Observations and Analysis. Univ.
Oklahoma Press, Norman, 1 9 1 - 2 1 4 .
Martin, D . W., and M. R. Howland (1986). Grid history: A geostationary satellite technique for
estimating daily rainfall in the tropics. / . Climate Appl. Meteor., 2 5 , 1 8 4 - 1 9 5 .
McCann, D . W. (1983). The enhanced-V: A satellite observable severe storm signature. Mon. Wea.
Rev., I l l , 8 8 7 - 8 9 4 .
Meneghini, R., J. Eckerman, and D. Atlas (1983). Determination of rain rate from a spaceborne radar
using measurements of total attenuation. IEEE Trans Geosci. Remote Sens. GE-21, 3 4 - 4 3 .
Meneghini, R., and D . Atlas (1986). Simultaneous ocean cross section and rainfall measurements
from space with a nadir-looking radar. / . Atmos. Ocean. Tech., 3 , 4 0 0 - 4 1 3 .
Moses, J. F., and E. C. Barrett (1986). Interactive procedures for estimating precipitation from satellite
imagery. Proceedings of the Workshop on Hydrologie Applications of Space Technology held in
Cocoa Beach, Florida, August 1985. IAHS Publ. N o . 160, pp. 2 5 - 3 9 .
Mugnai, Α., and E. A. Smith (1984). Passive microwave radiation transfer in an evolving cloud
medium. In G. Fiocco (ed.), IRS '84: Current Problems in Atmospheric Radiation, A. Deepak
Publishing, Hampton, VA, pp. 2 9 7 - 3 0 0 .
Mugnai, Α., and E. A. Smith (1988). Radiative transfer to space through a precipitating cloud at
multiple microwave frequencies. Part I: Model description. / . Appl. Meteor., 27, 1 0 5 5 - 1 0 7 3 .
Mugnai, Α., Η. J. Cooper, E. A. Smith, and G. J. Tripoli (1990). Simulation of microwave brightness
temperatures of an evolving hailstorm at SSM/I frequencies. Bull. Am. Meteor. Soc, 7 1 , 2 - 1 3 .
Nagle, R. E., and S. M. Serebreny (1962). Radar precipitation echo and satellite cloud observations
of a maritime cyclone. / . Appl. Meteor., 1, 2 7 9 - 2 9 5 .
NASA (1988). Preliminary Project Plan for the Tropical Rainfall Measuring Mission (TRMM). Phase
A, Part I: Technical Plan. NASA/Goddard Space Flight Center, Greenbelt, M D .
Negri, A. J., and Τ. H. Vonder Haar (1980). Moisture convergence using satellite-derived wind fields:
A severe local storm case study. Mon. Wea. Rev., 108, 1 1 7 0 - 1 1 8 2 .
Negri, A. J., and R. F. Adler (1981). Relation of satellite-based thunderstorm intensity to radar-
estimated rainfall. / . Appl. Meteor., 2 0 , 2 8 8 - 3 0 0 .
Negri, A. J. (1982). Cloud-top structure of tornadic storms on 10 April 1979 from rapid scan and
stereo satellite observations. Bull. Am. Meteor. Soc, 6 3 , 1 1 5 1 - 1 1 5 9 .
Negri, A. J., R. F. Adler, and P. J. Wetzel (1984). Rain estimation from satellites: An examination
of the Griffith-Woodley Technique. / . Climate Appl. Meteor., 2 3 , 1 0 2 - 1 1 6 .
Negri, A. J., and R. F. Adler (1987a). Visible and infrared rain estimation. I: A grid cell approach.
/. Climate Appl. Meteor., 1 5 5 3 - 1 5 6 4 .
Negri, A. J., and R. F. Adler (1987b). Visible and infrared rain estimation. II: A cloud definition
approach. / . Climate Appl. Meteor., 1565-1576.
9.4 Severe Thunderstorms 359
Peslen, C. Α. (1980). Short-interval SMS wind vector determinations for a severe local storms area.
Mon. Wea. Rev., 108, 1 4 0 7 - 1 4 1 8 .
Prabhakara, C , G. Dalu, G. L. Liberti, J. J. Nucciarone, and R. Suhasini (1992). Rainfall estimation
over oceans from S M M R and S M M / I microwave data. / . Appl. Meteor., 3 1 , 5 3 2 - 5 5 2 .
Rao, M. S. V., and J. S. Theon (1977). N e w features of global climatology revealed by satellite-
derived oceanic rainfall maps. Bull. Am. Meteor. Soc., 5 8 , 1 2 8 5 - 1 2 8 8 .
Rao, M. S. V., W. V. Abbott, and J. S. Theon (1976). Satellite-Derived Global Oceanic Rainfall Atlas
{1973 and 1974). NASA SP-410, Washington, D C , 31 pp. plus appendixes.
Reynolds, D . W., and E. A. Smith (1979). Detailed analysis of composited digital radar and satellite
data. Bull. Am. Meteor. Soc, 6 0 , 1 0 2 4 - 1 0 3 7 .
Reynolds, D . W. (1980). Observations of damaging hailstorms from geosynchronous satellite digital
data. Mon. Wea. Rev., 108, 3 3 7 - 3 4 8 .
Richards, F., and P. Arkin (1981). On the relationship between satellite-observed cloud cover and
precipitation. Mon. Wea. Rev., 1 0 9 , 1 0 8 1 - 1 0 9 3 .
Rodgers, Ε. Β., H. Siddalingaiah, A. T. C. Chang, and Τ. T. Wilheit (1979). A statistical technique
for determining rainfall over land employing Nimbus 6 ESMR measurements. / . Appl. Meteor.,
18, 9 7 8 - 9 9 1 .
Savage, R. C , and J. A. Weinman (1975). Preliminary calculations of the upwelling radiance from
rainclouds at 3 7 . 0 and 19.35 GHz. Bull. Am. Meteor. Soc, 56, 1 2 7 2 - 1 2 7 4 .
Savage, R. C. (1978). The radiative properties of hydrometeors at microwave frequencies. / . Appl.
Meteor., 17, 9 0 4 - 9 1 1 .
Scofield, R. Α., and V. J. Oliver (1977). A Scheme for Estimating Convective Rainfall from Satellite
Imagery. N O A A Tech. M e m o . NESS 86, Washington, D C , 4 7 pp.
Scofield, R. A. and L. E. Spayd Jr. (1984). A Technique That Uses Satellite, Radar, and Conventional
Data for Analyzing and Short-Range Forecasting of Precipitation from Extratropical Cyclones.
N O A A Tech. M e m o . NESDIS 8, Washington, D C , 51 pp.
Scofield, R. A. (1987). The NESDIS operational convective precipitation estimation technique. Mon.
Wea. Rev., 1 1 5 , 1 7 7 3 - 1 7 9 2 .
Shifrin, K. S., and Μ. M. Chernyak (1968). Microwave absorption and scattering by precipitation.
In Transfer of Microwave Radiation in the Atmosphere, Israel Program for Scientific Translations,
pp. 6 9 - 7 8 .
Sikdar, D . Ν . , V. E. Suomi, and C. E. Anderson (1970). Convective transport of mass and energy in
severe storms over the United States—an estimate from a geostationary altitude. Tellus, 2 2 ,
521-532.
Sikdar, D . N . (1972). A T S - 3 observed cloud brightness field related to a meso-to-subsynoptic scale
rainfall pattern. Tellus, 2 4 , 4 0 0 - 4 1 3 .
Simpson, J., and V. Wiggert (1969). Models of precipitating cumulus towers. Mon. Wea. Rev.,
97, 4 7 1 - 4 8 9 .
Simpson, J., R. F. Adler, and G. R. North (1988). A proposed tropical rainfall measuring mission
(TRMM) satellite. Bull. Am. Meteor. Soc, 6 9 , 2 7 8 - 2 9 5 .
Smith, Ε. Α., and A. Mugnai (1988). Radiative transfer to space through a precipitating cloud at
multiple microwave frequencies. Part II: Results and analysis. / . Appl. Meteor., 27, 1 0 7 4 - 1 0 9 1 .
Smith, Ε. Α., and A. Mugnai (1989). Radiative transfer to space through a precipitating cloud at
multiple microwave frequencies. Part III: Influence of large ice particles. / . Meteor. Soc. Japan,
67, 7 3 9 - 7 5 5 .
Smith, Ε. Α., A. Mugnai, H. J. Cooper, G. J. Tripoli, and X. Xiang (1992). Foundations for
statistical-physical precipitation retrieval from passive microwave satellite measurements. Part
I: Brightness-temperature properties of a time-dependent cloud radiation model. / . Appl.
Meteor., 3 1 , 5 0 6 - 5 3 1 .
Spayd, L. E., Jr. and R. A. Scofield (1984). A Tropical Cyclone Precipitation Estimation Technique
Using Geostationary Satellite Data. N O A A Tech. M e m o . NESDIS 5, Washington, D C , 36 pp.
Spencer, R. W., W. S. Olson, R. W u , D . W. Martin, J. A. Weinman, and D . A. Santek (1983a).
360 Chapter 9 Precipitation
World Climate Research Programme (1986). Global Large-Scale Precipitation Data Sets for the
WCRP. W C P - 1 1 1 , W M O / T D N o . 9 4 , Geneva, Switzerland.
Wu, R., and J. A. Weinman (1984). Microwave radiances from precipitating clouds containing
aspherical ice, combined phase, and liquid hydrometeors. / . Geophys. Res., 89, 7 1 7 0 - 7 1 7 8 .
W u , R., J. A. Weinman, and R. T. Chin (1985). Determination of rainfall rates from GOES satellite
images by a pattern recognition technique. / . Atmos. Ocean. Tech., 2, 3 1 4 - 3 3 0 .
Wylie, D . P. (1979). An application of a geostationary satellite rain estimation technique to an
extratropical area. / . Appl. Meteor., 18, 1 6 4 0 - 1 6 4 8 .
10
Earth
Radiation
Budget
The solar energy reaching the Earth is traditionally quantified as the solar
constant (see Section 3.7). To be precise, the solar constant is the annual average
solar irradiance received outside the Earth's atmosphere (at the top of the atmo-
363
364 Chapter 10 Earth Radiation Budget
sphere) on a surface normal to the incident radiation and at the Earth's mean
distance from the sun. The actual solar irradiance varies by ±3.4% from the
solar constant during the year due to the eccentricity of the Earth's orbit about
the sun, but it varies as a result of other causes also. Ground-based measurements
of the solar constant date back to 1913. The problem with ground-based measure-
ments is that they must be corrected for the effects of the atmosphere. Space-
based measurements require no such correction; they date from 1975.
Both the Nimbus 6 and 7 satellites carried an Earth Radiation Budget (ERB)
experiment (Smith et al., 1975; Jacobowitz et al., 1978). Part of the instrument
was designed to monitor the sun's output. Once each orbit, a 10-channel radiome-
ter viewed the sun for approximately 3 min. The 10 channels were designed to
measure the solar irradiance in several spectral intervals. On the Nimbus 7 ERB,
channel 10 was replaced with channel 10C, an electrically self-calibrating cavity
radiometer, which measures the total solar irradiance in the spectral range from
less than 0.2 μνα to greater than 50 /xm (Hickey et al., 1988a). The Nimbus
7 10C data began on 16 November 1978 and are still being collected as of
this writing.
The Solar Maximum Mission satellite carried an active cavity radiometer
[ACRIM; Willson (1979, 1984), Willson and Hudson (1988)]. Launched 16
February 1980, SMM made solar irradiance measurements until December 1989.
The Earth Radiation Budget Experiment (see Chapter 4) has flown on the
Earth Radiation Budget Satellite (launched 5 October 1984) as well as the NOAA
9 and 10 satellites (launched 12 December 1984 and 17 September 1986, respec-
tively).
Figure 10.1 shows the annually averaged Nimbus 7 ERB and Solar Max ACRIM
solar constant measurements. Shown for comparison is the annually averaged
sunspot number calculated at the Zurich Observatory. The solar constant seems
to follow the sunspot cycle. The mean Nimbus 7 solar constant, measured over
an 11-year solar cycle, is 1372.0 W m~ with a standard deviation of 0.6 W m~ .
2 2
The mean Solar Max solar constant over a period shorter than a solar cycle is
1367.5 W m " with a standard deviation of 0.4 W m . The ERBE measurements
2 - 2
however, that the measurements by the instruments are highly correlated. Thus
changes in the annually averaged solar constant can be determined quite accu-
rately, perhaps within a tenth of a watt per square meter, even though the absolute
magnitude can be determined only within a few watts per square meter.
The reader should note, however, that the above numbers must be characterized
as preliminary for several reasons.
1 374
200
a.
LU
CD
1 50 D
Ο
Q_
1 00 00
Ζ
D
(/)
X
u
D
Μ
1 366
7 8 7 9 8 0 81 8 2 8 3 8 4 8 5 8 6 8 7 8 8 8 9 9 0 91
FIGURE 10.1. Annual average solar constant (W m ) as measured by channel 10C of the Nimbus 7
- 2
ERB and by the Solar M a x i m u m Mission ACRIM. The dashed line is the annual average Zurich
sunspot number. The ERB and ACRIM data are from Table 4 of Hoyt et al. (1992). The sunspot
data are courtesy of the N O A A / N E S D I S / N a t i o n a l Geophysical Data Center, Boulder, Colorado.
The goal of radiation budget studies is to measure the incoming and outgoing
radiation as a function of time and space. Incoming radiation is irradiance and
366 Chapter 10 Earth Radiation Budget
which is also called the outgoing longwave radiation (OLR). With these quantities
we can calculate the albedo
Α ( ί > Θ ) ψ ) = Μ ^ Ψ ) ? ( 1 ( U )
•^sunMsun
and £ = E (f) is the solar constant S adjusted for distance from the sun,
s u n sun sun
E
-" -(ÄJ' =S
( ι ο
· >
4
where d {t) is the Earth-sun distance at time t, and d is the mean Earth-sun
sun sun
Since angular resolution for a sensor is stated in terms of the half-power angles,
flat-plate sensors have a 120° angular resolution. The ERBE and Nimbus 7 ERB
both have flat-plate sensors. An instrument which employs flat-plate sensors is
also called a nonscanner because nadir-pointing, flat-plate sensors detect radiation
from horizon to horizon without scanning.
NFOV sensors measure radiation in a narrow cone. AVHRR is a NFOV
instrument (Fig. 4.6), and the scanning instruments on the Nimbus 6 and 7 ERB
and on the ERBE are also NFOV sensors, but they are not as narrow as the
AVHRR sensor. An instrument that employs a NFOV sensor is often called a
scanner because to monitor the Earth disk, the sensor must scan.
Flat-plate sensors have few moving parts; thus they can operate for a decade
or longer, which makes them useful for climate studies. NFOV scanners typically
wear out in a few years. On the other hand, scanners make high-spatial-resolution
measurements, which are useful for regional radiation budget studies. WFOV
sensors make measurements which have been integrated over large areas, typically
a circle of radius 2500-3500 km at the surface. WFOV measurements are useful
for global or large-scale studies.
A polar-orbiting satellite is a very good platform for measuring the radiation
budget. A satellite-mounted instrument makes two observations daily, separated
by about 12 h, of every point on Earth. These are sufficient to calculate the
monthly average radiation budget, which is usually the time scale of interest.
There are some problems, however:
• The measurements are made at the satellite height, not at the top of the at-
mosphere (the inverse problem).
• The twice-daily observations may not adequately sample the diurnal varia-
tion, particularly of shortwave radiation. A diurnal model may be neces-
sary (the diurnal problem).
• Radiation budget quantities which have been integrated over wavelength,
shortwave or longwave, are necessary. Satellite sensors, however, do not
measure exactly these quantities (the spectral correction problem or unfil-
tering problem).
• To measure the radiant exitance from a point, one must observe the point
from all possible directions. Satellites, however, observe a single point
from a single direction (the angular dependence problem).
In summary, to retrieve accurate radiation budget quantities, we must know the
radiance in all directions, at all times, in all wavelengths, and at all heights.
Satellites can observe a larger fraction of the necessary data than other methods,
but not all. We must use modeling to fill the gaps in the measurements. The three
radiation budget data sets described below differ in their handling of these
problems.
Since 1974, with one gap, NOAA/NESDIS has derived radiation budget quanti-
ties from its polar-orbiting satellites. The data set has been constructed using data
3^8 Chapter 10 Earth Radiation Budget
from several satellites and two different instruments: (1) the Very High Resolution
Radiometer (VHRR), which flew on NOAA satellites prior to NOAA 6; and (2)
the AVHRR, since the launch of TIROS N. Retrieval of VHRR-based radiation
budget products between June 1974, and March 1978, is described by Gruber
(1978). Retrieval of AVHRR-based radiation budget products since January 1979,
is similar and is described by Gruber et al. (1983). Here we summarize the retrieval
scheme of Gruber et al. (1983).
Afternoon satellites (see Table 4.2) are used so as to have the highest solar
elevation angle and to avoid, to the extent possible, missing data at high latitudes
in the winter hemisphere. The five-channel, 1-km-resolution AVHRR data are
too voluminous to all be processed, especially since the final radiation budget
products have roughly 200-km resolution. Two of the five channels are processed:
channel 1 (0.58-0.68 μπι) represents solar wavelengths, and channel 5 (11.5-12.5
μπι) represents infrared wavelengths. Channel 5 was chosen because it is more
sensitive to water vapor than channel 4; thus it is thought to better represent the
outgoing longwave radiation. However, some satellites do not have channel 5;
channel 4 (10.5-11.5 /xm) is used for these satellites.
Next, to reduce data processing, global area coverage (GAC) data are used
instead of the full-resolution local area coverage (LAC) data. The scheme to
produce GAC data from the 1-km-resolution data is as follows. Along a scan
line, four contiguous scan spots are averaged, and a fifth one is skipped. Two
scan spots at each end of the 2048-element scan line are ignored. This results in
a 409-element scan line. The next two scan lines are skipped entirely, and the
fourth scan line is averaged as above. This results in a data set which at nadir
has about 4-km resolution in the cross-track direction and 3.3-km resolution in the
along-track direction. To further reduce processing, the GAC data are processed in
1 1 x 1 1 blocks. These blocks are called GAC targets. At nadir, they are about
36 km along-track by 44 km cross-track. If 60 or more of the 121 GAC scan
spots are present, the counts for all scan spots in a GAC target are averaged, and
the average counts in the two channels represent one radiation budget observation.
If fewer than 60 scan spots are present (due to data loss), the entire block is
treated as missing.
With the average count for a GAC target calculated, the visible count is con-
verted into an unnormalized albedo using a linear equation determined in the
prelaunch calibration. The infrared count is converted into a radiance using on-
board blackbody and space observations (Lauritson et al., 1979; see also Chapter
4). At this point, the data consist of calibrated, spatially averaged radiometric
quantities at the altitude of the satellite.
The next step converts the data into radiation budget estimates. The visible
albedo is normalized by dividing by the cosine of the solar zenith angle at the
center of the GAC target. The assumptions in this calculation are that (1) all
reflected radiation is isotropic (and thus that one viewing angle is sufficient to
estimate the total reflected radiation in all directions), (2) the albedo in the entire
solar spectrum is the same as the albedo measured in the narrow spectral band
of the AVHRR channel 1, and (3) the measured albedo is the daily average albedo.
10.2 Top of the Atmosphere Radiation Budget 3^9
These assumptions cause errors in the retrieved albedo, but they produce useful
albedo estimates, and the data processing is much simpler than in more accurate
schemes discussed below. Ascending and descending portions of the orbit are
processed separately. Albedo is only calculated for the daylight half orbit. In
addition, the available solar radiation (/x E , averaged over the entire day) is
sun sun
calculated as
C /-sunset / 1 \2
Ε
™ =
-Γ\ \ΤΊ^ J^u (t)dt, n (10.5)
*Ό •'sunrise V^sunW/
where t is the time in seconds, t is the day length (86,400 s), and the solar
O
M L W = σ TJ = OLR. (10.9)
OLR values are calculated and stored separately for ascending and descending
half orbits; thus two OLR observations of any point are made each day. In short,
both angular and spectral corrections are made to the OLR, and two observations
per day are available to estimate the daily average radiant exitance.
The final step in the procedure is to map the data. They are mapped onto two
125 x 125 polar stereographic maps (arrays), one Northern Hemisphere and one
Southern Hemisphere, which have about 100-km resolution at the equator and
200-km resolution at the poles. Each GAC target is assigned to the box in which
its center falls, and all values in the box are averaged for each day. From the
polar stereographic arrays, a 2.5° x 2.5° latitude—longitude map is constructed
using bilinear interpolation. The four parameters mapped and archived are OLR
daytime, OLR nighttime, absorbed solar radiation, and available solar radiation
Chapter 10 Earth Radiation Budget
370
( E , which is not archived in 2.5° x 2.5° format). The data are available from
sun
the Satellite Data Services Division of NESDIS. The OLR data have appeared in
1
the Nimbus 7 ERB radiation budget parameters have changed since 1985; thus
these accuracy figures may no longer be applicable.
The shortwave parameters (albedo, absorbed solar radiation, and thus net
radiation) are considered to be less accurate than the OLR. The main reason for
this conclusion is that the OLR estimates include angular and spectral corrections
while the shortwave parameters do not. We know of no published accuracy
estimates for the shortwave parameters.
The primary advantage to the NOAA radiation budget data set is that it is a
long-term data set which will continue for many years. The OLR data have been
used for many purposes and are a very valuable tool for climate studies.
The Earth Radiation Budget (ERB) instrument, flown on the Nimbus 6 and 7
satellites, was the first radiation budget instrument to combine nonscanning,
WFOV instruments with scanning, NFOV instruments on the same satellite. This
combination has become standard due to the complimentary properties the two
2
designs. The ERB is not the most accurate radiation budget instrument to have
flown (that honor belongs to the Earth Radiation Budget Experiment to be dis-
cussed in the next section), but it produced a long-term data set including the best
angular dependence models assembled to date. Its history is interesting because it
demonstrates the difficulty and challenge of making laboratory-quality measure-
ments from space. Kyle et al. (1993) review the 20-year history of the ERB in-
strument.
The Earth Radiation Budget (ERB) instrument was designed to be flown on
the Nimbus 6 satellite. It was launched 12 June 1975. A second ERB was flown
on the Nimbus 7 satellite, launched 24 October 1978. (See the Nimbus 6 and
Nimbus 7 User's Guides.) Each ERB is a 22-channel instrument. Channels 1-10
directly observe the sun to measure the solar flux density in several bands (see
Section 10.1). Channels 11-14 are nonscanning, WFOV (flat-plate) channels.
Channels 11 and 12 measure total flux density (0.2 to >50 μ,ιη). Channel 12 is
1
Kidwell (1991) describes the positioning of the maps and the format of the data tapes.
2
The EOS may not have W F O V sensors.
10.2 Top of the Atmosphere Radiation Budget 371
et al., 1993). The technique was originally developed by Raschke et al. (1973)
to process Nimbus 3 data, and for shortwave radiation it may be derived from
basic principles as follows.
Recall from Eq. 3.63 that the radiance reflected from a surface or scene is
given by
J
-2TT [π/2
(10.10)
ο j
o
sphere, Lj all comes from essentially one direction ( 0 , φ ) , so that §ιιη διιη
or
γ (0 ,φ ; 0 , , 0
Γ Γ Γ u n l i m ) c o s 0 s m M M 0 r - (10.13)
r
Since the solar radiation incident on the scene is p E , sun sun the albedo (A = M /E ) r {
must be
J
-2TT rn/2
γ (0 ,φ ; 0
Γ Γ Γ s u n ^ s u n )cos0 sin0 J0 ^ . r r r r (10.14)
0 J
o
Dividing the radiant exitance (Eq. 10.13) by the observed radiance (Eq. 10.11)
and substituting Eq. 10.14 gives
Λ/Γ — _ ΉL _ 7rL
r r
(10.15)
where ξ is the anisotropic reflectance factor (Eq. 3.67). Thus, with a knowledge
γ
land-ocean (coastal areas). They also chose four cloud categories: clear ( 0 - 5 %
cloud cover), partly cloudy (5-50%), mostly cloudy (50-95%), and overcast
(95-100%). There would be 20 scene types if all types of clouds were allowed over
all surfaces. However, Suttles et al. allowed only one overcast scene (regardless
of surface type), they eliminated the difficult-to-discriminate partly-cloudy-over-
snow and mostly-cloudy-over-snow categories, and they combined land and desert
for partly and mostly cloudy situations. Thus they arrived at 12 scene types: clear
over land, ocean, snow, desert, and land-ocean mix; partly cloudy over ocean,
land or desert, and land-ocean mix; mostly cloudy over ocean, land or desert,
and land-ocean mix; and overcast. Nimbus 7 ERB NFOV observations between
1 April 1979 and 22 June 1980 were binned in one of these scene types using the
Nimbus 7 THIR/TOMS cloud algorithm (see Chapter 8) to determine cloudiness.
The Nimbus 7 ERB NFOV channels could scan in many directions, including
parallel to the satellite subtrack (see Jacobowitz etal., 1978). This made it possible
for the ERB to view a given location from several viewing angles as it approached,
passed over, and receded from the location. For each of the scene types, the
NFOV data were further binned into 10 solar zenith angle bins, into seven viewing
zenith angle bins, and into eight relative azimuth angle bins (for a total of 5880 4
bins, when all azimuth angles were pooled for the vertical viewing angle). The
azimuth angles, </> , were relative to the sun's azimuth and were assumed to be
R
Earth-sun distance and to overhead sun [i.e., the radiances were multiplied by
(d /d )
sun sun sec 0 ] . The average radiance, L, was then calculated in each bin. For
2
sun
each surface type and for each solar zenith angle, the radiant exitance was calcu-
lated by integrating over viewing zenith angle and relative azimuth angle using
Eq. 10.12. The anisotropic factor is then given by
^ ( ^ ^ s u n ^ R ) - ^ f ; U n
f R )
- (10.16)
The albedo for each scene type and each solar zenith angle was calculated by
Α ( 0 =^ # ^ · (10.17)
^sun
(The solar constant was used because the radiances were normalized to overhead
sun and to the mean Earth-sun distance.)
Since Nimbus 7 is a sunsynchronous satellite with approximately a noon equa-
tor crossing time, not all solar zenith angles were observed for every surface type.
Deserts, for example, occur mainly in the subtropics, but Nimbus 7 can observe
high solar zenith angles only in the high latitudes. Empty angular bins were
filled using several techniques. The most important is the Helmholtz Reciprocity
4
Defined as the difference between the sun's azimuth and the viewing azimuth such that reflection
away from the sun has a zero relative azimuth, and reflection toward the sun has a 180° relative azimuth.
Chapter 10 Earth Radiation Budget
374
Principle, which states that the bidirectional reflectance is invariant when the
incoming and outgoing angles are interchanged:
(10.19)
5
N o t e that although reciprocity was used to fill empty bins, the final models do not satisfy
reciprocity in general. (See Green et al. (1990a).)
10.2 Top of the Atmosphere Radiation Budget 375
et al., 1985). Also, improved calibrations had been found, and improved algo-
rithms had been developed for the ERBE instrument. Therefore, the Nimbus 7
ERB data were reprocessed with the ERBE algorithms (Kyle et al., 1993).
The NFOV data (May 1979-May 1980) were processed in two ways to yield
albedo, OLR, and net radiation. The first method is the traditional method devel-
oped by Raschke et al. (1973) and refined into the ERBE algorithm. The corrected
NFOV radiances were binned daily into 18,630 boxes each approximately (166
km) and covering the globe. Each of these boxes is called a subtarget area. Each
2
of the radiances was transformed into a radiant exitance (M = ττΤΙξ^ using the r
suming that the radiances for each cloud-cover category are described by a bivari-
ate normal frequency distribution. The cloud cover category with the highest
probability is selected. Diekmann and Smith (1989) found that the cloud-cover
category is correctly identified about two thirds of the time, and incorrectly
identified by one cloud category about one third of the time.
During processing, it was discovered that at large satellite viewing zenith angles
the retrieved albedos were too high and the longwave radiant exitances were too
low. To alleviate this problem, observations from satellite viewing zenith angles
greater than 75° (70° for ERBE) were rejected.
At this point in the processing, the instantaneous values of OLR and albedo
have been calculated. The next step produces daily averaged values. The daily
averaged OLR (M ) is simply the mean of the noon (ascending) and the midnight
LW
6
The ERBE team decided that the s h o r t w a v e - l o n g w a v e correlations were not important and they
do not use them for ERBE processing.
Chapter 10 Earth Radiation Budget
observes a target area and that the albedo calculated using the anisotropic reflect-
ance factor is A*[0 (f )] Let A'[6 {t)] be the albedo of the scene as a function
sun obs sun
In other words, the albedo estimates of Suttles et al., coupled with a knowledge
of the sun's elevation angle, supply the time variation of albedo, and the observed
albedo supplies the magnitude. The daily mean reflected shortwave radiation can
then be estimated as
_ 1 /-sunset
M s w =r /* E A[0 (i)]
sun sun sun (10.21)
^sunrise
Ä = Sss, (10.23)
-^sun
and the daily net radiation is
£ n e t =(l-Ä)E s u n -M L W . (10.24)
Note that the solar constant was assumed to be 1371.5 W m~ in all of the 2
Science Data Center for the days that the instrument was turned on between May
1979 and May 1980.
The second way in which the NFOV data were processed was with the sorting
into angular bins (SAB) algorithm (Kyle et al., 1993). Developed by Arking and
Vemury (1986) as a means of checking the predecessor to the MLCE method,
the SAB method does not use angular dependence models. Instead, observations
of an area are accumulated until the area has been observed from sufficiently
many angles that the radiant exitance may be evaluated by directly integrating
Eq. 10.12. Clearly the SAB method cannot produce high temporal resolution.
The NSSDC archive contains zonal mean quantities calculated daily, but the
regional quantities are only available as monthly averages. Also, the retrievals
were not in the (166 km) subtarget areas but in (500-km) areas called target
2 2
areas. (Each target area is composed of nine subtarget areas; the globe is covered
by 2070 target areas.) The SAB method is still primarily useful for checking the
MLCE algorithm, and the two methods compare well, according to Kyle et
al. (1993).
10.2 Top of the Atmosphere Radiation Budget 377
Ψ is
_ 1
L (Θ,Ψ; Θ ,Ψ ) — y M<sun^sun — ~ ~ / ^ s u n ^ s u n >
r 8ιιη (10.25)81ιη r
7Γ
shortwave irradiance that a satellite above this Lambertian Earth would measure
is then
J
-ITT r-πΙΙ
L ; (Θ,Ψ;Θ ,Ψ 8ιιη 51ιη ) cos0 sin0 άθάφ (10.26)
0 J
o
or
E * (r ,0 ,Y )=-E
s w s s s s u n μ 5ϋη cosö sin0 άθ αφ, (10.27)
TT Jo Jo
where ( Γ , Θ , Ψ ) is the position of the satellite, and (0,φ) are directions from the
8 8 8
flat-plate sensor. In this calculation, the point (Θ,Ψ) is the location where the ray
from the satellite intersects the top of the atmosphere; i.e., both Θ and Ψ are
functions of r , Θ , Ψ , 0, and φ. (Note that cos0 is the response of the flat-plate
8 8 8
A(r , s Θ , Ψ) =
8 8
£
sw(r ,®s>^s).
s ( 1 0 > 2 8 )
E£ (r ,e ,¥ )
w s s s
This estimate assumes that the albedo is constant throughout the field of view of
the flat-plate sensor and that the reflection is isotropic. The albedo for the near-
7
We will call this the maximum irradiance even though greater irradiances could be measured if
reflectance is not isotropic.
8
Radiance is usually specified as a function of zenith and azimuth angles. Here, to emphasize the
Earth-sun-satellite geometry, w e specify radiance as a function of the latitude and longitude of the
viewed location and of the sun.
Chapter 10 Earth Radiation Budget
378
mean albedo for the target area is the sum of the measured irradiances divided
by the sum of the maximum irradiances.
In the first algorithm, the longwave irradiance is calculated by subtracting the
shortwave irradiance from the total irradiance ( £ = £ ο τ "~ ^sw)- Then the L W Τ
irradiance is converted into a radiant exitance at the top of the atmosphere. The
historical way to do this (Smith et al., 1977; Jacobowitz et al., 1979) is to assume
that the radiation from Earth is isotropic, which means that the radiation can be
thought of as emanating from a point source at the center of the Earth and
decreasing in intensity as the square of the distance from the center of the Earth.
If the measured longwave irradiance is £ , then the TO A radiant exitance must be
L W
(10.29)
where r is the radius of the Earth, h is the height of the satellite, and h
e s is TOA
the height of the assumed top of the atmosphere (15 km for Nimbus 7 ERB
retrievals). This is called a shape factor retrieval (Green, 1981) or inverse square
law retrieval. The shape factor is
and is constant for a circular orbit. M is assigned to the target area which
L W
contains the subsatellite point, and all observations within a target area are av-
eraged.
The net radiation is calculated for each target area as
£ n e t = (1 - kA)E an - M, (10.30)
where M is the average of the noon and midnight OLR observations, £ (Eq.
L W s u n
10.22) is the mean daily solar insolation on the target area, and k is a factor to
transform A into a mean daily value. The factor k is based on a hybrid directional
reflectance model (90% cloud/ocean, 10% land) derived from Raschke et al.
(1973); the model is applied globally.
Nine years (November, 1978-October, 1987) of albedo, OLR, and net radia-
tion computed from Nimbus 7 WFOV observations are archived at the National
Space Science Data Center for daily and monthly time periods. Figure 10.2 shows
the zonally averaged albedo, OLR, and net radiation for the period November
1978 to October 1986. The Earth's radiation budget is quite consistent from year
to year, but important interannual variations occur, for example, the 1982-1983
El Nino event.
The WFOV instrument is still operational, and radiation budget products will
be produced past October, 1987, as funds become available.
From an altitude of 955 km, a Nimbus 7 WFOV sensor simultaneously views
over 130 (500 km) target areas. Each observation, however, is assigned to the
2
10.2 Top of the Atmosphere Radiation Budget 379
- 9 0 LI I I I I I I I
1979 1980 1981 1982 1983 1984 1985 1986
single target area which is directly below the satellite. The resulting radiation
budget fields therefore appear to be very smooth in comparison with NFOV fields.
A large effort has gone into desmoothing or deconvoluting the Nimbus 7 ERB
data. This is the second algorithm with which the WFOV data have been processed.
The longwave data can be desmoothed in a straightforward manner. Green
(1981) explains two ways to desmooth the longwave data. One way involves
a spherical harmonic expansion of the longwave fields. The satellite-measured
Chapter 10 Earth Radiation Budget
380
irradiance is given by
£ L W ( © S , ^ S ) = - Γ Γ *Μ^(®,Ψ)ξ(®,%ζ)€θ8θ3ίηθάθάφ,
1
(10.31)
TT Jo Jo
M (e,Y) LW = SΣ
n=0 m= —η
b?Y%&,V). (10.32b)
The a™ are calculated from observations obtained over a sufficiently long period
of time that all the "boxes" (perhaps target areas) are filled. This takes a minimum
of several days. The fcJJ are the spherical harmonic expansion coefficients of the
1
sought solution for M . Smith and Green (1981) showed that if the anisotropic
L W
where the λ are easily calculated eigenvalues of Eq. 10.33. Therefore, the a™ are
η
Κ = Κ αϊ. 1
(10.34)
This technique is called deconvolution? Since λ = [(r + h )/(r + /? )] , 0 e TOA e s
2
deconvolution starts at the same point as the shape factor retrieval. The λ decrease η
tropic reflectance factor ξ^(ζ). Suppose that there are Ν such regions. Each satellite
irradiance observation is given by
£i = Σ ß,kM , k (10.35)
k=l
where
9
Care must be taken to choose the optimal number of spherical harmonics. T o o small a number
will not desmooth the data as much as they could be. T o o large a number will introduce artifacts
into the data.
10.2 Top of the Atmosphere Radiation Budget 381
and i l is the solid angle of area k subtended at the ith position of the satellite.
ik
A large number of observations (>N) can be used to find the least squares best
solution for the M .
10
k
The shortwave data are more difficult to desmooth than the longwave data
because the shortwave radiant exitance depends strongly on the position of the
sun and because the shortwave anisotropic factor is more complicated than the
longwave anisotropic factor. In particular, the deconvolution technique does not
work because spherical harmonics are not eigenfunctions of the measurement
equation. Parameter estimation can work, but retrieval of M is not really s w
£ s w ( ® , ^ ) = - Γ Γ *Μ^(Θ,Ψ)ξ(ζ)οο3θύηθάθάφ,
1
(10.37)
π Jo Jo
s s
spherical harmonic series. For each shortwave WFOV observation, Eq. 10.37 is
a linear equation in the unknown spherical harmonic expansion coefficients.
Proceeding as in parameter estimation, a large number of £ measurements are s w
used to calculate the spherical harmonic coefficients. Hucek et al. use 289 coeffi-
cients (N = 16), which they estimate to be the number of independent pieces of
information that are retrievable from the WFOV data. The albedo of each target
area is then calculated using the coefficients.
Figure 10.3 compares the smooth and desmoothed WFOV zonal averages.
Desmoothing results in a field in which the highs are higher and the lows are
lower; thus they are more like scanner observations. The field mean does not
change in the desmoothing process. Desmoothing makes the WFOV data more
useful for regional radiation budget studies. If one only wants to study the global
mean radiation budget, desmoothing is not necessary.
One year (July 1983 to June 1984) of Nimbus 7 ERB WFOV data have been
desmoothed and are archived at the National Space Science Data Center (Kyle
et al., 1993). The shortwave observations were desmoothed by the method of
Hucek et al. (1987), and the longwave data were processed by the deconvolution
method. Ardanuy et al. (1987) detail the specific process for retrieval of the
longwave radiant exitance and the net radiation.
The best estimates of the accuracy of the Nimbus 7 ERB parameters are
determined by comparing ERB data with ERBE data. We therefore defer until
the end of the next section a discussion of ERB accuracy.
Again, care must be taken when choosing N . T o o large a number creates a system of equations
1 0
which is difficult to solve; t o o small a number results in less than optimal desmoothing. Also, there
can be stability problems in finding the solution (see Twomey, 1977).
332 Chapter 10 Earth Radiation Budget
160 I 1 1 1 I I I I I . I . I
-60 -50 -40 -30 -20 -10 0 10 20 30 40 50 60
LATITUDE (DEGREES)
FIGURE 10.3. Comparison of desmoothed Nimbus 7 W F O V data with standard (smooth) data: (a)
albedo (percent) and (b) longwave radiant exitance (W m~ ). [After Ardanuy et al. (1987).]
2
Second, while the scanner was operational, NFOV scene types were used to
improve retrieval of WFOV albedo. The ERB algorithm assumed that reflected
shortwave radiation was isotropic (Eq. 10.27). When both the ERBE scanner and
the ERBE non-scanner are operating, the scanner sweeps out the area viewed by
the non-scanner. The scene types identified with the scanner can be used to include
an anisotropic factor into Eq. 10.27:
(10.38)
The albedo is then calculated by dividing the observed irradiance by Εξ^. In this
calculation, the albedo is still assumed constant over the field of view, but the
1 1
The vegetated desert was included in the land scene, and the non-vegetated desert became the
desert model (R. N . Green, personal communication, 1990).
384 Chapter 10 Earth Radiation Budget
M
.= Σ (10.39)
j= -N
To conserve energy, the weights must sum to the reciprocal of the shape factor.
The advantage of this technique over deconvolution is that it does not smear
observations across local times because observations separated by several minutes
are used instead of observations over several days. The disadvantage is that the
numerical filter technique only enhances the resolution along the ground track
and does not improve the across-track resolution. Deconvolution enhances the
radiation field in both directions.
Fourth, better spatial and temporal averaging procedures are used (Brooks et
al., 1986). A table of observations during a month is kept for each location. The
table is formed of day and local time bins. The observations from multiple satellites
are included in the table, and averages are carefully performed to determine
monthly mean quantities. This averaging includes, for example, fitting a half sine
curve to the daytime longwave measurements over land, which Brooks and Minnis
(1984) showed is an improvement over simple averaging. The scanner data are
mapped onto a 2.5° x 2.5° latitude-longitude grid system, which is finer resolution
than the (500 km) grid used for the Nimbus 7 ERB data. The enhanced resolution
2
(numerically filtered) WFOV and MFOV data are mapped onto a 5° x 5° grid,
and the shape factor WFOV and MFOV data are mapped onto a 10° x 10° grid.
The reader is referred to Brooks etal. (1986) for further details on the complicated
task of producing accurate monthly mean radiation budget quantities from instan-
taneous satellite-measured irradiances.
An important aspect of the ERBE data processing is that clear scenes are
evaluated by themselves in addition to their use in the averages; ERBE produces
an estimate of average clear-sky radiation for each month. The difference between
clear-sky and mean radiation is caused mainly by clouds. This difference can be
used to investigate cloud forcing, the effect of clouds on the radiation budget of
Earth. Ramanathan et al. (1989) showed using ERBE data that in April, 1985,
averaged over the globe, clouds caused an extra 44.5 W m~ in shortwave energy
2
1985, clouds lowered the globally averaged net radiation by 13.2 W m~ . The 2
effects of clouds on radiation budget and climate remain an active research topic
10.2 Top of the Atmosphere Radiation Budget 385
compared to the ERB value of 1371.5 W m" . The ERBE algorithms locate the
2
Clear Albedo
FIGURE 10.4. Combined ERBS and N O A A 9 scanner data for June 1985: (a) mean albedo (percent),
(b) clear albedo (percent), (c) mean longwave radiant exitance (W m" ), (d) clear longwave radiant
2
were constructed from combined ERBS and NOAA 9 scanner data. The zonal
averages for the same time period are shown in Fig. 10.5.
There are basically three ways to estimate the accuracy of satellite-based radia-
tion budget measurements: (1) compare the measurements made by different
satellites, (2) compare measurements made by the same satellite, and (3) estimate
the accuracy by estimating the uncertainty in each aspect of the measurement
and/or retrieval and compounding the uncertainties. All three types of analysis
have been performed.
Kyle et al. (1989b) have compared Nimbus 7 ERB WFOV measurements with
ERBE measurements. They found that "the Nimbus 7 WFOV shortwave sensor
reads about 2.5% higher than the ERBS sensor, while the Nimbus 7 total channel
reads about 1% below the ERBS channel near midnight and 1% above it at
noon." The global average of albedo agreed with combined ERBS/NOAA 9
NFOV albedo within 0.03%, and the OLR agreed within 0.16%. They conclude
that the agreement between the ERB and ERBE data sets is good enough to allow
the sets to be combined for long-term studies of the radiation budget. They warn,
however, that the characteristics of each instrument must be kept in mind.
a
90 Ν ι 1 1 1 1 1 ~p 1 1 1
— — — 1
Clear Albedo
( Mean Albedo
90S I ι ι ι ι 1 ι ι ι 1
^ 0 100
b
90N| 1 1 1 1 1 1 1 1 1
C l e a r Long W a v e
N
« M e a n Long W a v e
90S* • » · · 1 1 1 1 1
^^^^ZT -
90 Ν ι 1 . 1 1 ^ 1 1 1 1
EQ - ^ ^ ^ ^
^ ^ • ^ ^
90S' 1
1 » » 1 1 ' 1 ι
-200 0 200
FIGURE 10.5. E R B S / N O A A 9 Zonal averages for June 1 9 8 5 : (a) albedo (percent), (b) longwave radiant
exitance (W m" ), (c) net radiation (W m ) . [Courtesy of G. Garrett Campbell, Cooperative Institute
2 - 2
Green et al. (1990b) have compared the ERBE sensors with themselves. The
scanner has longwave, shortwave, and total channels. The total channel was
assumed to be correct. At night, the longwave channel should give the same
reading as the total channel. They found that this was true within about 1% for
both ERBS and NOAA 9. During the day, the shortwave channel should be the
difference between the total channel and the longwave channel. They found that
for ERBS the agreement was within 0.5%, but the NOAA 9 shortwave channel
appears to be about 2% too high. Using the scanner measurements, the non-
scanner measurements can be simulated. Green et al. found that "Generally, the
MFOV measurements are within 2% to 3 % of the scanner measurements. The
WFOV TOT channel is very good, and the WFOV SW channel is about 3 % high
for ERBS and excellent for NOAA 9."
Barkstrom et al. (1989) list ten validation criteria and many researchers who
are pursuing them using various data for comparison. They note that "ERBE
data products are complex assemblages of data and models. Thus, their uncertaint-
ies are difficult to compute." They expect uncertainties in instantaneous scanner
radiances to be about 1% for longwave observations and 2 - 3 % for shortwave
observations. For instantaneous radiant exitances from 2.5° x 2.5° boxes they
expect ±5 W m~ for longwave observations and ±15 W m~ for shortwave
2 2
ments. Finally, they estimate that the uncertainty in ERBE measurements of the
annual, global average of the net radiation is about ± 5 W m~ . 2
concentration. There are trends in the radiation budget data, but the uncertainties
prevent them from being statistically significant. We hope that this problem will
be resolved with next-generation radiation budget instruments.
cc
12
LOCAL TIME (hr)
FIGURE 10.6. Diurnal variation of albedo (top, percent) and outgoing longwave radiation (bottom,
W m~ ) for an area over South America centered at 25.9° South and 68.7° West during November
2
the surface radiation budget is dominated by clouds, however, satellite data are
well suited for this purpose, and they can provide world-wide estimates.
The surface radiation budget may be divided into five quantities: downward
shortwave radiation (called insolation), upward shortwave (reflected) radiation,
downward longwave radiation, upward longwave radiation, and net radiation.
irradiance at the surface, and A is the surface albedo. This equation implicitly
sfc
includes scattering and multiple reflections between the surface and clouds.
Fritz et al. correlated A, measured by TIROS 3, with atmospheric transmittance
( a t m — ^ s f c ^ M s u n ^ s u n ) ? measured by ground-based instruments. A correlation coef-
T
ficient of —0.9 was found, which confirms the hypothesis that when clouds (with
their high albedo) cover an area, the fraction of solar radiation reaching the
surface decreases. Fritz et al. went on to explore the fraction of incident radiation
which is absorbed in the atmosphere. Hanson et al. used more accurate TIROS
10.3 Surface Radiation Budget 391
4 data from a longer period to show that the solar energy absorbed in the
atmosphere over the U.S. during the period March through May 1962 ranged
from 9% to 26%, and that absorbed at the ground ranged from 34% to 5 8 % .
Non-satellite estimates of the surface albedo were necessary for these studies.
Vonder Haar and Ellis (1975) estimated surface albedo from satellite data and
surface pyranometer measurements. All of these studies used data from low Earth
orbiters, which means that only one observation was made per day during daylight.
Tarpley (1979) developed and Justus et al. (1986) improved a statistical tech-
nique to use GOES data to estimate £ over the U.S. Solving Eq. 10.40 for £
sfc sfc
we have
1 ~ ^sfc
£ s f c = ^ u n £ s u n - ^ - £ a t m t (10.43)
The two most important terms in the Eq. 10.43 are the solar insolation ( / L i E ) sun sun
and the reflected radiance (ttL). It follows that a simple regression equation using
these quantities as predictors ought to estimate the surface insolation well. To
account for geographic variations in surface albedo and in atmospheric absorption,
Tarpley and Justus et al. used deviations from the expected minimum radiance
rather than radiance itself. Their final regression equation (Justus et al., 1986) is
£sfc = M s A u n ( a
U 1 + ^sun + ^sun) + ^(C " 2
C^J, (10.44)
where C and C min are the mean observed and minimum (respectively) GOES
visible brightness count (approximately proportional to L ) . The mean is calcu- _1/2
lated over a 30-50-km square, which means that the results have this resolution.
The most difficult part of this analysis is in keeping an up-to-date map of minimum
counts for each satellite viewing time. Justus et al. used a 1° latitude-longitude
grid for this purpose.
Using this method, Justus et al. were able to reproduce hourly surface pyrano-
meter measurements within about 16% rms of the mean hourly insolation and
daily surface pyranometer measurements within about 9.5% rms of the daily
insolation or about 1.4 MJ m T . Figure 10.7 shows the mean daily surface insol-
2
ation over the United States for July and December 1983.
Gautier et al. (1980) and Diak and Gautier (1983) developed a physical retrieval
scheme based on a simple radiative transfer model. Their model atmosphere
consists of an ozone layer; an atmospheric layer, which absorbs and scatters solar
radiation; a cloud layer; a subcloud atmospheric layer; and the surface. The
absorption and scattering properties of the atmosphere are taken from the litera-
JULY
UNITEO STATES
D E C E M B E R
UNITEO STATES
FIGURE 10.7. Solar insolation (W m~ ) at the surface for (a) July and (b) December 1 9 8 3 . [Reprinted
2
by permission of the publisher from "Satellite Measured Insolation in the United States, Mexico, and
South America" by C. G. Justus, Μ. V. Paris, and J . D. Tarpley, Remote Sensing of Environment,
Vol. 2 0 , pp. 5 7 - 8 3 . Copyright 1 9 8 6 by Elsevier Science, Inc.]
10.3 Surface Radiation Budget
393
ture, and the surface dewpoint (from an analysis) provides an estimate of precipita-
ble water, which is the chief atmospheric absorber of sunlight. A satellite measure-
ment of reflected radiation yields an albedo. If the albedo is lower than a threshold,
clear sky is assumed, and the solar insolation at the surface is calculated with a
clear-sky model. If the albedo is higher than the threshold, an overcast sky is
assumed, and surface solar insolation is calculated from the cloudy model. The
technique is simple and computationally easy. Its advantage is that it can be used
at every satellite pixel, which provides an estimate of horizontal variability of
solar radiation.
In comparison with pyranometers, the method of Diak and Gautier had a
standard error of 5.3% of the mean daily insolation, which is roughly the accuracy
cited for a well-maintained pyranometer.
Dedieu et al. (1987) developed a method similar to that of Diak and Gautier
for use with Meteosat data.
Moser and Raschke (1984) also used Meteosat data, but they employed a
sophisticated radiative transfer model to convert radiances to surface insolation.
Basically, Moser and Raschke assume that clouds perturb the clear-sky insolation
Eff . They find the minimum and maximum observed radiance over an area during
c
a month's time for each observing time. The observed radiances L are then obs
normalized by
(10.45)
E c(£su„) = [ W ( ^ £ s u „ ) £ £ ( 0 ] ,
si
(10.46)
where /*(L', £ ) is a tabulated function determined from the radiative transfer
sun
model. Moser and Raschke report daily rms errors in Europe of 1 0 - 1 4 % (1.8-2.2
MJ m~ ) and mean monthly rms errors of 3 - 7 % (0.5-1.3 MJ m " day—2).
2 2
The reflected solar radiation is the product of the surface albedo and the
downward solar radiation. Thus a knowledge of the surface albedo is the key
element in estimating reflected solar radiation.
The basic approach to estimating surface albedo has historically been through
the minimum albedo technique (e.g., Raschke and Preuss, 1975). The region of
interest is monitored over a period of time, say one month, and the minimum
albedo during that time is noted. Since few locations are likely to be cloud-covered
for an entire month, the minimum albedo is likely to represent the clear-sky
planetary albedo. Corrections for atmospheric absorption and scattering may then
be applied to retrieve the albedo of the surface itself (e.g., Preuss and Geleyn, 1980).
Surface albedo is also produced as a byproduct of solar insolation calculations
(Gautier et. al., 1980). Dedieu et al. (1987) have mapped surface albedos over
Europe using Meteosat data.
394 Chapter 10 Earth Radiation Budget
Pinker and Ewing (1986) note that the surface type is important when using
narrowband measurements to infer surface albedo. The albedo of snow, for
example is high for wavelengths less than 1 μπι, but drops rapidly for longer
wavelengths. The albedo of vegetated surfaces, on the other hand, is low in the
portion of the spectrum sampled by visible detectors (such as AVHRR channel
1), but rises in the near infrared. Simply assuming that the broadband surface
albedo is the same as the narrowband albedo can cause errors if surface type
is ignored.
Because Eq. 10.40 is linear in A and A , it is reasonable to assume that an
sfc
A sfc = a + bA (10.47)
longwave radiation. The largest problem occurred when one site was foggy. The
TOVS cloud amount was only 8%, which caused the satellite estimate to be too
low by about 80 W m~ . 2
10.3 Surface Radiation Budget y)$
Darnell et al. (1986) improved their data handling and corrected the pyrgeome-
ter data for shortwave heating of the dome. For monthly averaged data, they
found a correlation coefficient of 0.98 and a standard error of 10 W m~ , or 2
treatment produced the best results, which indicates the importance of clouds in
downward longwave irradiance estimation.
Pinker and Tarpley (1988) found that GOES brightness count alone has nearly
as high a correlation with surface net radiation as does TOA net radiation, which
supports the argument that clouds are the chief moderator of surface net radiation.
Bibliography
Abel, P., and A. Gruber (1979). An Improved Model for the Calculation of Longwave Flux at 11
jitm. N O A A Tech. M e m o . NESS 106, Washington, D.C.
Ardanuy, P. E., and H. Jacobowitz (1984). A calibration adjustment technique combining ERB
parameters from different remote sensing platforms into a long-term data set. / . Geophys. Res.,
89, 5 0 1 1 - 5 0 1 9 .
Ardanuy, P. E., H. L. Kyle, R. R. Hucek, and B. S. Groveman (1987). Nimbus 7 Earth Radiation
Budget wide field of view climate data set improvement. 2. Deconvolution of Earth radiation
budget products and consideration of 1 9 8 2 - 1 9 8 3 El N i n o event./. Geophys. Res., 9 2 , 4 1 2 5 - 4 1 4 3 .
Ardanuy, P., and J. Rea (1984). Degradation asymmetries and recovery of the NIMBUS 7 Earth
Radiation Budget shortwave radiometer. / . Geophys. Res., 8 9 , 5 0 3 9 - 5 0 4 8 .
Arking, Α., and J. S. Levine (1967). Earth albedo measurements: July 1963 to June 1 9 6 4 . / . Atmos.
Sei., 2 4 , 7 2 1 - 7 2 4 .
Arking, Α., and S. Vemury (1984). The Nimbus 7 ERB data set: A critical analysis. / . Geophys. Res.,
89, 5 0 8 9 - 5 0 9 7 .
Atwater, Μ. Α., and J. T. Ball (1981). A surface solar radiation model for cloudy atmospheres. Mon.
Wea. Rev., 1 0 9 , 8 7 8 - 8 8 8 .
Bandeen, W. R., R. A. Hanel, J. Lickt, R. A. Stampl, and W. G. Stroud (1961). Infrared and solar
radiation measurements from the TIROS II meteorological satellite. / . Geophys. Res., 66,
3169-3185.
Barkstrom, Β. R. (1984). The Earth Radiation Budget Experiment (ERBE). Bull. Amer. Meteor. Soc,
65, 1 1 7 0 - 1 1 8 5 .
Barkstrom, Β. R., and G. L. Smith (1986). The Earth Radiation Budget Experiment: Science and
implementation. Rev. Geophys., 2 4 , 3 7 9 - 3 9 0 .
Barkstrom, Β. R., Ε. Harrison, G. Smith, R. Green, J. Kibler, R. Cess, and the ERBE Science Team
(1989). Earth Radiation Budget Experiment (ERBE) archival and April 1985 results. Bull. Amer.
Meteor. Soc, 70, 1 2 5 4 - 1 2 6 2 .
Bess, T. D., and G. L. Smith (1987a). Atlas of Wide-Field-of-View Outgoing Longwave Radiation
Derived from Nimbus 6 Earth Radiation Budget Data Set-July 1976 to June 1978. N A S A Ref.
Pub. RP-1185, Langley Research Center, Hampton, VA.
Bess, T. D., and G. L. Smith (1987b). Atlas of Wide-Field-of-View Outgoing Longwave Radiation
Derived from Nimbus 7 Earth Radiation Budget Data Set-November 1978 to October 1985.
N A S A Ref. Pub. RP-1186, Langley Research Center, Hampton, VA.
Bess, T. D . , G. L. Smith, and T. P. Charlock (1989). A ten-year monthly data set of outgoing longwave
radiation from N i m b u s - 6 and N i m b u s - 7 satellites. Bull. Amer. Meteor. Soc, 70, 4 8 0 - 4 8 9 .
Bess, T. D., R. N . Green, and G. L. Smith (1980). Deconvolution and Analysis of Wide-Angle
Longwave Radiation Data from Nimbus 6 Earth Radiation Budget Experiment for the First Year.
N A S A Tech. Paper 1746, Langley Research Center, Hampton, VA.
Bess, T. D., R. N . Green, and G. L. Smith (1981). Deconvolution of wide field-of-view radiometer
measurements of Earth-emitted radiation. Part II: Analysis of first year of Nimbus 6 ERB data.
/. Atmos. Sei., 3 8 , 4 7 4 - 4 8 8 .
Brooks, D . R., and P. Minnis (1984). Comparison of longwave diurnal models applied to simulations
of the Earth Radiation Budget Experiment. / . Climate Appl. Meteor., 2 3 , 1 5 5 - 1 6 0 .
10.3 Surface Radiation Budget 397
Gruber, Α., and Η. Jacobowitz (1985). The longwave radiation estimated from N O A A polar orbiting
satellites, an update and comparison with N i m b u s - 7 ERB results. Adv. Space Res., 5, 1 1 1 - 1 2 0 .
Gruber, Α., and Τ. S. Chen (1988). Diurnal variation of outgoing longwave radiation. / . Climatology,
8, 1 - 1 6 .
Gruber, Α., I. Ruff, and C. Earnest (1983). Determination of the Planetary Radiation Budget from
TIROS-N Satellites. N O A A Tech. Rep. NESDIS 3 , Washington, D C .
Gruber, Α., Μ. Varnadore, P. A. Arkin, and J. S. Winston (1986). Monthly and Seasonal Mean
Outgoing Longwave Radiation and Anomalies. N O A A Tech. Rep. NESDIS 2 6 , Washington, D C .
Gube, M. (1980). Outgoing longwave flux computation from Meteosat data. ESA Journal, 4 , 3 8 1 - 3 9 6 .
Gube, M. (1982). Planetary albedo estimates from Meteosat data. ESA Journal, 6, 5 3 - 6 9 .
Hall, J. B., Jr. (ed.) (1982). Earth Radiation Science Seminars. N A S A Conf. Pub. 2 2 3 9 , Langley
Research Center, Hampton, VA.
Hanson, K. J, Τ. H. Vonder Haar, and V. E. Suomi (1967). Reflection of sunlight to space and
absorption by the Earth and atmosphere over the United States during spring 1 9 6 2 . Mon. Wea.
Rev., 9 5 , 3 5 4 - 3 6 2 .
Hanson, K. J. (1971). Studies of cloud and satellite parameterization of solar irradiation at the Earth's
surface. Proceedings: Miami Workshop on Remote Sensing. U.S. Department of Commerce,
133-148,
Harrison, E. F., D. R. Brooks, P. Minnis, B. A. Wielicki, W. F. Staylor, G. G. Gibson, D . F. Young,
F. M. Denn, and the ERBE Science Team (1988). First estimates of the diurnal variation of
longwave radiation from the multiple-satellite Earth Radiation Budget Experiment. Bull. Amer.
Meteor. Soc, 6 9 , 1 1 4 4 - 1 1 5 1 .
Hartmann, D., V. Ramanathan, A. Berrior, and G. E. Hunt (1986). Earth radiation budget data and
climate research. Rev. Geophys., 2 4 , 4 3 9 - 4 6 8 .
Hickey, J. R., Β. M. Alton, F. J. Griffin, H. Jacobowitz, P. Pelligrino, R. H. Maschhoff, Ε. A. Smith,
and Τ. Η. Vonder Haar (1982). Extraterrestrial solar irradiance variability: T w o and one-half
years of measurements from N i m b u s - 7 . Sol. Energy, 2 9 , 1 2 5 - 1 2 7 .
Hickey, J. R., Β. M. Alton, H. L. Kyle, and D . Hoyt (1988a). Total solar irradiance measurements
by ERB/Nimbus 7: A review of nine years. Space Sei. Rev., 4 8 , 3 2 1 - 3 4 2 .
Hickey, J. R., Β. M. Alton, H. L. Kyle, and E. R. Major (1988b). Observation of total irradiance
variability from Nimbus satellites. Adv. Space Res., 8(7), 5 - 1 0 .
Hickey, J. R., L. L. Stowe, H. Jacobowitz, P. Pelligrino, R. H. Maschhoff, F. Β. House, and Τ. Η.
Vonder Haar (1980). Initial solar irradiance determinations from Nimbus 7 cavity radiometer
measurements. Science, 2 0 8 , 2 8 1 - 2 8 3 .
House, F. B., A. Gruber, G. E. Hunt, and A. T. Mecherikunnel (1986). History of satellite missions
and measurements of the Earth radiation budget ( 1 9 5 7 - 1 9 8 4 ) . Rev. Geophys., 2 4 , 3 5 7 - 3 7 7 .
House, F. B., and J. C. Jafolla (1980). One dimensional technique for enhancing Earth radiation
budget observations from Nimbus 7 satellite. 1980 International Radiation Symposium, IAMAP
Radiation Commission, Fort Collins, C O , 292—294.
Hoyt, D . V., H. L. Kyle, J. R. Hickey, and R. H. Maschhoff (1992). The N i m b u s - 7 total solar
irradiance: A new algorithm for its derivation. / . Geophys. Res., 97, 5 1 - 6 3 .
Hucek, R. R., H. L. Kyle, and P. E. Ardanuy (1987). Nimbus 7 Earth Radiation Budget wide field
of view climate data set improvement. 1. The Earth albedo from deconvolution of shortwave
measurements. / . Geophys. Res., 9 2 , 4 1 0 7 - 4 1 2 3 .
Hunt, G. E., R. Kandel, and A. T. Mecherikunnel (1986). A history of presatellite investigations of
the Earth's radiation budget. Rev. Geophys., 2 4 , 3 5 1 - 3 5 6 .
Hwang, P. H., L. L. Stowe, Η. Y. M. Yeh, H. L. Kyle, and the Nimbus 7 Cloud Data Processing
Team (1988). The N i m b u s - 7 global cloud climatology. Bull. Amer. Meteor. Soc, 6 9 , 7 4 3 - 7 5 2 .
Jacobowitz, Η., Η. V. Soule, H. L. Kyle, F. B. House, and the Nimbus 7 ERB Experiment Team (1984a).
The Earth Radiation Budget (ERB) experiment: An overview. / . Geophys. Res., 8 9 , 5 0 2 1 - 5 0 3 8 .
Jacobowitz, H., L. L. Stowe, and J. R. Hickey (1978). The Earth Radiation Budget (ERB) experiment.
In The Nimbus 7 Users' Guide, N A S A Goddard Space Flight Center, Greenbelt, M D , 3 3 - 6 9 .
Jacobowitz, H., R. J. Tighe, and the Nimbus 7 ERB Experiment Team (1984b). The Earth radiation
budget derived from the N i m b u s - 7 ERB experiment. / . Geophys. Res., 8 9 , 4 9 9 7 - 5 0 1 0 .
10.3 Surface Radiation Budget 399
Jacobowitz, H., W. L. Smith, Η. B. Howell, F. W. Nagle, and J. R. Hickey (1979). The first 18
months of planetary radiation budget measurements from the Nimbus 6 ERB experiment./. Atmos.
Sei., 3 6 , 5 0 1 - 5 0 7 .
Janowiak, J. E., A. F. Krueger, P. A. Arkin, and A. Gruber (1985). Atlas of Outgoing Longwave
Radiation Derived from NOAA Satellite Data. N O A A Atlas N o . 6, Washington, D C .
Justus, C. G., Μ. V. Paris, and J. D . Tarpley (1986). Satellite measured insolation in the United States,
Mexico, and South America. Rem. Sens. Env., 2 0 , 5 7 - 8 3 .
Kidwell, Κ. B. (1991). NOAA Polar Orbiter Data User's Guide. N O A A / N E S D I S Satellite Data
Services Division, Washington, D . C .
King, M. D., and R. J. Curran (1980). The effect of a nonuniform planetary albedo on the interpretation
of Earth radiation budget observations. / . Atmos. Sei., 37, 1 2 6 2 - 1 2 7 8 .
Kopia, L. P., Earth Radiation Budget Experiment scanner instrument. Rev. Geophys., 2 4 , 4 0 0 - 4 0 6 .
Koepke, P. (1980). Calibration of the Meteosat IR channel by ground measurements. Contrib. Atmos.
Phys., 5 3 , 4 4 2 - 4 4 5 .
Koepke, P., and Κ. T. Kriebel (1987). Improvements in the shortwave cloudfree radiation budget
accuracy. Part I: Numerical study including surface anisotropy. / . Climate Appl. Meteor., 26,
374-395.
Koepke, P. (1989). Removal of atmospheric effects from A V H R R albedos. / . Appl. Meteor., 2 8 ,
1341-1348.
Kriebel, Κ. T. (1978). Measured spectral bidirectional reflection properties of vegetated surfaces.
Appl. Opt., 17, 2 5 3 - 2 5 9 .
Kriebel, Κ. T. (1981). Calibration of the Meteosat VIS channel by airborne measurements. Appl.
Opt., 2 0 , 1 1 - 1 2 .
Kung, E. C , R. A. Bryson, And D . H. Lenschow (1964). Study of a continental surface albedo on
the basis of flight measurements and structure of the earth's surface cover over North America.
Mon. Wea. Rev., 9 2 , 5 4 3 - 5 6 4 .
Kyle, H. L., A. Mecherikunnel, P. Ardanuy, L. Penn, B. Groveman, G. G. Campbell, and Τ. H. Vonder
Haar (1990). A comparison of t w o major earth radiation budget data sets. / . Geophys. Res.,
95, 9 9 5 1 - 9 9 7 0 .
Kyle, H. L., F. B. House, P. E. Ardanuy, H. Jacobowitz, R. H. Maschhoff, and J. R. Hickey (1984).
N e w in-flight calibration adjustment of the Nimbus 6 and 7 Earth Radiation Budget wide field
of view radiometers. / . Geophys. Res., 8 9 , 5 0 5 7 - 5 0 7 6 .
Kyle, H. L., J. R. Hickey, P. E. Ardanuy, H. Jacobowitz, A. Arking, G. G. Campbell, F. B. House,
R. Maschhoff, G. L. Smith, L. L. Stowe, and T. Vonder Haar (1993). The Nimbus Earth Radiation
Budget (ERB) experiment: 1975 to 1 9 9 2 . Bull. Amer. Meteor. Soc., 74, 8 1 5 - 8 3 0 .
Kyle, H. L., P. E. Ardanuy, and E. J. Hurley (1985). The status of the N i m b u s - 7 Earth radiation
budget data set. Bull. Amer. Meteor. Soc, 66, 1 3 7 8 - 1 3 8 8 .
Lauritson, L. G., Nelson, G. J., and F. W. Porto (1979). Data Extraction and Calibration ofTIROS-
N/NOAA Radiometers. N O A A Tech. M e m o . NESS 107, Washington, D.C.
London, J., and T. Sasamori (1971). Radiative energy budget of the atmosphere. Space Res., 1 1 ,
639-649.
Luther, M. R., J. E. Cooper, and G. R. Taylor (1984). Earth Radiation Budget Experiment nonscanner
instrument. Rev. Geophys., 2 4 , 3 9 1 - 3 9 9 .
Maschhoff, R., A. Jalink, J. Hickey, and J. Swedberg (1984). NIMBUS-Earth radiation budget sensor
characterization for improved data reduction fidelity. / . Geophys. Res., 8 9 , 5 0 4 9 - 5 0 5 6 .
Minnis, P., and E. F. Harrison (1984a). Diurnal variability of regional cloud and clear-sky radiative
parameters derived from GOES data. Part I: Analysis method. / . Climate Appl. Meteor., 2 3 ,
993-1011.
Minnis, P., and E. F. Harrison (1984b). Diurnal variability of regional cloud and clear-sky radiative
parameters derived from GOES data. Part II: November 1978 cloud distributions./. Climate Appl.
Meteor., 2 3 , 1 0 1 2 - 1 0 3 1 .
Minnis, P., and E. F. Harrison (1984c). Diurnal variability of regional cloud and clear-sky radiative
parameters derived from GOES data. Part III: November 1978 radiative parameters. / . Climate
Appl. Meteor., 2 3 , 1 0 3 2 - 1 0 5 1 .
400 Chapter 10 Earth Radiation Budget
Morcrette, J., and P. Deschamps (1986). D o w n w a r d longwave radiation at the surface in clear-sky
atmospheres: Comparisons of measured, satellite-derived, and calculated fluxes. ISLSCP: Proceed-
ings of an international conference held in Rome, Italy, 1 9 8 5 ; ESA SP-248, ESA, Paris, France,
257-262.
Morgan, J. (1980). Meteosat 1 calibration report. Issue 8, ESOC M D M D / M E T .
Moser, W., and E. Raschke (1984). Incident solar radiation over Europe from METEOSAT data. / .
Climate Appl. Meteor., 2 3 , 1 6 6 - 1 7 0 .
NASA (1978). The Nimbus 7 Users' Guide. NASA/Goddard Space Flight Center, Greenbelt, M D .
Ohring, G., A. Gruber, and R. G. Ellingson (1984). Satellite determinations of the relationship
between total longwave radiation flux and infrared w i n d o w radiance. / . Climate Appl. Meteor.,
23, 4 1 6 - 4 2 5 .
Pinker, R. T., and J. D . Tarpley (1988). The relationship between the planetary and surface net
radiation: An update. / . Appl. Meteor., 27, 9 5 7 - 9 6 4 .
Pinker, R. T., and J. A. Ewing (1985). On the modeling of surface solar radiation: Model formulation
and validation. / . Climate Appl. Meteor., 2 4 , 3 8 9 - 4 0 1 .
Pinker, R. T., and J. A. Ewing (1986). Effect of surface properties on the narrow to broadband
spectral relationship in clear sky satellite observations. Rem. Sens. Env., 2 0 , 2 6 7 - 2 8 2 .
Pinker, R. T., and L. A. Corio (1984). Surface radiation budget from satellites. Mon. Wea. Rev.,
112, 2 0 9 - 2 1 5 .
Pinker, R. T., J. A. Ewing, and J. D . Tarpley (1985). The relationship between the planetary and
surface net radiation. / . Climate Appl. Meteor., 2 4 , 1 2 6 2 - 1 2 6 8 .
Preuss, H. J., and J. F. Geleyn (1980). Surface albedos derived from satellite data and their impact
on forecast models. Arch. Meteorol. Geophys. Bioklmatol, Ser. A, 2 9 , 3 4 5 .
Ramanathan, V., R. D. Cess, E. F. Harrison, P. Minnis, B. R.. Barkstrom, Ε. Ahmad, and D . Hartmann
(1989). Cloud-radiative forcing and climate: Results from the Earth Radiation Budget Experiment.
Science, 2 4 3 , 5 7 - 6 3 .
Raschke, Ε., Τ. H. Vonder Haar, W. R. Bandeen, and M. Pasternak (1973). The annual radiation
balance of the Earth-atmosphere system during 1 9 6 9 - 1 9 7 0 from Nimbus 3 measurements. / .
Atmos. Sei., 3 0 , 3 4 1 - 3 6 4 .
Raschke, E., and H. J. Preuss (1975). The determination of the solar radiation budget of the Earth's
surface from satellite measurement. Meteorol. Rundsch., 3 2 , 1 8 - 2 8 .
Rasool, S. I., and J. H. Bolle (1984). International Satellite Land Surface Climatology Project. Bull.
Amer. Meteor. Soc, 6 5 , 1 4 1 .
Saunders, R. W., and G. E. Hunt (1983). Some radiation budget and cloud measurements derived
from Meteosat 1 data. Tellus, 35B, 1 7 7 - 1 8 8 .
Saunders, R. W., and G. E. Hunt (1980). Meteosat observations of diurnal variation of radiation
budget parameters. Nature, 2 8 3 , 6 4 5 - 6 4 7 .
Saunders, R. W., L. L. Stowe, G. E. Hunt, and C. F. England (1983). An intercomparison between
radiation budget estimates from Meteosat, Nimbus 7 and T I R O S - N satellites. / . Climate Appl.
Meteor., 22, 5 4 6 - 5 5 9 .
Smith, G. L., and R. N . Green (1981). Deconvolution of wide field-of-view radiometer measurements
of Earth-emitted radiation. Part I: Theory. / . Atmos. Set., 3 8 , 4 6 1 - 4 7 3 .
Smith, G. L., R. N . Green, E. Raschke, L. M. Avis, J. T. Suttles, B. A. Wielicki, and R. Davies (1986).
Inversion methods for satellite studies of the Earth's radiation budget: Development of algorithms
for the ERBE mission. Rev. Geophys., 2 4 , 4 0 7 - 4 2 1 .
Smith, W. L., D . T. Hilleary, H. Jacobowitz, Η. B. Howell, J. R. Hickey, and A. J. Drummond (1975).
The Earth Radiation Budget (ERB) experiment. In The Nimbus 6 User's Guide. N A S A Goddard
Space Flight Center, Greenbelt, M D , 1 0 9 - 1 3 9 .
Smith, W. L., D . T. Hilleary, H. Jacobowitz, Η. B. Howell, J. R. Hickey, and A. J. Drummond (1977).
N i m b u s - 6 Earth Radiation Budget Experiment. Appl. Opt., 16, 3 0 6 - 3 1 8 .
Stephens, G. L., Campbell, G. G., and Τ. H. Vonder Haar (1981). Earth radiation budgets./. Geophys.
Res., 86, 9 7 3 9 - 9 7 6 0 .
Stowe, L. L. (ed.) (1988). Report of the Earth Radiation Budget Requirements Review-1987. NOAA
Tech. Rep. NESDIS 4 1 , Washington, D C .
10.3 Surface Radiation Budget
401
Stowe, L. L., and M. Fromm (1983). Nimbus-7 ERB Sub-Target Radiance Tape (STRT) Data Base.
N O A A Tech. M e m o . NESDIS 3 , Washington, D C .
Stowe, L. L., C. G. Wellemeyer, T. F. Eck, Η. Υ. Μ. Yeh, and the Nimbus 7 Cloud Data Processing
Team (1988). N i m b u s - 7 global cloud climatology. Part I: Algorithms and validation./. Climate,
1, 4 4 5 - 4 7 0 .
Stowe, L. L., H. Jacobowitz, and V. R. Taylor (1980). Reflectance characteristics of Earth and cloud
surfaces as measured by the ERB scanning channels on the Nimbus 7 satellite. Proceedings:
International Radiation Symposium, Fort Collins, Colorado, 4 3 0 - 4 3 2 .
Suomi, V. E. (1958). The radiation balance of the Earth from a satellite. Ann. IGY, 1, 3 3 1 - 3 4 0 .
Suttles, J. T., R. N . Green, G. L. Smith, B. A. Wielicki, I. J. Walker, V. R. Taylor, and L. L. Stowe
(1989). Angular Radiation Models for Earth-Atmosphere System, Volume II: Longwave Radiation.
N A S A Ref. Pub. RP 1 1 8 4 vol. II, Langley Research Center, Hampton, VA.
Suttles, J. T., R. N . Green, P. Minnis, G. L. Smith, W. F. Staylor, B. A. Wielicki, I. J. Walker, D . F.
Young, V. R. Taylor, and L. L. Stowe (1988). Angular Radiation Models for Earth-Atmosphere
System, Volume I: Shortwave Radiation. N A S A Ref. Pub. RP 1 1 8 4 vol. I, Langley Research
Center, Hampton, Virginia.
Suttles, J. Τ., B. A. Wielicki, and S. Vemury (1992). Top-of-atmosphere radiative fluxes: Validation
of ERBE scanner inversion algorithm using N i m b u s - 7 ERB data. / . Appl. Meteor., 3 1 , 7 8 4 - 7 9 6 .
Tarpley, D . (1979). Estimating incident solar radiation at the surface from geostationary satellite
data. / . Climate Appl. Meteor., 18, 1 1 7 2 - 1 1 8 1 .
Tarpley, J. D . (1979). Estimating incident solar radiation at the surface from geostationary satellite
data. / . Appl. Meteor., 18, 1 1 7 2 - 1 1 8 1 .
Taylor, V. R., and L. L. Stowe (1984). Reflectance characteristics of uniform earth and cloud surfaces
derived from Nimbus 7 ERB. / . Geophys. Res., 8 9 , 4 9 8 7 - 4 9 9 6 .
Thompson, S. L., and S. G. Warren (1982). Parameterization of outgoing infrared radiation derived
from detailed radiative calculations. / . Atmos. Set., 3 9 , 2 6 6 7 - 2 6 8 0 .
Twomey, S. (1977). Introduction to the Mathematics of Inversion in Remote Sensing and Indirect
Measurements. Elsevier, N e w York, 2 4 3 pp.
Vonder Haar, Τ. Η., and V. Ε. Suomi (1971). Measurements of the Earth's radiation budget from
satellites during a five-year period. Part 1: Extended time and space means. / . Atmos. Sei., 2 8 ,
305-314.
Vonder Haar, Τ. Η., Ε. Raschke, Μ. Pasternak, and W. Bandeen (1972). The radiation budget of
the Earth-atmosphere system as measured from the Nimbus 3 satellite ( 1 9 6 9 - 1 9 7 0 ) . Space Res.,
12, 4 9 1 - 4 9 7 .
Vonder Haar, Τ. Η., and J. S. Ellis (1975). Solar energy microclimate as determined from satellite
observations. Proc. Soc. Photo-Opt. Instrum. Eng., 6 8 , 1 8 - 2 2 .
Wielicki, Β. Α., and R. N . Green (1989). Cloud identification for ERBE radiative flux retrieval. / .
Appl. Meteor., 2 8 , 1 1 3 3 - 1 1 4 6 .
Willson, R. C. (1984). Measurements of solar total irradiance and its variability. Space Sei. Rev.,
38, 2 0 3 - 2 4 2 .
Willson, R. C , and H. S. Hudson (1988). Solar luminosity variations in solar cycle 2 1 . Nature,
322, 8 1 0 - 8 1 2 .
Willson, R. C , S. Gulkis, M. Jansson, H. S. Hudson, and G. A. Chapman (1981). Observations of
solar irradiance variability. Science, 2 1 1 , 7 0 0 - 7 0 3 .
Willson, R. C. (1979). Active cavity radiometer type IV. Appl. Optics, 18(2), 1 7 9 - 1 8 8 .
Winston, J. S., A. Gruber, Τ. I. Gray, Jr., M. S. Varnadore, C. L. Earnest, and L. P. Mannello
(1979). Earth-Atmosphere Radiation Budget Analyses Derived from NOAA Satellite Data, June
1974-February 1978. Vol. 1 and 2 , N O A A , Washington, D C .
Wydick, J. E., P. A. Davis, and A. Gruber (1987). Estimation of Broadband Planetary Albedo from
Operational Narrowband Satellite Measurements. N O A A Tech. Rep. NESDIS 2 7 , Washing-
ton, D C .
77
The Future
11.1 NOAA K, L, Μ
NOAA and NASA have completed the specifications for the third generation
of TIROS Ν satellites (NASA, 1988a), which are called NOAA K, L, M. The
first of this series is scheduled for launch in 1996.
403
404 Chapter 11 The Future
NOAA Κ X X X X X X X X
NOAAL X X X X X X X
NOAA Μ X X X X X X X X
11.1.1 AVHRR/3
The AVHRR/3, which will fly on NOAA K, L, Μ (NASA, 1988b), will have
a sixth channel. Known as channel 3A, the new channel will have a center
wavelength of 1.6 μπι. This channel will be used during daylight to gain informa-
tion on the phase of clouds (ice or water) and to discriminate between low cloud
and snow cover or ice. It is an improved version of the 1.6-μπι Special Sensor C,
which flew on the DMSP F4 satellite (Bunting and d'Entremont, 1982). The old
channel 3 (3.7 μπι) will be called channel 3B. The two channel 3s will operate
at different times: 3A during daylight, 3Β at night. The instrument will be pro-
grammed to automatically switch channels at the terminator.
Several other small changes will be made in AVHRR/3 to improve performance
or calibration. Otherwise, the AVHRR/3 will be virtually identical to the
AVHRR/2. AVHRR/3 is being build by ITT Aerospace/Optical Division.
11.1.2 HIRS/3
The HIRS/3 will also be very similar to its predecessor, the HIRS/2 (NASA,
1988c). The only major difference is that five of the channels will have different
wavelengths (Table 11.2). All of the changes are made to better sense tropospheric
Central Wavelength
Half-Power
Old New Bandwidth Specified N E A L
Channel (μτη) [μηί) (cm- )
1
(mW m~ sr~ cm)
2 l
a
Differences from HIRS/2; compare with Table 4.5.
11.1 NOAA K L, Μ
} 4Q5
temperature and water vapor. HIRS/3 is also being build by ITT Aerospace/
Optical Division.
11.1.3 AMSU
The major change on NOAA K, L, Μ is that the Microwave Sounding Unit
will be significantly improved and renamed the Advanced Microwave Sounding
Unit (AMSU). In fact it will become two instruments: AMSU-A (being built by
1
Aerojet) and AMSU-B (being supplied by the British Met. Office). Table 11.3
compares current and future microwave instruments.
AMSU-A will have twice the horizontal resolution of the MSU, and AMSU-B
will have six times the resolution. This resolution increase will substantially im-
prove the scale on which analysis of the data can be performed.
A second major improvement in AMSU is that it will have a total of 20 channels
in comparison with only four on the MSU. Table 11.4 lists the frequencies sensed
by these instruments. Further, the new channels will allow the retrieval of water
vapor profiles, integrated liquid water, precipitation, and such surface parameters
as soil moisture and wind speed (over the ocean), in addition to atmospheric tem-
perature.
AMSU-A will replace the infrared Stratospheric Sounding Unit (SSU) on the
current NOAA satellites. Channels 10-14 are stratospheric-sensing channels.
Finally, the retrieval process will also change significantly for the NOAA K,
L, Μ sounding system. On the current NOAA satellites, the HIRS/2 is the primary
sounding instrument. The MSU is an important part of the retrieval process, but
its chief use is in overcast situations when the infrared sounders cannot be used
for tropospheric measurements. On NOAA K, L, M, the primary sounding instru-
ment will be the AMSU. HIRS/3 will be used when cloud conditions permit to
improve vertical resolution.
and A M S U - A 2 .
Chapter 11 The Future
a
Notation: χ ± y ± z\ χ is the center frequency. If y appears, the center frequency is not sensed,
but t w o bands, one on either side of the center frequency, are sensed; y is the distance from the center
frequency to the center of the two passbands. If ζ appears, it is the width of the two passbands. This
pattern is easily implemented with radiofrequency receivers, and it effectively doubles the signal (two
passbands instead of one).
V = vertical, Η = horizontal, R = rotates with scan angle.
b
with contributions from Japan, Europe, and Canada as well as from non-NASA
U.S. agencies. The space component of MTPE will include satellites in sunsynchro-
nous orbit, geostationary orbit, and other orbits as required to make the observa-
tions necessary to understand the Earth system. The centerpiece of the MTPE is
the Earth Observing System (EOS)—"a series of polar-orbiting and low inclination
satellites for long-term global observations of the land surface, biosphere, solid
Earth, atmosphere, and oceans" (Asrar and Dokken, 1993). MTPE is currently
envisioned as having three phases: a pre-EOS era, the EOS era, and a post-EOS
or EOS-follow-on era.
2
Most of the information in Section 11.2 came from the 1993 edition of the EOS Reference
Handbook (Asrar and Dokken, 1993), which is the most comprehensive and up-to-data source on
future plans for Earth observation that we have found. NASA publishes a new edition of the EOS
Reference Handbook approximately annually. We strongly suggest that readers obtain a copy of the
latest edition. (See the bibliography for the address.)
11.2 Mission to Planet Earth 407
The pre-EOS era satellites are operational and experimental satellites that were
planned before the advent of EOS and are scheduled for launch before the first
EOS satellite in 1998. Table 11.5 lists currently planned instruments and satellites
that are part of the pre-EOS MTPE program. In keeping with the MTPE philoso-
phy, most of the satellites are interdisciplinary—they make measurements of
several components of the Earth system. We briefly discuss two yet-to-be-launched
satellites that will make extensive atmospheric measurements.
77.27.7 TRMM
The Tropical Rainfall Measuring Mission (TRMM) is a joint NASA/NASDA
mission designed to measure tropical rainfall and its diurnal variability on a
monthly time scale and in areas ("boxes") of 10 km . TRMM will be launched
5 2
on a Japanese H-2 rocket and will fly in a 350-km orbit with a 35° inclination
angle, which causes it to sample all local times each 23 days. Five instruments,
which measure various related aspects of precipitation, are planned for TRMM:
(Provided by NASDA.)
• TRMM Microwave Imager (TMI)—A five-channel passive microwave im-
ager making measurements from 10 to 91 GHz. (Provided by NASA.)
• Visible Infrared Scanner (VIRS)—A five-channel imaging radiometer
(0.63, 1.6, 3.75, 10.7, and 12.0 μηι) with nominal 2-km resolution at na-
dir and a 1500-km swath width; similar in design to AVHRR/3. (Pro-
vided by NASA.)
• Lightning Imaging Sensor (LIS)—See Section 11.2.2.10. (Provided by
NASA.)
• Clouds and Earth's Radiant Energy System (CERES)—See Section
11.2.2.4. (Provided by NASA.)
The basic idea of TRMM is to mount all of the instruments that can be
used to measure precipitation from satellite platforms on one satellite, where the
measurements can be compared and combined. For example, the 220-km swath
width of the PR samples only a small fraction of the area sensed by the other
instruments, but it will help calibrate the visible, infrared, and microwave sensors
to yield improved precipitation estimates. Similarly, the LIS, which will sense
lightning flashes, will help determine the location of the more active thunder-
storm cells.
77.27.2 ADEOS
The Advanced Earth Observing Satellite (ADEOS) is a joint Japanese/U.S./
French mission to observe the atmosphere, land, and oceans. An 800-km, 98.6°-
inclination sunsynchronous orbit with a 10:30 am equator crossing time is
TABLE 11.5. Mission to Planet Earth Pre-EOS Era Satellites*
N A S A Satellites
ERBS (operating) Radiation budget, aerosol, and ozone data from 57°
Earth Radiation Budget Satellite inclination orbit
T O M S / M e t e o r (operating) Ozone mapping and monitoring (joint with Russia)
Total Ozone Mapping Spectrometer
UARS (operating) Stratospheric and mesospheric chemistry and related
Upper Atmosphere Research Satellite processes
N A S A Spacelab Series (1992 on) A series of Shuttle-based experiments to measure at-
mospheric and solar dynamics (ATLAS), atmo-
spheric aerosols (LITE), and surface radar backscat-
ter polarization, and phase information [SIR—C
and X — S A R (joint with Germany)]
T O P E X / P o s e i d o n (operating) Ocean circulation (joint with France)
Ocean Topography Experiment
LAGEOS 2 (operating) Satellite laser-ranging target for monitoring crustal
Laser Geodynamics Satellite motions and Earth rotation variations (joint with
Italy)
SeaWiFS Purchase of ocean color data to monitor ocean pro-
Sea-Viewing Wide Field Study ductivity
T O M S / E a r t h Probe Ozone mapping and monitoring
Total Ozone Mapping Spectrometer
N S C A T / A D E O S (1996) Ocean surface wind vectors (joint with Japan)
NASA Scatterometer/Advanced Earth
Observing Satellite
T O M S / A D E O S (1996) Ozone mapping and monitoring (joint with Japan)
Total Ozone Mapping Spectrometer/
Advanced Earth Observing Satellite
T R M M (1998) Precipitation, clouds, and radiation processes in lower
Tropical Rainfall Measuring Mission latitudes (joint with Japan)
Landsat 7 (1997) High-spatial-resolution visible and infrared radiance/
Land Remote-Sensing Satellite reflectance to monitor land surface (joint with
DoD)
N o n - N A S A Satellites
N O A A 9 trough J (U.S.—opera- Visible and infrared radiance/reflectance, infrared at-
tional) mospheric sounding, and ozone measurements
Landsat 4 / 5 / 6 (U.S.—operational) High spatial resolution visible and infrared radiance/
Land Remote-Sensing Satellite reflectance
DMSP (U.S.—operational) Visible, infrared, and passive microwave atmospheric
Defense Meteorological Satellite and surface measurements
Program
ERS 1 (ESA—pre-operational) C-band SAR, microwave altimeter, scatterometer, and
European Remote-Sensing Satellite sea surface temperature
JERS 1 (ESA—preoperational) L-band SAR backscatter and high spatial resolution
Japan's Remote-Sensing Satellite visible and infrared radiance—reflectance
ERS 2 (ESA—1994) Same as ERS 1, plus ozone mapping and monitoring
European Remote-Sensing Satellite
Radarsat (Canada—1995) C-band SAR measurements of Earth's surface (joint
Radar Satellite U.S./Canadian mission)
N O A A Κ through Ν (U.S.—1995 Visible, infrared, and microwave radiance—re-
flectance infrared atmospheric sounding, and ozone
measurements
ADEOS (Japan—1996) Visible and near-infrared radiance—reflectance, scat-
Advanced Earth Observing Satellite terometery, and tropospheric and stratospheric
chemistry (joint USA and France)
a
See Asrar and Dokken (1993).
11.2 Mission to Planet Earth 409
system. (Readers are again referred to the EOS Reference Handbooks.) Table
11.6 lists the currently planned EOS-era satellites.
In the remainder of Section 11.2.2, we cannot resist listing the instruments 3
that are currently planned for the EOS-era satellites and noting what, in our
view, are the important aspects for satellite meteorology and related fields. The
descriptions are sketchy and will probably change, but we think it provides readers
3
Abbreviations may be found in Appendix B.
4] 2 Chapter 11 The Future
with information that they need about the future of satellite meteorology. Much
more information can be found in Asrar and Dokken (1993).
/1.2.2.1 ACRIM
The Active Cavity Radiometer Irradiance Monitor, selected for flight on the
EOS-CHEM Series satellites, will extend the NASA/ACRIM solar luminosity
data base to three decades, which is essential for understanding the variability of
the sun's output and its affect on the Earth's atmosphere. [See Willson and
Hudson (1991).]
11.2.2.3 ASTER
The Advanced Spaceborne Thermal Emission and Reflection Radiometer
(ASTER) will have 14 channels between 0.5 and 12 μηι and spatial resolutions
between 15 and 90 m. The high-spatial-resolution images of clouds and the
Earth's surface will allow atmospheric scientists to better understand the physical
processes that affect climate change. In addition, ASTER will have an aft pointing
channel, which will allow stereo images of clouds to be made as the satellite flies
over. This should considerably increase our knowledge of cloud properties. ASTER
will complement the more frequent return period but lower-spatial-resolution
MODIS. ASTER is planned to fly on the EOS-AMI satellite. [See Fujisada and
Ono (1991).]
11.2.2.4 CERES
The Clouds and Earth's Radiant Energy System (CERES) instrument is a follow-
on of the ERBE Scanner. It will have a cross-track scanner and a biaxial scanner,
each with three channels: shortwave, longwave, and total. CERES will fly on the
EOS-AM and EOS-PM satellite series as well as TRMM 1. [See Barkstrom (1990).]
11.2.2.5 DORIS, SSALT, and TMR
The Solid-State Altimeter (SSALT) is a radar altimeter that measures ocean
wave height, wind speed, and information on the ocean surface current velocity.
It also maps the topography of the sea surface and polar ice sheets. DORIS
(Doppler Orbitography and Radiopositioning Integrated by Satellite) precisely
locates the satellite in orbit to correct the SSALT measurements. TMR (TOPEX
Microwave Radiometer) provides atmospheric water vapor measurements for
11.2 Mission to Planet Earth 413
112.2.6 EOS-COLOR
The EOS Ocean Color Instrument (EOS-COLOR) is a follow-on to the Nimbus
7 Coastal Zone Color Scanner and SeaWiFS, and it is designed primarily to
measure marine phytoplankton. However, it will yield information on the global
carbon cycle, and it may also be useful for marine aerosol studies. EOS-COLOR
is the only instrument slated for the EOS-COLOR satellite.
11.2.2.7 EOSP
The Earth Observing Scanning Polarimeter (EOSP) is derived from planetary
instruments on the Pioneer and Galileo satellites. It measures radiance and linear
polarization of reflected sunlight in 12 spectral bands from 0.41 to 2.25 μηι.
EOSP data will provide information on the global aerosol and cloud distribution
and on such properties as optical depth, phase, particle size, and cloud-top pres-
sure. EOSP is scheduled for flight on the EOS-AM2 and - A M 3 satellites.
11.2.2.8 GLAS
The Geoscience Laser Altimeter System (GLAS) will measure cloud heights,
planetary boundary-layer heights, and vertical aerosol structure, as well as ice
sheet, land, and water topography. GLAS is planned for flight on the EOS-ALT
satellite. [See Cohen et al. (1987) and Curran et al. (1987a).]
11.2.2.9 HIRDLS
The High-Resolution Dynamics Limb Sounder (HIRDLS) is the latest in a long
line of limb sounders (LRIR, LIMS, SAMS, ISAMS, CLAES). It measures gas and
aerosol concentrations in the upper troposphere, stratosphere, and mesosphere
using the spectral range 6-18 μπι. HIRDLS has been selected for the EOS-CHEM
Series satellites.
11.2.2.10 LIS
The Lightning Imaging Sensor (LIS) is a staring imager (no moving parts)
designed to detect, locate, and quantize the radiant energy of lightning flashes. It
utilizes a two-dimensional (128 x 128) CCD array to image a 600 x 600-km
area with 5-km resolution. The array is read out each 2 ms and processed to
yield the location of lightning flashes since the last readout. To be flown on
TRMM 1, LIS will complement the other precipitation-sensing instruments by
yielding an indication of storm intensity—the lightning flash rate. LIS will detect
both cloud-to-cloud and cloud-to-ground lightning, whereas ground-based light-
ning sensors detect only cloud-to-ground strokes. LIS will detect lightning both
day and night over both land and ocean. On TRMM, of course, each location
will be sensed only twice per day for a few minutes as the satellite passes over.
Someday, a successor instrument may be flown on a geostationary satellite (see
414 Chapter 11 The Future
112.111 MIMR
The Multifrequency Imaging Microwave Radiometer (MIMR) is derived from
the SSM/I on the DMSP satellites and from the SMMR on the Nimbus 7 and
Seasat satellites. It will sense horizontally and vertically polarized radiation at six
frequencies between 6.8 and 90 GHz in a 50° conical scan with ground resolution
of from 5 to 60 km, depending on frequency. MIMR will extend previous measure-
ments of a wide variety of geophysical parameters by including more frequencies
and by having a somewhat greater swath width. MIMR is scheduled for flight
on the EOS-PM Series.
11.2112 MISR
The Multi-Angle Imaging SpectoRadiometer (MISR) will be the only EOS
instrument that will routinely provide multiangle observations. Using nine separate
charge-coupled device (CCD) pushbroom cameras, MISR will make images at
nine fixed angles in the along-track direction: one at nadir, and four both fore
and aft out to ±70.5°. Four spectral bands centered at 0.443, 0.555, 0.670, and
0.865 μπι will be simultaneously imaged. Ground resolution will be selectable
between 240 m and 1.92 km. Due to the fact that each point will be imaged nine
times as the satellite passes over, stereo observations are automatic. MISR will
be used for cloud, aerosol, and radiation budget studies as well as for land,
biosphere, and ocean studies. MISR is scheduled for flight on the EOS-AM
Series satellites.
11.2.2.13 MLS
The Microwave Limb Sounder (MLS) is derived from the UARS MLS. It will
measure temperature, moisture, and a variety of gases in the stratosphere, includ-
ing the chlorine compounds involved in ozone chemistry. MLS will fly on the
EOS-CHEM Series.
11.2.2.14 MODIS
The Moderate-Resolution Imaging Spectrometer (MODIS) (Esias et al., 1986;
Salomonson and Toll, 1991; King et al., 1992) is derived from the AVHRR, from
the Nimbus 7 Coastal Zone Color Scanner, and from Landsat. It is designed to
make 0.25-1-km resolution observations in 36 visible and infrared bands. The
orbit and scan geometries are to be such that each point on Earth is sensed every
day or two. A wide variety of oceanic, atmospheric, and land surface processes
will be observable with MODIS. Atmospheric applications include cloud and
aerosol observations. Of note are a set of channels in the Oxygen Α-band near
0.76 μηι that will determine cloud-top pressure (by measuring the amount of
oxygen above the cloud), and a pair of polarized 0.5-μ,ηι channels that should
be interesting in both cloud and aerosol studies. Because of its versatility and
synergism with other EOS sensors, MODIS is slated to fly on both the EOS-AM
and - P M satellites.
11.2 Mission to Planet Earth 415
η2115 Μοριπ
Measurements of Pollution in the Troposphere (MOPITT) is a four-channel
gas-correlation spectrometer, similar to the PMR, SSU, SAMS, and ISAM, which
will measure profiles of CO concentration and column-integrated C H amounts. 4
11.1116 NSCAUII
The NASA Scatterometer II (NSCAT II) is derived from the Seasat SASS and
the NSCAT to fly on ADEOS. It will utilize six fan-beam antennas to measure
ocean surface wind speed and direction under all weather conditions. Having six
antennas, as opposed to SASS's four antennas, should improve accuracy and
reduce uncertainty in wind direction. NSCAT II is scheduled for the ADEOS
II satellite.
N 0 and OCIO. SAGE III has been selected for flight on the EOS-AERO and
3
11.2.2.18 SOLSTICE II
The Solar Stellar Irradiance Comparison Experiment II (SOLSTICE II) is de-
rived from the SOLSTICE instrument on UARS. It will measure the solar irradiance
at four ultraviolet wavelengths. SOLSTICE II is scheduled for the EOS-CHEM
Series satellites.
11.2.2.19 TES
The Tropospheric Emission Spectrometer (TES) is a high-resolution infrared
imaging Fourier-transform spectrometer with spectral coverage of 2.3-15.4 μηι
at a spectral resolution of 0.025 c m . It will be capable of imaging virtually all
-1
chemical species that are active in the infrared. TES is slated for the EOS-AM2
and - A M 3 satellites.
Possible satellites in the future of MTPE include the Geostationary Earth Obser-
vatory (GEO) to measure processes that are too fast to be studied by the polar-
orbiting EOS satellites or that require accurate observations of the diurnal cycle.
Also envisioned are additional Earth Probes to make measurements which EOS
satellites cannot. Plans for the EOS follow-on era will be formulated over the
next decade or so.
There have been few active meteorological sensors in space. A Lidar Atmo-
spheric Sounder and Altimeter (LASA) was proposed as part of the Earth Observ-
ing System (Curran et al., 1987a). LASA was to be capable of measuring vertical
profiles of ozone, water vapor, aerosol concentration, temperature, and pressure.
Unfortunately, LASA is not currently scheduled for the Earth Observing System.
A lidar (light detection and ranging) is like radar but uses light instead of radio
waves. A pulse of laser-generated radiation is transmitted, and the amplitude of
the returned (reflected) signal is measured. The reflection can be from the surface
or from atmospheric aerosols or clouds. Since aerosols are ubiquitous, a continu-
ous return is measured. A single laser wavelength is useful for measuring distances.
If the satellite orbit and attitude are accurately known, the height of the ocean
or of the tops of clouds can be measured. Aerosol optical depth profiles can also
be measured by a single-wavelength lidar.
Most atmospheric measurements require two laser wavelengths. To measure
the concentration of a gas (ozone and water vapor, for example) the two wave-
lengths are chosen such that one is affected by a gaseous absorption line and the
other is not. The difference between the returned signals (assuming that other
11.3 Other Possibilities 4]7
4
Except in a relative sense around tropical storms (see section 7.3.3.2).
418 Chapter 11 The Future
would "stare" at the Earth and the entire array would be read out approximately
every 2 ms.
Two major problems confront the Lightning Mapper: filtering out the bright
background during daylight hours and the high data rate (500 frames per second).
Background filtration would rely on (1) the slow time rate of change of the
background versus the approximately 0.5-ms duration of a lightning stroke, (2)
the nonblackbody spectrum of lightning radiation, and (3) differences in spatial
scale of sunlit clouds and lightning flashes. The data-rate problem would have to
be solved with some sort of on-board background subtraction routine, which
might work as follows. If a lightning stroke has occurred in a 10-km pixel, that
pixel should be significantly brighter than the same pixel in the last frame. If the
two frames are subtracted, the only data which need to be transmitted to Earth
are for those small number of pixels that are significantly different than in the
last frame.
There currently exists a network of ground-based lightning detectors in the
United States. The Lightning Mapper would significantly enhance this network
in two ways. Of course, it would extend the range to a large fraction of the
hemisphere visible from geostationary orbit. It also would be able to detect both
cloud-to-cloud and cloud-to-ground lightning. Ground-based detectors only sense
cloud-to-ground strokes. A wide range of storm physics, atmospheric electricity,
and atmospheric chemistry applications would benefit from the Lightning Map-
per data.
Though it has been discussed for years, a microwave imager—sounder for use
in geostationary orbit is not in current plans. Because of its ability to measure
atmospheric temperature, precipitation, and a wide variety of other important
atmospheric and surface properties through clouds, we believe that a geostationary
microwave instrument is extremely important. The large antenna needed for
this instrument presents technical challenges, but we believe that the return on
investment would be substantial.
Lastly, we feel compelled to reiterate our suggestion (Kidder and Vonder Haar,
1990) that meteorological instruments on satellites in Molniya orbit (see Chapter
2) could significantly improve observations of the high latitudes, including Alaska,
northern Europe, and the North and South Poles..
It has been more than thirty years since the launch of the first meteorological
satellites. In that time satellite meteorology has grown rapidly. If even a fraction
11.4 Α Final Comment
of the planned satellite instruments are launched in the coming decades, we expect
that growth not only to continue but to accelerate. To those of you who have
just finished reading this book, we hope that it will be of use in whatever role
you play in this exciting and expanding field.
Bibliography
AIRS Science Team (1990). The Atmospheric Infrared Sounder (AIRS) Science and Measurements
Requirements. Jet Propulsion Laboratory document N o . D6665, Pasadena, CA.
Asrar, G., and D . J. Dokken (eds.) (1993). EOS Reference Handbook. NASA, Washington, D C .
[Available from the Earth Science Support Office, Document Resource Facility, 3 0 0 D Street, SW,
Suite 8 4 0 , Washington, D C 2 0 0 2 4 , telephone: (202) 4 7 9 - 0 3 6 0 . ]
Atlas, R., E. Kalnay, W. E. Baker, J. Susskind, D . Reuter, and M. Halem (1985). Simulation studies
of the impact of future observing systems on weather prediction. Preprints: Seventh Conference
on Numerical Weather Prediction, American Meteorological Society, Boston, pp. 1 4 5 - 1 5 1 .
Barkstrom, Β. R. (1990). Long-term monitoring of the Earth's radiation budget. Proceedings of the
Society of Photo-Optical Instrumentation Engineers, vol. 1299.
Bunting, J. T., and R. P. d'Entremont (1982). Improved Cloud Detection Utilizing Defense Meteorolog-
ical Satellite Program Near Infrared Measurements. Air Force Geophysics Lab. Tech. Rep. AFGL-
T R - 8 2 - 0 0 2 7 , Hanscom AFB, MA.
Butler, D. M., R. E. Hartle, M. Abbott, R. Cess, R. Chase, P. Christensen, J. Dutton, L.-L. Fu, C.
Gautier, J. Gille, R. Gurney, P. Hays, J. Hovermale, S. Martin, J. Melack, D. Miller, V. Mohnen,
B. Moore III, J. Norman, S. Schneider, J. Sherman III, V. Suomi, B. Tapley, R. Watts, E. W o o d ,
and P. Zinke (1986). From Pattern to Process: The Strategy of the Earth Observing System. Eos
Science Steering Committee Report, Vol. II, NASA, Washington, D C .
Christian, H. J., R. J. Blakeslee, and S. J. Goodman (1989). The detection of lightning from geostation-
ary orbit. / . Geophys. Res., 9 4 , 1 3 3 2 9 - 1 3 3 3 7 .
Christian, H. J., R. J. Blakeslee, and L. J. Goodman (1992). Lightning Imaging Sensor (LIS) for
the Earth Observing System. N A S A Tech. M e m o . 4 3 5 0 , NASA/Marshall Space Flight Center,
Huntsville, AL 3 5 8 1 2 .
Cohen, S. J. Degnan, J. Bufton, J. Garvin, and J. Abshire (1987). The Geoscience Laser Altimetry/
Ranging System. IEEE Trans. Geoscience Remote Sensing, GE-25, 5 8 1 - 5 9 2 .
Curran, R. J., R. R. Nelms, F. Allario, R. A. Bindschadler, Ε. V. Browell, J. L. Bufton, R. L. Byer,
S. Cohen, J. J. Degnan, W. B. Grant, R. Greenstone, W. S. Heaps, Β. M . Herman, A. Jalink, D .
K. Killinger, L. Korb, J. B. Laudenslager, Μ. P. McCormick, S. H. Melfi, R. T. Menzies, V. Mohnen,
J. Spinhirne, and H. J. Zwally (1987a). LASA—Lidar Atmospheric Sounder and Altimeter. Earth
Observing System Instrument Panel Report, Vol. lid, NASA, Washington, D C .
Curran, R. J., D. E. Fitzjarrald, J. W. Bilbro, V. J. Abreu, R. A. Anthes, W. E. Baker, D. A. Bowdle,
D. Burridge, G. D. Emmitt, S. E. D . Ferry, P. H. Flamant, R. Greenstone, R. M. Hardesty, K. R.
Hardy, R. M. Huffaker, M. P. McCormick, R. T. Menzies, R. M. Schotland, J. K. Sparkman, Jr., J.
M. Vaughan, and C. Werner (1987b). LAWS—Laser Atmospheric Wind Sounder. Earth Observing
System Instrument Panel Report, Vol. Ilg, NASA, Washington, D C .
Esias, W., W. Barnes, M. Abbott, S. Cox, R. Evans, R. Fraser, A. Goetz, C. Justice, E. P. McClain,
M. Maxwell, R. Murphy, J. Prospero, B. Rock, S. Running, R. Smith, J. Solomon, and J. Susskind
(1986). MODIS—Moderate-Resolution Imaging Spectrometer. Earth Observing System Instru-
ment Panel Report, Vol. IIb, NASA, Washington, D C .
Fujisada, Η. and Μ. O n o (1991). Overview of ASTER design concept. Proceedings of the Society of
Photo-Optical Instrumentation Engineers, vol. 1490.
Kidder, S. Q., and Τ. H. Vonder Haar (1990). On the use of satellites in Molniya orbits for meteorologi-
cal observation of middle and high latitudes. / . Atmos. Ocean. Tech., 7, 5 1 7 - 5 2 2 .
King, M. D . , Y. J. Kaufman, W. P. Menzel, and D. Tanre (1992). Remotes sensing of cloud, aerosol,
and water vapor properties from M O D I S . IEEE Trans. Geosci. Remote Sens., 3 0 , 2 - 2 7 .
420 Chapter 11 The Future
NASA (1988a). Performance Specification for the NOAA-K through -M Satellites. GSFC-S-480-25,
Goddard Space Flight Center, Greenbelt, M D .
NASA (1988b). Performance Specification for the NOAA-K, L&M Advanced Very High Resolution
Radiometerl3 (AVHRR/3) Flight Models A301, A302, and A303. GSFC-S-480-28.3, Revision B,
Goddard Space Flight Center, Greenbelt, M D .
N A S A (1988c). Performance Specification for the NOAA-K, L&M High Resolution Infrared
Radiation Sounder 13 (HIRS/3) Flight Models H301, H302, andH303. GSFC-S-480-28.2, Revision
B, Goddard Space Flight Center, Greenbelt, M D .
Salomonson, V. V., and D. L. Toll (1991). The Moderate Resolution Imaging Spectrometer-Nadir
( M O D I S - N ) facility instrument. Adv. Space Res., 1 1 , 2 3 1 - 2 3 6 .
Willson, R. C , and H. S. Hudson (1991). The sun's luminosity over a complete solar cycle. Nature,
351, 4 2 - 4 5 .
A
APPENDIX
List of
Meteorological
Satellites
since 1957 made atmospheric measurements. The list contains only successfully
launched satellites. Manned flights are listed only if they included a specific atmo-
spheric experiment.
The reader should be aware that the list probably includes some satellites that
did not make atmospheric measurements, and it certainly omits some that did.
1
Thompson, T. D . (ed.) ( 1 9 8 8 - 1 9 9 3 ) . TRW Space Log, Vols. 2 3 - 2 9 . T R W Space & Technology
Group, One Space Park, Mail Station R 4 / 2 1 3 5 , Redondo Beach, CA 9 0 2 7 8 .
Dubach, L. L., and C. N g (1988). Compendium
2
of Meteorological Space Programs, Satellites,
and Experiments. National Space Science Data Center, Goddard Space Flight Center, Greenbelt, M D .
421
422 Appendix A
ches for that year, and a letter that identifies the various spacecraft which
resulted from that launch. For example, 1984 108B refers to the second
spacecraft (B) resulting from the 108th successful launch in 1984. Before
1963 a different numbering system was used. We have attempted to re-
number these early launches using the TRW Space Log as a guide.
• The Date of launch is for Coordinated Universal Time (UTC).
• For the launch Site, we have used a consistent name, even though names
have changed through the years. For example, we have used "Cape Ca-
nav." to refer to the launch sites at Cape Canaveral, Florida; they are vari-
ously called the Kennedy Space Center, the Eastern Test Range, and the
Eastern Space and Missile Center.
• The Payload Mass is in many cases approximate.
• The Orbital Data are taken from the TRW Space Logs. The Period has
been adjusted slightly using Eq. 2.16 so that it agrees with the Perigee,
Apogee, and Inclination. Note that for satellites still in orbit these are the
data that were current when the Space Logs were compiled. They are not
necessarily representative of the orbital data during the useful lifetime of
the satellite. For example, the inclination angle of geostationary satellites
is maintained near zero. When they are no longer operational, their incli-
nation angles drift. The Compendium contains "initial" orbital data,
which is often more representative of the data when the satellite was oper-
ational. For the sake of consistency, however, we have used the Space Log
data.
• In the Comments section, we have tried to list the instrumentation which
the satellite carried or the type of measurements which it made. In some
cases this information is missing, and in others it may be incorrect.
"Successful" here means that something reached orbit. Whether any useful data were collected
3
is another question.
Name Infi ID Country LAUNCH DATA Payload ORBITAL DATA Comments
or Date Site Vehicle Mass Per- Peri- Apo- Incli-
Agency (kg) iod gee gee) nation
(min) (km) (km) (de9)
Vanguard 2 1959 002A NASA Feb 17 Cape Canav. Vanguard 9.8 123.6 559 3142 32.9 Optical Scanner; wobble degraded data
Explorer 6 1959 006A NASA Aug 7 C a p e Canav. Thor Able 64 764.4 245 42400 47.0 F i r s t E a r t h p h o t o ; first r a d i a t i o n b u d g e t
data
Explorer 7 1959 012A NASA Oct 13 Cape Canav. Juno 2 41.5 101.3 556 1088 50.2 Earth radiation budget d a t a
TIROS 1 1960 002A NASA Apr1 C a p e Canav. Thor Able 120 98.7 677 722 48.4 First d e d i c a t e d m e t e o r o l o g i c a l satellite;
1 T V - W A , 1 T V - N A ; returned 2 2 9 5 5
p h o t o s u p to 1 7 J u n 1 9 6 0
Midas 2 1960 006A USAF May 24 C a p e Canav. Atlas A g e n a A 2 3 0 0 94.3 484 511 33.0 Non-Scanning Radiometer,
atmospheric density
Explorer 8 1960 014A NASA Nov 3 Cape Canav. Juno 2 41 106.2 407 1699 49.9 Ionospheric research
TIROS 2 1960 016A NASA Nov 23 C a p e Canav. Delta 130 97.1 585 667 48.5 1 T V - W A , 1 TV-NA, Scanning
Radiometer, Widefield Radiometer
Explorer 9 1961 004A NASA Feb 16 W a l l o p s Is. Scout 6.6 118.2 634 2583 38.6 3 . 7 m balloon to study a t m o s p h e r i c
density
TIROS 3 1961 017A NASA Jul 12 C a p e Canav. Delta 129 100.0 730 801 47.9 2 T V - W A , Scanning Radiometer,
Widefield Radiometer, Suomi
Radiometer
Midas 3 1961 018A USAF Jul 12 Vandenberg Atlas A g e n a Β 1 6 0 0 161.5 3345 3538 91.1 Scanning Radiometer
Midas 4 1961 028A USAF Oct 21 Vandenberg Atlas A g e n a Β 1 8 0 0 166.0 3485 3760 91.2 Scanning Radiometer
Discoverer 3 5 1961 030A USAF Nov 15 Vandenberg Thor Agena Β 2100 89.8 238 278 81.6 Scanning Radiometer
TIROS 4 1962 002A NASA Feb 8 C a p e Canav. Delta 129 100.3 712 840 48.3 1 T V - W A , 1 TV-ΜΑ, Scanning
Radiometer, Widefield Radiometer,
Suomi Radiometer
Midas 5 1 9 6 2 0 1 OA USAF Apr 9 Vandenberg Atlas A g e n a Β 1 8 6 0 153.0 2782 3406 86.7 Scanning Radiometer
Kosmos 4 1962 014A USSR Apr 2 9 Tyuratam A-1 4600 90.6 285 317 65.0 Possible met. applications
Ariel 1 1962 015A NASA/UK Apr 2 6 C a p e Canav. Delta 60 100.9 389 1214 53.9 Joint U S - U K ionospheric satellite
TIROS 5 1962 025A NASA Jun 19 C a p e Canav. Delta 129 99.8 578 922 58.1 1 TV-WA, 1 TV-MA
Kosmos 7 1962 033A USSR Jul 2 8 Tyuratam A-1 4600 90.1 197 356 65 Possible met. applications
TIROS 6 1962 047A NASA Sep 18 C a p e Canav. Delta 127 98.7 686 713 58.3 1 TV-WA, 1 TV-MA
Kosmos 9 1962 048A USSR Sep 27 Tyuratam A-1 4600 90.9 292 346 65 Possible met. applications
Explorer 17 1963 009A NASA Apr 2 C a p e Canav. Delta 185 96.4 255 917 57.6 A t m o s p h e r i c r e s e a r c h d a t a until 1 0 J u l
1963
Kosmos 14 1 9 6 3 0 1 OA USSR Apr 13 K a p u s t i n Y a r B-1 400 92.0 252 499 49.0 Possible metsat
Kosmos 15 1963 011A USSR Apr 22 Tyuratam A-1 4730 89.7 160 358 65.0 Possible metsat
Midas 6 1963 014A USAF May 9 Vandenberg Atlas A g e n a Β 2 0 0 0 166.5 3606 3677 87.3 Scanning Radiometer
TIROS 7 1963 024A NASA Jun 19 C a p e Canav. Delta 135 95.8 548 572 58.2 2 T V - W A , Scanning Radiometer, Suomi
Radiometer
Kosmos 23 1963 050A USSR Dec 13 K a p u s t i n Y a r B-1 347 93.0 240 613 49 Possible metsat
Explorer 19 1963 053A NASA Dec 19 Vandenberg Scout 7 115.9 590 2395 78.8 3 . 7 m balloon identical to Explorer 9
424
Name Infi ID Country LAUNCH DATA Payload ORBITAL DATA Comments
or Date Site Vehicle Mass Per- Peri- Apo- Incli-
Agency (kg) iod gee gee) nation
(min) (km) (km) de
( 9)
TIROS 8 1963 054A NASA Dec 21 Cape Canav. Delta 119 98.8 682 725 58.5 First A P T , 1 T V - W A
Ariel 2 1964 015A NASA/UK Mar 27 W a l l o p s Is. Scout 168 101.4 810 843 51.6 R e t u r n e d d a t a f r o m British e x p e r i m e n t s
until N o v 1 9 6 4
Explorer 2 0 1964 051A NASA Aug 25 Vandenberg Scout 44 103.8 859 1006 79.9 R e t u r n e d i o n o s p h e r i c d a t a until J u l y
1966
Nimbus 1 1964 052A NASA Aug 28 Vandenberg Thor Agena Β 376 98.5 429 937 98.6 First s u n s y n c h r o n o u s m e t s a t ; A V C S ,
APT, HRIR
Kosmos 44 1964 053A USSR Aug 28 Tyuratam A-1 3800 99.0 604 813 65.1 Probable metsat
Kosmos 45 1964 055A USSR Sep 13 Tyuratam A-1 4730 89.7 207 313 64.9 W e a t h e r r e s e a r c h film c a p s u l e
recovered
Explorer 2 2 1964 064A NASA Oct 10 Vandenberg Scout 52 104.6 875 1062 79.7 Ionospheric a n d geodetic d a t a satellite
Kosmos 49 1964 069A USSR Oct 2 4 Kapustin Yar B-1 355 91.7 264 466 49 M e a s u r e d Earth's IR a n d U V radiation
flux
Explorer 2 4 1964 076A NASA Nov 21 Vandenberg Scout 8.6 116.4 525 2498 81.4 3 . 7 m balloon for a t m o s p h e r i c density
s t u d i e s ; p a r t o f first N A S A dual-payload
launch
San Marco 1 1964 084A Italy Dec 15 W a l l o p s Is. Scout 254 94.9 198 846 37.8 M e a s u r e d atmospheric density
TIROS 9 1965 004A NASA Jan 22 Cape Canav. Delta 138 119.1 702 2568 96.4 Cartwheel configuration; 2 T V - W A ,
global coverage
Kosmos 58 1965 014A USSR Feb 26 Tyuratam A-1 4730 94.6 470 523 65.0 Probable metsat
Gemini 3 1965 024A NASA Mar 23 C a p e Canav. Titan 2 3220 88.2 160 240 32.5 Synoptic W e a t h e r Photography
Kosmos 65 1965 029A USSR Apr 17 Tyuratam A-1 4730 89.8 207 319 65 W e a t h e r r e s e a r c h film c a p s u l e
recovered
Explorer 2 7 1965 032A NASA Apr 2 9 W a l l o p s Is. Scout 59.7 107.6 933 1315 41.2 Geodetic a n d ionospheric research
satellite; battery failed 1 9 6 8
Gemini 4 1965 043A NASA Jun 3 C a p e Canav. Titan 2 3569 88.7 162 281 32.0 Synoptic W e a t h e r Photography
T I R O S 10 1965 051A NASA Jul 2 Cape Canav. Delta 127 100.4 728 817 98.4 Sunsynchronous; 2 T V - W A
Gemini 5 1965 068A NASA Aug 21 C a p e Canav. Titan 2 3600 89.3 197 303 32.6 Synoptic W e a t h e r Photography, Cloud
Top Specrometer (Oxygen A Band),
Space Object Radiometry
Kosmos 92 1965 083A USSR Oct 16 Tyuratam A-1 4730 89.9 202 334 65 W e a t h e r r e a s e a r c h film c a p s u l e
recovered
Alouette 2 1965 098A Canada Nov 29 Vandenberg Thor Agena Β 145 119.9 503 2837 79.8 Ionospheric research
Explorer 31 1965 098B NASA Nov 29 Vandenberg Thor Agena Β 99 120.6 502 2906 79.8 Ionospheric data complementing
Alouette 2; l a u n c h e d with Alouette 2
Gemini 7 1965 100A NASA Dec 4 C a p e Canav. Titan 2 3658 90.1 292 298 28.9 Synoptic W e a t h e r Photography, S p a c e
Object Radiometry
Gemini 6A 1965 104A NASA Dec 15 Cape Canav. Titan 2 3546 89.5 258 271 28.9 Flight d e l a y e d ; S y n o p t i c W e a t h e r
Photography
Name Infi ID Country LAUNCH DATA Payload ORBITAL DATA Comments
or Date Site Vehicle Mass Per- Peri- Apo- Incli-
Agency (kg) iod gee gee) nation
(min) (km) (km) (deg)
Kosmos 100 1965 106A USSR Dec 17 Tyuratam A-1 3800 96.5 533 641 65 Probable metsat
ESSA1 1966 008A ESSA Feb 3 Cape Canav. Delta 138 100.0 689 818 97.9 2 wide-angle vidicon c a m e r a s ; switched
off 8 M a y 1 9 6 7
ESSA2 1966 016A ESSA Feb 28 Cape Canav. Thor Agena D 132 113.5 1352 1412 101.2 2 A P T ; global operational A P T
Gemini 8 1966 020A NASA Mar 16 C a p e Canav. Titan 2 3789 88.5 159 265 28.9 Synoptic W e a t h e r Photography, Cloud
T o p Spectrometer (no data)
Molniya-1 3 1966 035A USSR Apr 25 Tyuratam A-2-e 1750 710.6 506 39492 65 Comsat, also transmitted cloud cover
photos
Kosmos 118 1966 038A USSR M a y 11 Tyuratam A-1 4730 93.0 415 428 65 Probable metsat
Nimbus 2 1966 040A NASA May 15 Vandenberg TAT-Agena Β 414 108.1 1092 1175 100.4 AVCS, HRIR, MRIR, APT/DRIR
Explorer 3 2 1966 044A NASA May 25 Cape Canav. Delta 225 99.3 250 1194 64.5 Six a e r o n o m y experiments; returned
d a t a until D e c 1 9 6 6
Gemini 9 1966 047A NASA Jun 3 Cape Canav. Titan 2 3750 89.7 270 272 28.9 Synoptic W e a t h e r Photography
Kosmos 121 1966 054A USSR Jun 17 Plesetsk A-1 4730 89.9 200 333 72.9 Possible metsat
Kosmos 122 1966 057A USSR Jun 25 Tyuratam A-1 4730 94.3 469 495 65 Dual Vidicon C a m e r a s , Scanning H R I R ,
A c t i n o m e t r i c Inst.
Gemini 10 1966 066A NASA Jul 1 8 C a p e Canav. Titan 2 3757 88.5 160 268 28.9 Synoptic W e a t h e r Photography
G e m i n i 11 1966 081A NASA Sep 12 C a p e Canav. Titan 2 3793 88.6 161 280 28.8 Synoptic W e a t h e r Photography
D M S P - 4 A F1 1966 082A USAF S e p 16 Vandenberg Thor Burner 2 125 100.5 680 872 98.8 Video Camera System, C System
ESSA3 1966 087A ESSA Oct 2 Vandenberg Thor Agena D 145 114.6 1384 1484 100.9 2 A V C S , F P R ; r e p l a c e d E S S A 1 in T O S
system; silenced 9 Oct 1968
Molniya-1 4 1966 092A USSR Oct 2 0 Tyuratam A-2-e 1750 714.5 505 39685 64.9 Comsat, also transmitted cloud cover
photos
Gemini 12 1966 104A NASA N o v 11 C a p e Canav. Titan 2 3655 89.8 243 310 28.8 Synoptic W e a t h e r Photography
ATS 1 1966110A NASA Dec 7 Cape Canav. Atlas-Agena D 352 1434.9 35734 35797 10.5 First g e o s y n c h r o n o u s m e t . d a t a ; S p i n
Scan Cloudcover Camera
ESSA4 1967 006A ESSA Jan 26 Vandenberg Thor Agena D 290 113.5 1324 1437 102.0 2 A P T ; r e p l a c e d E S S A 2; c a m e r a failure
caused deactivation 6 D e c 1 9 6 7
DMSP-4A F2 1 9 6 7 0 1 OA USAF Feb 8 Vandenberg Thor Burner 2 250 101.3 778 854 98.9 Video Camera System, C System
Kosmos 144 1967 018A USSR Feb 28 Plesetsk A-1 4730 97.0 574 644 81.2 M e t s a t similar to K o s m o s 1 2 2
Kosmos 149 1967 024A USSR Mar 21 Kapustin Y a r B-1 375 89.7 245 285 48.4 3 ch. wide-angle radiometers, narrow-
a n g l e IR radiometer, 3 ch.
telephotometer, T V camera system
ESSA5 1967 036A ESSA Apr 2 0 Vandenberg Thor Agena D 145 113.6 1352 1419 102.0 2 A V C S , FPR; replaced E S S A 3
San Marco 2 1967 038A Italy Apr 2 6 Indian O c e a n Scout 129 90.5 135 498 2.9 E q u a t o r i a l l a u n c h ; r e t u r n e d air density
data
Kosmos 156 1967 039A USSR Apr 27 Plesetsk A-1 4730 94.2 467 483 81.2 Dual Vidicon C a m e r a s , Scanning H R I R ,
A c t i n o m e t r i c Inst.
Ariel 3 1967 042A UK May 5 Vandenberg Scout 90 95.8 497 608 80.2 First all-British r e s e a r c h satellite
425
426
Name Infi ID Country LAUNCH DATA Payload ORBITAL DATA Comments
or Date Site Vehicle Mass Per- Peri- Apo- Incli-
Agency (kg) iod gee gee) nation
(min) (km) (km) (de9)
Molniya-1 5 1967 052A USSR May 24 Tyuratam A-2-e 1750 715.6 460 39785 64.8 Comsat, also transmitted cloud cover
photos
Aurora 1 1967 065B USA/USN Jun 29 Vandenberg Thor Burner 2 21 172.2 3793 3946 89.8 I n v e s t i g a t e d f o r m a t i o n of A u r o r a
Borealis
DODGE 1967 066F DOD/NASA Jul 1 C a p e C a n a v . Titan 3 C 102 1319.0 33270 33659 6.2 D O D Gravity Experiment; Dual Vidicon
C a m e r a s ; first c o l o r p i c s f r o m n e a r - g e o
altitude
DMSP-4A F3 1967 080A USAF Aug 23 Vandenberg Thor Burner 2 195 102.0 822 878 98.8 Video C a m e r a System, C System
Molniya-1 6 1967 095A USSR Oct 3 Tyuratam A-2-e 1750 718.2 502 39868 65 Comsat, also transmitted cloud cover
photos
DMSP-4A F4 1967 096A USAF O c t 11 Vandenberg Thor Burner 2 195 99.6 650 822 99.2 Video C a m e r a System, C System
Molniya-1 7 1967 101A USSR Oct 22 Tyuratam A-2-e 1750 715.1 508 39710 64.7 Comsat, also transmitted cloud cover
photos
Kosmos 184 1967 102A USSR Oct 25 Plesetsk A-1 4730 93.7 441 459 81.2 Metsat
Kosmos 185 1967 104A USSR Oct 2 7 Tyuratam F-1-m 4000 98.7 518 873 64.1 Probable metsat
ATS 3 1967 111A NASA Nov 5 C a p e Canav. Atlas A g e n a D 3 6 5 1436.0 35730 35844 9.2 Multicolor S p i n S c a n Cloudcover
Camera
ESSA6 1967 114A ESSA Nov 10 Vandenberg Thor Agena D 132 114.9 1407 1483 102.1 2 APT; deactivated 4 Nov 1969
WRESAT 1 1967 118A Australia Nov 29 Woomera SPARTA 45 99.4 193 1259 83.2 R e t u r n e d solar radiation a n d upper
atmospheric data
OV3 6 1967 120A USAF Dec 5 Vandenberg Scout 95 93.2 412 439 90.6 Conducted ionospheric studies;
transmission e n d e d 5th day
Kosmos 206 1968 019A USSR Mar 14 Plesetsk A-1 4730 93.6 449 437 81.2 Returned w e a t h e r a n d infrared photos,
thermal data
Apollo 6 1968 025A NASA Apr 4 C a p e Canav. Saturn 5 36806 90.2 205 392 32.5 E a r t h orbit; E a r t h p h o t o g r a p h y
Molniya-1 8 1968 035A USSR Apr 21 Tyuratam A-2-e 1750 713.3 391 39738 65 Comsat, also transmitted cloud cover
photos
D M S P - 4 B F1 1968 042A USAF May 23 Vandenberg Thor Burner 2 195 102.0 809 888 98.8 Video C a m e r a System, C System
Kosmos 226 1968 049A USSR Jun 12 Plesetsk A-1 4730 97.0 579 639 81.2 Returned cloud cover photos, weather
data
Molniya-1 9 1968 057A USSR Jul 5 Tyuratam A-2-e 1750 714.8 401 39803 65 Comsat, also transmitted cloud cover
photos
O V 1 15 1968 059A USAF Jul 11 Vandenberg Atlas F 470 105.0 154 1818 89.8 Air density/solar radiation correlation
experiment
OV1 16 1968 059B USAF Jul 11 Vandenberg Atlas F 600 91.8 163 554 89.7 Ionospheric drag experiment; launched
w i t h O V I 15
Kosmos 232 1968 060A USSR Jul 16 Plesetsk A-1 4730 88.9 189 248 65 Possible metsat
Explorer 3 9 1968 066A NASA Aug 8 Vandenberg Scout 9.3 105.0 633 1341 80.6 Air D e n s i t y Explorer; density & t e m p ,
variations at intermediate latitudes
Name lnt'l I D Country LAUNCH DATA Payload ORBITAL DATA Comments
or Date Site Vehicle Mass Per- Peri- Apo- Incli-
Agency (kg) iod gee gee) nation
(min) (km) (km)
ESSA7 1968 069A ESSA Aug 16 Vandenberg Long-Tank Delta145 115.0 1429 1471 101.8 2 A V C S , F P R , S-band; replaced E S S A
5; d e a c t i v a t e d 1 9 Jul 1 9 6 9
Kosmos 243 1968 080A USSR Sep 23 Tyuratam A-2 5900 89.6 213 293 71.3 C a r r i e d first m i c r o w a v e r a d i o m e t e r ,
m e a s u r e d w a t e r v a p o r & liquid w a t e r
OV2 5 1968 081A USAF Sep 26 C a p e C a n a v . Titan 3 C 204 Comprehensive environmental
r e s e a r c h satellite, carried 11
experiments
Aurorae 1968 084A ESRO Oct 3 Vandenberg Scout 81 103.1 258 1538 93.7 E S R O 1; Investigated auroral
p h e n o m e n a & polar ionosphere
Molniya-1 10 1968 085A USSR Oct 5 Tyuratam A-2-e 1750 712.1 436 39633 65 Comsat, also transmitted cloud cover
photos
Apollo 7 1968 089A NASA O c t 11 Cape Canav. Saturn 1Β 14674 89.5 231 297 31.64 E a r t h orbit; E a r t h p h o t o g r a p h y
D M S P - 4 B F2 1968 092A USAF Oct 23 Vandenberg Thor Burner 2 195 101.3 792 838 98.5 Video Camera System, C System
Kosmos 258 1968 111A USSR Dec 10 Tyuratam A-1 4730 89.4 202 282 65 Possible metsat
ESSA8 1968 114A ESSA Dec 15 Vandenberg Long-Tank Delta132 114.7 1411 1461 101.2 2 APT
Kosmos 261 1968 117A USSR Dec 20 Plesetsk B-1 347 93.1 207 642 71 S o v i e t b l o c k j o i n t e x p e r i m e n t s in a i r
density, polar auroras
Apollo 8 1968 118A NASA Dec 21 C a p e Canav. Saturn 5 28833 - - - - L u n a r orbit; E a r t h p h o t o g r a p h y
ISIS 1 1969 009A Canada Jan 30 Vandenberg TAI Delta 241 128.1 575 3494 88.4 Continued ionospheric studies begun
with Alouette 1
ESSA9 1969 016A ESSA Feb 26 C a p e Canav. TAI Delta 145 115.3 1423 1502 102 Final E S S A launch; 2 A V C S , F P R , S -
band; replaced E S S A 7
Apollo 9 1969 018A NASA Mar 3 C a p e Canav. Saturn 5 36511 88.6 203 229 32.6 E a r t h orbit; E a r t h p h o t o g r a p h y
OV1 18 1969 025B USAF Mar 18 Vandenberg Atlas F 125 95.2 466 583 98.8 Studied i o n o s p h e r e m e a s u r i n g radio
interference, electric fields, radiation
Meteor-1 1 1969 029A USSR Mar 26 Plesetsk A-1 3800 97.2 600 639 81.2 O p e r a t i o n a l m e t s a t ; cloud, s n o w , ice
cover data during daylight & dark
Nimbus 3 1969 037A NASA Apr 14 Vandenberg Thor Agena D 575 107.4 1070 1130 99.6 IRIS, M R I R , HRIR, IDCS, M U S E , SIRS,
IRLS
Apollo 10 1969 043A NASA May 18 Cape Canav. Saturn 5 42530 - - - - L u n a r orbit; E a r t h p h o t o g r a p h y
Apollo 11 1969 059A NASA Jul 16 Cape Canav. Saturn 5 43811 - - - - L u n a r orbit; E a r t h p h o t o g r a p h y
DMSP-4B F3 1969 062A USAF Jul 2 3 Vandenberg Thor Burner 2 195 101.2 775 844 98.5 Video Camera System, C System
Boreas 1969 083A ESRO Oct 1 Vandenberg Scout 80 91.5 291 389 86.0 E S R O 1B; Ionospheric and auroral
e x p e r i m e n t s ; early d e c a y d u e to
booster malfunction
Meteor-1 2 1969 084A USSR Oct 6 Plesetsk A-1 3800 96.7 573 613 81.2 M o n i t o r e d c l o u d a n d ice cover, IR, a n d
thermal energy reflected and radiated
from Earth's a t m o s p h e r e
427
428
Name Int'l I D Country LAUNCH DATA Payload ORBITAL DATA Comments
or Date Site Vehicle Mass Per- Peri- Apo- Incli-
Agency (kg) iod gee gee) nation
de
(min) (km) (km) ( 9)
Interkosmos 1 1969 088A USSR Oct 14 Kapustin Yar B-1 320 93.3 254 626 48.4 Experiments from 3 East European
countries to study sun's U V & X - r a y
effects on upper a t m o s p h e r e
Apollo 12 1969 099A NASA Nov 14 Cape Canav. Saturn 5 43848 - - - - L u n a r orbit; E a r t h p h o t o g r a p h y
Interkosmos 2 1969 110A USSR Dec 25 Kapustin Yar B-1 320 98.4 200 1178 48.4 Carried East European-made
ionospheric observation a n d survey
experiments
Kosmos 320 1970 005A USSR Jan 16 Kapustin Yar B-1 375 90.1 247 326 48.5 Possible metsat
ITOS 1970 008A NASA Jan 23 Vandenberg TAT-Delta Μ 309 115.1 1432 1477 101.8 T I R O S M ; first of 2 n d g e n e r a t i o n U S
metsats; 2 A P T , 2 A V C S , 2 S R , F P R , 3-
axis stabilization
D M S P - 5 A F1 1970 012A USAF F e b 11 Vandenberg Thor Burner 2 195 101.1 759 850 98.8 Scanning Radiometer
Meteor-1 3 1970 019A USSR Mar 17 Plesetsk A-1 3800 93.5 419 457 81.1 -
Nimbus 4 1970 025A NASA Apr 8 Vandenberg Thorad-Agena D620 107.2 1087 1097 99.5 IDCS, IRIS, BUV, S C R , F W S , THIR,
SIRS, MUSE, IRLS
Apollo 13 1970 029A NASA Apr 11 Cape Canav. Saturn 5 43924 - - - - L u n a r orbit; E a r t h p h o t o g r a p h y
Meteor-1 4 1970 037A USSR Apr 2 8 Plesetsk A-1 3800 98.2 625 710 81.2 -
Meteor-1 5 1970 047A USSR Jun 23 Plesetsk A-1 3300 102.0 821 879 81.2 -
Interkosmos 3 1970 057A USSR Aug 7 Kapustin Yar B-1 340 99.7 200 1295 48.4 Studied ionospheric protons, electons,
a n d alpha particles
DMSP-5A F2 1970 070A USAF Sep 3 Vandenberg Thor Burner 2 101.4 764 874 99.1 Scanning Radiometer
Interkosmos 4 1970 084A USSR Oct 14 K a p u s t i n Y a r B-1 320 93.5 255 649 43.3 Studied solar U V a n d X-ray effects
Meteor-1 6 1970 085A USSR Oct 15 Plesetsk A-1 3800 96.2 565 574 81.2 -
Kosmos 384 1970 105A USSR Dec 10 Plesetsk A-2 5900 89.5 203 292 72.8 Microwave Radiometers, Narrow-Angle
IR R a d i o m e t e r
NOAA 1 1970 106A NOAA D e c 11 Vandenberg Long-Tank Delta306 114.9 1422 1471 101.8 S e c o n d in t h e I T O S s e r i e s ; 2 A P T , 2
AVCS, 2 SR, FPR
PEOLE 1 1970 109A France Dec 12 Kourou Diamant Β 70 97.0 517 747 15.0 Preliminary E O L E ; gathered data from a
met. balloon system
Kosmos 389 1970 113A USSR Dec 18 Plesetsk A-1 3800 97.6 598 679 81.2 Possible metsat
Meteor-1 7 1971 003A USSR Jan 20 Plesetsk A-1 3800 96.7 586 603 81.2 -
Apollo 14 1971 008A NASA Jan 31 C a p e Canav. Saturn 5 44456 - - - - L u n a r orbit; E a r t h p h o t o g r a p h y
DMSP-5A F3 1971 012A USAF Feb 17 Vandenberg Thor Burner 2 195 100.7 755 817 98.3 Scanning Radiometer
ISIS 2 1971 024A Canada Apr 1 Vandenberg Thor Delta 264 113.7 1354 1423 88.2 3 r d in a s e r i e s o f i n t e r n a t i o n a l
ionospheric studies satellite
Meteor-1 8 1971 031A USSR Apr 17 Plesetsk A-1 3800 95.3 512 540 81.2 -
Salyut 1 1971 032A USSR Apr 19 Tyuratam D-1 18500 88.5 200 210 51.5 S p a c e s t a t i o n ; m e t . o b s . b y c r e w of
S o y u z 11
San Marco 3 1971 036A Italy Apr 2 4 San Marco Scout 164 93.6 222 718 3.2 3 r d Italian a t m o s p h e r i c satellite
Name Infi ID Country LAUNCH DATA Payload ORBITAL DATA Comments
or Date Site Vehicle Mass Per- Peri- Apo- Incli-
Agency (kg) iod gee gee) nation
(min) (km) (km) de
( 9)
Meteor-1 9 1971 059A USSR Jul 16 Plesetsk A-1 3800 95.7 538 554 81.2
Apollo 15 1971 063A NASA Jul 2 6 C a p e Canav. Saturn 5 46723 - - - - L u n a r orbit; E a r t h p h o t o g r a p h y
EOLE 1 1971 071A France Aug 16 W a l l o p s Is. Scout 84 100.1 664 870 50.1 G a t h e r e d data from a met. balloon
system
D M S P - 5 B F1 1971 087A USAF Oct 14 Vandenberg Thor Burner 2 195 101.5 782 865 99.1 Scanning Radiometer
Interkosmos 5 1971 104A USSR Dec 2 Kapustin Yar B-1 340 98.5 198 1181 48.5 U S S R a n d C z e c h i n s t r u m e n t s to s t u d y
e f f e c t s of s o l a r a c t i v i t y o n n e a r - E a r t h
radiation
Ariel 4 1971 109A UK D e c 11 Vandenberg Scout 99.5 95.5 477 593 82.99 U K - 4 ; 4 British a n d 1 U S e x p e r i m e n t s to
survey t h e i o n o s p h e r e a n d radio
signals
Oreol 1 1971 119A France/USSR Dec 27 Plesetsk C-1 630 111.4 391 2184 74.0 Studied upper atmosphere
in h i g h l a t i t u d e s a n d t h e p o l a r l i g h t s ; 1 s t
satellite l a u n c h e d by U S S R with F r e n c h
Assistance
Meteor-1 10 1971 120A USSR Dec 29 Plesetsk A-1 3300 102.6 835 922 81.3 -
DMSP-5B F2 1972 018A USAF Mar 24 Vandenberg Thor Burner 2 195 101.6 787 868 99.1 Scanning Radiometer, V T P R (SSE)
M e t e o r - 1 11 1972 022A USSR Mar 30 Plesetsk A-1 3300 102.5 858 883 81.2 -
Interkosmos 6 1972 027A USSR Apr 7 Tyuratam A-2 6000 89.0 203 256 51.8 -
Apollo 16 1972 031A NASA Apr 16 C a p e Canav. Saturn 5 46733 - - - L u n a r orbit; E a r t h p h o t o g r a p h y
Interkosmos 7 1972 047A USSR Jun 30 K a p u s t i n Y a r B-1 375 92.6 260 555 48.4 -
Meteor-1 12 1972 049A USSR Jun 30 Plesetsk A-1 3300 102.9 880 898 81.2 -
Landsat 1 1972 058A NASA Jul 2 3 Vandenberg Delta 816.4 103.2 899 911 98.8 Also called E R T S 1; Multispectral
Scanner ( M S S ) , Return B e a m Vidicon
(RBV)
Denpa 1972 064A Japan Aug 19 Kagoshima Mu4S 75 154.3 232 6093 31.0 Investigated ionospheric p h e n o m e n a
NOAA 2 1972 082A NOAA Oct 15 Vandenberg Delta 344 115.0 1447 1453 101.6 First o p e r a t i o n a l s o u n d e r ; 2 V H R R , 2
VTPR, 2 SR, S P M
Meteor-1 13 1972 085A USSR Oct 26 Plesetsk A-1 3300 102.5 858 883 81.2 -
DMSP-5B F3 1972 089A USAF Nov 9 Vandenberg Thor Burner 2 195 101.5 797 855 98.8 Scanning Radiometer, V T P R (SSE)
E S R O IV 1972 092A ESRO Nov 22 Vandenberg Scout 114 99.2 245 1187 91.1 Investigated polar ionosphere
Interkosmos 8 1972 094A USSR Nov 30 Plesetsk B-1 340 93.6 214 679 71 -
Apollo 17 1972 096A NASA Dec 7 Cape Canav. Saturn 5 46743 - - - - L u n a r orbit; E a r t h p h o t o g r a p h y
Nimbus 5 1972 097A NASA D e c 11 Vandenberg Delta 772 107.2 1087 1100 99.7 First s c a n n i n g m i c r o w a v e r a d i o m e t e r ;
THIR, S C M R , E S M R , ITPR, S C R , N E M S
Aeros 1 1972 100A W. Germany Dec 16 Vandenberg Scout 127 95.7 223 867 96.9 U p p e r air r e s e a r c h satellite
Meteor-1 14 1973 015A USSR Mar 20 Plesetsk A-1 2000 102.6 866 883 81.2 -
Interkosmos 9 1973 022A USSR Apr 19 K a p u s t i n Y a r B-1 400 102.1 199 1526 48.4 "Copurnicus-500" solar radiation
monitor
429
430
Name Infi ID Country LAUNCH DATA Payload ORBITAL DATA Comments
or Date Site Vehicle Mass Per- Peri- Apo- Incli-
Agency (kg) iod gee gee) nation
(min) (km) ( k m ) ( d e 9)
Skylab 1973 027A NASA M a y 14 Cape Canav. Saturn 5 74783 93.1 422 442 50.0 Infrared Spectrometer, M i c r o w a v e
Radiometer/Scatterometer/Altimeter.L-
Band Microwave Radiometer,
Multispectral S c a n n e r
Meteor-1 15 1973 034A USSR May 29 Plesetsk A-1 2000 102.4 842 891 81.2 -
DMSP-5B F4 1973 054A USAF Aug 17 Vandenberg Thor Burner 2 195 101.3 795 836 98.5 Scanning Radiometer
Interkosmos 10 1973 0 8 2 A USSR Oct 30 Plesetsk C-1 550 102.2 260 1454 74.0 S c i e n t i f i c r e s e a r c h in c o o p e r a t i o n w i t h
Czechoslovakia and East G e r m a n y
NOAA 3 1973 086A NOAA Nov 6 Vandenberg Delta 345 116.2 1499 1508 101.7 2 VHRR, 2 VTPR, 2 SR, S P M
Oreol 2 1973 107A France/USSR Dec 26 Plesetsk C-1 550 106.1 393 1691 74.0 -
Meteor-1 16 1974 011A USSR Mar 5 Plesetsk A-1 2200 102.2 825 885 81.2 -
DMSP-5B F5 1974 015A USAF Mar 16 Vandenberg Thor Burner 2 A 195 101.3 767 859 99.0 Scanning Radiometer, V T P R (SSE)
Meteor-1 17 1974 025A USSR Apr 2 4 Plesetsk A-1 2200 102.5 855 887 81.2 -
SMS 1 1974 033A NASA M a y 17 Cape Canav. Thorad Delta 627 1340.2 32345 35439 1.9 First d e d i c a t e d g e o s y n c h r o n o u s
metsat; V I S S R , S E M , D C P
I n t e r k o s m o s 11 1 9 7 4 0 3 4 A USSR May 17 Kapustin Yar C-1 400 95.3 483 591 50.6 -
ATS 6 1974 039A NASA May 30 Cape Canav. Titan 3 C 930 1411.9 35191 35438 8.0 Geosynchronous Very High Resolution
Radiometer
Salut 3 1974 046A USSR Jun 24 Tyuratam D-1 18500 89.3 232 257 51.1 S p a c e station; met. obs. by S o y u z
crews
Meteor-1 18 1974 052A USSR Jul 9 Plesetsk A-1 2200 103.1 887 911 81.2 -
Aeros 2 1974 055A W . G e r m a n y Jul 16 Vandenberg Scout 127 95.7 224 869 97.4 U p p e r air r e s e a r c h satellite
D M S P - 5 C F1 1974 063A USAF Aug 9 Vandenberg Thor Burner 2 A 195 101.6 792 862 98.7 Scanning Radiometer, V T P R (SSE)
Meteor-1 19 1974 083A USSR Oct 2 8 Plesetsk A-1 2200 102.4 834 899 81.2 -
Interkosmos 12 1974 0 8 6 A USSR Oct 31 Plesetsk C-1 400 94.2 243 707 74.0 Atmosphere-ionosphere research
NOAA 4 1974 089A NOAA Nov 15 Vandenberg Thorad Delta 340 115.0 1443 1457 101.4 2 VHRR, 2 VTPR, 2 SR, S P M
Meteor-1 2 0 1974 099A USSR Dec 17 Plesetsk A-1 2200 102.3 841 882 81.2 -
Salyut 4 1974 104A USSR Dec 26 Tyuratam D-1 18500 91.3 337 350 51.6 S p a c e station; met. obs. by S o y u z
crews
Landsat 2 1975 004A NASA Jan 22 Vandenberg Delta 953 103.3 901 914 98.9 Multispectral S c a n n e r ( M S S ) , Return
B e a m Vidicon ( R B V )
SMS 2 1975 011A NASA Feb 6 Cape Canav. Delta 627 1436.5 35752 35844 1.8 VISSR, SEM, DCP
Interkosmos 13 1975 0 2 2 A USSR Mar 27 Plesetsk C-1 400 104.5 285 1643 82.9 U p p e r a t m o s p h e r e r e s e a r c h satellite
Meteor-1 21 1975 023A USSR Apr1 Plesetsk A-1 2200 102.5 858 887 81.2 -
Ariabata 1975 033A India Apr 19 Kapustin Yar C-1 360 95.0 512 537 50.7 A t m o s p h e r i c research satellite
DMSP-5C F2 1975 043A USAF May 24 Vandenberg Thor Burner 2 194 101.8 797 881 98.7 Scanning Radiometer, V T P R (SSE)
Nimbus 6 1975 052A NASA Jun 12 Vandenberg Delta 829 107.5 1099 1112 99.6 LRIR, PMR, T W E R L E , ERB, THIR,
HIRS/1, SCAMS, ESMR, T & D R E
Name Infi ID Country LAUNCH DATA Payload ORBITAL DATA Comments
or Date Site Vehicle Mass Per- Peri- Apo- Incli-
Agency (kg) iod gee gee) nation
(min) (km) (km) (de9)
Meteor-2 1 1975 064A USSR Jul 11 Plesetsk A-1 2800 102.4 847 885 81.3 E x p e r i m e n t a l w e a t h e r satellite
Soyuz 19 1975 065A USSR Jul 15 Tyuratam A-2 6182 88.5 186 220 51.8 S o v i e t p a r t of A p o l l o - S o y u z
Apollo 18 1975 066A NASA Jul 15 Cape Canav. Saturn 1Β 14768 88.4 170 228 51.8 U S p a r t of A p o l l o - S o y u z
Meteor-1 22 1975 087A USSR Sep 18 Plesetsk A-1 2200 102.3 801 920 81.3 -
Explorer 5 4 1975 096A NASA Oct 6 Vandenberg Delta 676 118.7 141 3093 90.0 A t m o s p h e r i c satellite
GOES 1 1975 100A NASA Oct 16 C a p e Canav. Delta 627 1436.7 35797 35804 6.7 First o p e r a t i o n a l g e o s y n c h r o n o u s
metsat; V I S S R , S E M , D C P
Explorer 5 5 1975 107A NASA Nov 20 C a p e Canav. Delta 721 93.2 447 449 19.6 A t m o s p h e r i c satellite
Interkosmos 14 1975 115A USSR D e c 11 Plesetsk C-1 700 96.6 300 883 74.0 -
Meteor-1 23 1975 124A USSR Dec 25 Plesetsk A-1 2200 102.3 842 885 81.3 -
UME 1976 019A Japan Feb 29 T a n e g a s h i m a N-1 139 105.2 988 1009 69.7 Ionospheric S o u n d i n g Satellite ( I S S 1)
Meteor-1 2 4 1976 032A USSR Apr 7 Plesetsk A-1 2200 102.2 831 888 81.3 -
Meteor-1 25 1976 043A USSR M a y 15 Plesetsk A-1 2200 102.2 832 884 81.3 -
Interkosmos 15 1976 0 5 6 A USSR Jun 19 Plesetsk C-1 550 94.7 483 515 74.0 -
Salyut 5 1976 057A USSR Jun 22 Tyuratam D-1 19000 89.2 230 245 51.5 S p a c e station; met. obs. by S o y u z
crews
Interkosmos 16 1976 0 7 6 A USSR Jul 2 7 Kapustin Y a r C-1 550 94.3 463 518 50.5 Solar X-rays
NOAA 5 1976 077A NOAA Jul 2 9 Vandenberg Delta 340 116.3 1503 1519 101.9 2 V H R R , 2 V T P R , 2 S R , S P M ; I T O S H,
l a s t of t h e I T O S s e r i e s
D M S P - 5 D F1 1976 091A USAF S e p 11 Vandenberg Thor Burner 2 450 101.4 806 832 98.6 Block 5 D - 1 ; O L S , S S H
Meteor-1 26 1976 102A USSR Oct 15 Plesetsk A-1 2200 102.4 848 885 81.3 -
Meteor-2 2 1977 002A USSR Jan 6 Plesetsk A-1 2750 102.9 881 898 81.3 -
Meteor-1 27 1977 024A USSR Apr 5 Plesetsk A-1 2200 102.4 846 892 81.3 -
DMSP-5D F2 1977 044A USAF Jun 5 Vandenberg Thor Burner 2 450 101.4 789 853 99.0 Block 5 D - 1 ; O L S , S S H
GOES 2 1977 048A NOAA Jun 16 Cape Canav. Delta 627 1436.1 35757 35821 5.8 VISSR, SEM, DCP
Meteor-1 28 1977 057A USSR Jun 29 Tyuratam A-1 2200 96.0 543 583 97.4 First s u n s y n c h r o n o u s Meteor
GMS 1 1977 065A Japan Jul 14 C a p e Canav. Delta 670 1435.9 35771 35801 4.8 Also called Himawari; V I S S R
Interkosmos 17 1977 0 9 6 A USSR Sep 24 Plesetsk C-1 550 94.5 467 511 82.9 -
Salyut 6 1977 097A USSR Sep 29 Tyuratam D-1 9000 92.2 380 391 51.6 S p a c e station; met. obs. by S o y u z
crews
Meteosat 1 1977 108A ESA Nov 23 C a p e Canav. Delta 697 1436.1 34755 36823 6.1 Imaging Radiometer (0.5-0.9, 5.7-7.1,
and 10.5-12.5 micrometers)
Meteor-2 3 1977 117A USSR Dec 14 Plesetsk A-1 2700 102.4 846 886 81.2 -
Kyokko 1978 014A Japan Feb 4 Kagoshima Mu3H 103 134.2 636 3967 65.4 Scientific satellite investigating a u r o r a e
UME 2 1978 018A Japan Feb 16 Tanegashima N-1 140 107.2 974 1217 69.4 Ionospheric S o u n d i n g Satellite ( I S S )
Landsat 3 1978 026A NASA Mar 5 Vandenberg Delta 960 103.3 897 919 98.8 Multispectral S c a n n e r ( M S S ) , Return
B e a m Vidicon ( R B V )
431
432
Name Infi ID Country LAUNCH DATA Payload ORBITAL DATA Comments
or Date Site Vehicle Mass Per- Peri- Apo- Incli-
Agency (kg) iod gee gee) nation
(min) (km) ( k m ) (dec))
HCMM 1978 041A NASA Apr 2 6 Vandenberg Scout-F 134 97.3 618 627 97.6 Heat Capacity Mapping Mission ( A E M
1)
DMSP-5D F3 1978 042A USAF May 1 Vandenberg Thor Burner 2 513 101.2 804 817 98.6 Block 5 D - 1 ; O L S , S S H
GOES 3 1978 062A NOAA Jun 16 C a p e Canav. Delta 627 1436.0 35781 35795 3.7 VISSR, SEM, DCP
Seasat 1978 064A NASA Jun 27 Vandenberg Atlas/Agena D2300 100.5 778 783 108.0 E x p e r i m e n t a l o c e a n s u r v e y satellite;
failed after 9 0 days
TIROS Ν 1978 096A NASA Oct 13 Vandenberg Atlas F 1421 101.9 835 853 99.0 F i r s t of 3 r d g e n e r a t i o n m e t s a t s ;
AVHRR/1, HIRS/2, SSU, M S U , SEM,
DCS
Nimbus 7 1978 098A NASA Oct 24 Vandenberg Delta 907 104.1 927 969 99.4 L a s t of N i m b u s s e r i e s ; S M M R ,
S B U V / T O M S , C Z C S , E R B , S A M II,
LIMS, S A M S , THIR
I n t e r k o s m o s 1fJ 1 9 7 8 0 9 9 A USSR Oct 2 4 Plesetsk C-1 1050 92.2 320 428 82.9 -
Meteor-1 2 9 1979 005A USSR Jan 25 Tyuratam A-1 3800 97.1 581 645 97.6 S u n s y n c h r o n o u s orbit
SAGE 1979 013A NASA Feb 18 W a l l o p s Is. Scout-F 147 94.4 460 521 55.0 Stratospheric aerosol and gas
e x p e r i m e n t ( A E M 2); solar occultation
experiment
Solwind P78-1 1979 017A US Feb 24 Vandenberg Atlas 1331 95.4 513 547 97.8 Ionosphere and magnetosphere
research; destroyed by U S A F A S A T 13
Sep 1985
Interkosmos 19 1 9 7 9 0 2 0 A USSR Feb 27 Plesetsk C-1 1015 98.4 483 874 74.0 -
Meteor-2 4 1979 021A USSR Marl Plesetsk A-1 3800 102.2 845 872 81.2 -
DMSP-5D F4 1979 050A USAF Jun 6 Vandenberg Thor Burner 2 513 101.3 806 828 98.7 Block 5 D - 1 ; O L S , S S H , S S M / T , S S C
(1.6 micrometers)
Bhaskara 1 1979 051A India Jun 7 Kapustin Yar C-1 441 92.7 407 412 50.7 D u a l T V C a m e r a , Satellite M i c r o w a v e
Radiometer
NOAA 6 1979 057A NOAA Jun 27 Vandenberg Atlas F 1421 101.1 797 813 98.5 AVHRR/1, HIRS/2, SSU, M S U , S E M ,
DCS
Meteor-2 5 1979 095A USSR Oct 31 Plesetsk A-1 3800 102.6 866 882 81.2 -
Interkosmos 2 0 1979 0 9 6 A USSR Nov1 Plesetsk C-1 1100 91.7 343 363 74.0 -
SMM 1980 014A NASA Feb 14 C a p e Canav. Delta 2500 94.1 488 493 28.5 Solar M a x i m u m Mission; carries Active
C a v i t y R a d i o m e t e r s to m e a s u r e solar
constant
Meteor-1 30 1980 051A USSR Jun 18 Tyuratam A-1 3475 96.2 540 600 97.7 C o m b i n e d meteorological and Earth
resources satellite
Meteor-2 6 1980 073A USSR Sep 9 Plesetsk A-1 3300 102.4 840 889 81.2 -
GOES 4 1980 074A NOAA Sep 9 C a p e Canav. Delta 835 1435.9 35771 35802 3.1 First V A S ; V A S , S E M , D C P
Interkosmos 21 1981 0 1 1 A USSR Feb 6 Plesetsk C-1 550 92.9 397 427 74.0 O c e a n o g r a p h i c a n d terrestrial
experiments
Name Infi ID Country LAUNCH DATA Payload ORBITAL DATA Comments
or Date Site Vehicle Mass Per- Peri- Apo- Incli-
Agency (kg) iod gee gee) nation
(min) (km) (km) (deg)
Meteor-2 7 1981 043A USSR May 14 Plesetsk A-1 2750 102.4 849 887 81.3
GOES 5 1981 049A NOAA May 22 Cape Canav. Delta 835 1435.8 35781 35788 0.1 VAS, SEM, DCP
Meteosat 2 1981 057A ESA Jun 19 Kourou Ariane 4 697 1436.2 35790 35793 0.3 Imaging Radiometer (0.5-0.9, 5.7-7.1,
and 10.5-12.5 micrometers)
NOAA 7 1981 059A NOAA Jun 23 Vandenberg Atlas F 1421 102.0 838 858 99.1 AVHRR/2, HIRS/2, SSU, MSU, SEM,
DCS
Meteor-1 31 1981 065A USSR Jul 1 0 Plesetsk A-1 2200 97.5 607 659 97.7 Last Meteor-1
DE 1 1981 070A NASA Aug 3 Vandenberg Delta 403 410.6 464 23370 89.5 Dynamics Explorer 1
DE 2 1981 070B NASA Aug 3 Vandenberg Delta - 97.8 298 996 90.0 Dynamics Explorer 2
Interkosmos 2 2 1981 0 7 5 A USSR Aug 7 Plesetsk A-1 1500 101.9 795 890 81.2 Ionospheric and magnetospheric
research
GMS 2 1981 076A Japan Aug 10 Tanegashima N-2 670 1436.0 35766 35811 2.9 VISSR
Oreol 3 1981 094A France/ Sep 21 Plesetsk A-2 1000 108.5 400 1903 82.5 Soviet/French Magnetosphere and
USSR Ionosphere Explorer
SME 1981 100A NASA Oct 6 Vandenberg Thor Delta 437 94.8 504 506 97.7 Solar M e s o s p h e r e Explorer
STS2 1981 111A NASA Nov 12 Cape Canav. STS 8516 89.5 259 259 38 N i g h t / D a y O p t i c a l S u r v e y of L i g h t n i n g ,
M e a s u r e m e n t of A i r P o l l u t i o n f r o m S a t . ,
Shuttle Imaging R a d a r A
Bhaskara 2 1981 115A India Nov 20 Kapustin Yar C-1 444 94.7 498 518 50.6 Dual T V C a m e r a , Satellite M i c r o w a v e
Radiometer
Meteor-2 8 1982 025A USSR Mar 25 Plesetsk F-2 2750 104.2 941 962 82.5 F i r s t M e t e o r - 2 in t h i s o r b i t a l r e g i m e
Salyut 7 1982 033A USSR Apr 19 Tyuratam D-1-h 18900 94.0 473 474 51.6 S p a c e station; met. obs. by S o y u z
crews
STS4 1982 065A NASA Jun 27 Cape Canav. STS 11107 90.6 318 318 28.5 N i g h t / D a y O p t i c a l S u r v e y of L i g h t n i n g
Landsat 4 1982 072A NASA Jul 1 6 Vandenberg Thor Delta 1942 98.9 699 701 98.2 Multispectral S c a n n e r ( M S S ) , Thematic
Mapper (TM)
Meteor-2 9 1982 116A USSR Dec 14 Plesetsk A-1 2750 102.0 810 890 81.2 P o s s i b l e r e p l a c e m e n t for M e t e o r - 2 6
D M S P - 5 D F6 1982 118A USAF Dec 21 Vandenberg Atlas Ε - 101.3 811 823 98.7 Block 5 D - 2 ; O L S , S S H - 2 , S S M / T
NOAA 8 1983 022A NOAA Mar 28 Vandenberg Atlas F 1712 101.3 803 825 98.7 AVHRR/1, HIRS/2, SSU, MSU, SEM,
DCS, SAR
STS6 1983 026A NASA Apr 4 Cape Canav. STS 21277 89.9 284 284 28.0 N i q h t / D a y O p t i c a l S u r v e y of L i g h t n i n q
GOES 6 1983 041A NOAA Apr 2 8 C a p e Canav. Delta 3 9 1 4 835 1437.8 35782 35864 0.1 VAS, SEM, DCP
STS7 1983 059A NASA Jun 18 Cape Canav. STS 16817 90.5 314 314 28.0 Modular Optoelectric Multispectral
Scanner
Insat 1B 1983 089B India Aug 31 STS8 PAM-D 1152 1436.0 35636 35939 0.2 L a u n c h e d f r o m S p a c e S h u t t l e ; first 3 -
axis stabilized geosynchronous metsat
Meteor-2 10 1983 109A USSR Oct 2 8 Plesetsk A-1 2750 101.4 752 889 81.1 P h a s e d with M e t e o r - 2 7 a n d M e t e o r - 2 9
DMSP-5D F7 1983 113A USAF Nov 18 Vandenberg Atlas F - 101.5 815 832 98.7 Block 5 D - 2 ; O L S
STS9 1983 116A NASA Nov 28 Cape Canav. STS 15068 89.5 250 250 57.0 Spacelab 1; several met. experiments
433
434
Name Int'l I D Country LAUNCH DATA Payload ORBITAL DATA Comments
or Date Site Vehicle Mass Per- Peri- Apo- Incli-
Agency (kg) iod gee gee) nation
(min) (km) (km) (de9)
STS10 1984 011A NASA Feb 3 Cape Canav. STS 15342 90.8 326 326 28.4 S T S 41-B; Modular Optoelectric
Multispectral S c a n n e r
Ohzora 1984 015A Japan Feb 14 Tanegashima Mu-3S 180 92.9 317 503 74.6 E X O S - C ; Middle Atmosphere
P r o g r a m m e r ; optically s e n s i n g
atmosphere and ionosphere
Landsat 5 1984 021A NASA Marl Vandenberg Delta 1938 98.9 699 700 98.2 Multispectral S c a n n e r ( M S S ) , T h e m a t i c
Mapper (TM)
Uosat 2 1984 021Β UK Marl Vandenberg Delta 52 98.5 674 693 98.2 Magnetospheric studies, Earth
imaging, space dust detection, a m a t e u r
radio digital c o m m u n i c a t i o n s
M e t e o r - 2 11 1984 072A USSR Jul 5 Plesetsk F-2 2750 104.1 939 958 82.5 -
GMS3 1984 080A Japan Aug 2 Tanegashima N-2 303 1435.9 35783 35787 0.8 VISSR
STS12 1984 093A NASA Aug 30 Cape Canav. STS 21524 90.9 331 331 28.4 S T S 41-D; C L O U D S
STS13 1984 108A NASA Oct 5 Cape Canav. STS 10629 91.6 352 352 57.0 S T S 4 1 - G ; M e a s u r e m e n t of Air Pollution
from Sat., Large Format C a m e r a ,
Shuttle Imaging Radar Β
ERBS 1984 108B NASA Oct 5 STS13 Arm 226 96.7 601 605 57.0 C a r r i e s E R B E a n d S A G E II
NOAA 9 1984 123A NOAA Dec 12 Vandenberg Atlas 1712 102.1 841 862 99.0 AVHRR/2, HIRS/2, SSU, M S U , DCS,
ERBE, SBUV/2, SAR
Meteor-2 12 1985 013A USSR Feb 6 Plesetsk F-2 2200 104.1 934 960 82.5 -
Meteor-3 1 1985 100A USSR Oct 24 Plesetsk F-2 - 109.5 1185 1210 82.5 First of 3 r d g e n e r a t i o n m e t s a t s
Meteor-2 13 1985 119A USSR Dec 26 Plesetsk F-2 2200 104.1 936 958 82.5 -
SPOT 1 1986 019A France Feb 22 Kourou Ariane 1 1830 101.5 824 828 98.7 Earth resources
Viking 1986 019B Sweden Feb 22 Kourou Ariane 1 538 258.9 790 13359 98.8 S t u d y i n g magnetic, electric, ultraviolet
p r o p e r t i e s of a u r o r a l r e g i o n s
Meteor-2 14 1986 039A USSR May 27 Plesetsk F-2 2000 104.2 941 960 82.5 -
N O A A 10 1986 073A NOAA Sep 17 Vandenberg Atlas Ε 1712 101.3 808 826 98.7 AVHRR/1, HIRS/2, M S U , S E M , DCS,
ERBE, SAR
Meteor-2 15 1987 001A USSR Jan 5 Plesetsk F-2 2000 104.1 939 957 82.5 -
MOS1 1987 018A Japan Feb 19 Tanegashima N-2 745 103.3 909 909 99.1 M a r i n e Observation Satellite; M E S S R ,
VTIR, M S R
GOES 7 1987 022A NOAA Feb 26 C a p e Canav. Delta 3 9 1 4 835 1436.1 35783 35796 0.1 VAS, SEM, DCP
DMSP-5D F8 1987 053A USAF Jun 20 Vandenberg Atlas Ε 750 102.0 836 856 98.8 Block 5 D - 2 ; O L S , S S H - 2 , S S M / T ,
SSM/I
Meteor-2 16 1987 068A USSR Aug 18 Plesetsk F-2 2000 104.1 940 957 82.6 -
Meteor-2 17 1988 005A USSR Jan 20 Plesetsk F-2 2000 104.1 933 959 82.6 -
DMSP-5D F9 1988 006A USAF Feb 3 Vandenberg Atlas 750 101.4 815 826 98.7 Block 5 D - 2 ; O L S
San Marco 4 1988 026A Italy Mar 25 San Marco Scout 236 93.0 263 615 3.0 U p p e r a t m o s p h e r e studies, joint
Italian/FRG/U.S. mission
Name Infi ID Country LAUNCH DATA Payload ORBITAL DATA Comments
or Date Site Vehicle Mass Per- Peri- Apo- Incli-
Agency (kg) iod gee gee) nation
(min) (km) (km)
Kosmos 1940 1988 034A USSR Apr 2 6 Tyuratum D-1-e 2000 1430.4 35572 35784 0.2 Oceanographic, atmospheric studies;
located at 1 2 ° W
Meteosat 3 1988 051A ESA Jun 15 Kourou Ariane 4 696 1436.1 35786 35794 0.9 First A r i a n e 4 l a u n c h ; s t a t i o n e d a b o v e
0°E
Insat 1 C 1988 063A India Jul 21 Kourou Ariane 3 1190 1436.0 35763 35812 0.2 B a c k u p for Insat 1B; stationed a b o v e
93.5°East
Meteor-3 2 1988 064A USSR Jul 2 6 Plesetsk F-2 2000 109.5 1186 1208 82.5 -
Fengyun-1 1 1988 080A PRC Sep 6 Taiyuan CZ-4 750 102.8 877 899 99.2 E x p e r i m e n t a l m e t s a t ; first C Z - 4 l a u n c h ;
n e w l a u n c h site
N O A A 11 1988 089A NOAA Sep 24 Vandenberg Atlas Ε 1712 102.1 845 863 98.9 AVHRR/2, HIRS/2, MSU, SSU, SBUV/2,
SEM, DCS, S&R
Exos-D 1989 016A Japan Feb 21 Kagoshima M-3S-2 295 205.2 262 10021 75.1 S t u d y i n g a u r o r a bortalis, a u r o r a
australis
Meteor-2 18 1989 018A USSR Feb 28 Plesetsk F-2 1500 104.2 941 960 82.5 -
Meteosat 4 1989 020B ESA Mar 6 Kourou Ariane 4 4 L P 681 1436.0 35779 35798 0.7 Meteosat Operational Program 1 ( M O P
1).
Resurs-F 1 1989 038A USSR May 25 Plesetsk A-2 6300 89.9 254 273 82.3 Remote sensing. Deployed two
satellites
Pion 1 1989 038C USSR Jun 9 Resurs-F 1 A-2 78 89.9 257 270 82.3 Passive a t m o s p h e r i c research satellite
Pion2 1989 038D USSR Jun 9 Resurs-F 1 A-2 78 89.9 257 268 82.3 Passive atmospheric research satellite
Resurs-F 2 1989 049A USSR Jun 27 Plesetsk A-2 6300 90.0 260 274 82.5 Remote sensing
Resurs-F 3 1989 055A USSR Jul 1 2 Plesetsk A-2 6300 90.0 260 275 82.5 Remote sensing. Deployed two
satellites
Pion 3 1989 055C USSR Aug 7 Resurs-F 3 A-2 78 89.9 254 272 82.5 P a s s i v e atmospheric research satellite
Pion 4 1989 055D USSR Aug 7 Resurs-F 3 A-2 78 89.9 255 272 82.5 P a s s i v e atmospheric research satellite
GMS 4 1989 070A Japan Sep 5 Tanegashima H-1 725 1435.9 35780 35792 1.4 -
Interkosmos 2 4 1989 0 8 0 A USSR Sep 28 Plesetsk F-2 1000 116.1 505 2492 82.6 Particles a n d fields exp. US
participation
Meteor-3 3 1989 086A USSR Oct 2 4 Plesetsk F-2 2150 109.6 1190 1213 82.5 -
SPOT 2 1990 005A France Jan 22 Kourou Ariane 4 1870 101.5 821 823 98.7 2 n d in s e r i e s of F r e n c h r e m o t e s e n s i n g
satellites
MOS1B 1990 013A Japan Feb 7 Tanegashima H-1 740 103.3 908 909 99.2 M a r i n e observation satellite. C a l l e d
M o m o 1 Β after launch
LACE 1990 015A US Feb 14 C a p e Canav. Delta 2 1430 95.0 514 539 43.1 Low-power Atmospehric Compensation
E x p e r i m e n t for Strategic D e f e n s e
Initiative Office ( S D I O )
Okean 2 1990 018A USSR Feb 28 Plesetsk F-2 1900 97.7 632 658 82.5 Oceanographic remote sensing
Resurs F6 1990 047A USSR May 29 Plesetsk A-2 6300 89.9 255 268 82.3 Remote sensing
435
436
Name Infi I D Country LAUNCH DATA Payload ORBITAL DATA Comments
or Date Site Vehicle Mass Per- Peri- Apo- Incli-
Agency (kg) iod gee gee) nation
(min) (km) (km) (de9)
Insat 1 D 1990 051A India Jun 12 C a p e Canav. Delta 1190 1436.0 35781 35795 0.1 Domestic comsat/metsat. Last of
lnsat-1 series.
Meteor-2 1 9 1990 057A USSR Jun 27 Plesetsk F-2 1750 104.1 937 961 82.5 -
Resurs F7 1990 060A USSR Jul 1 7 Plesetsk A-2 6300 89.3 193 276 82.3 Remote sensing
Resurs F 8 1990 073A USSR Aug 16 Plesetsk A-2 6300 88.7 176 229 82.3 Remote sensing
Fengyun-1 2 1990 081A PRC Sep 3 Taiyuan CZ-4 881 102.9 881 896 99.0 -
P R C 31 1990 081Β PRC Sep 3 Taiyuan CZ-4 4 100.8 775 804 99.0 Atmospheric balloon
PRC 32 1990 081C PRC Sep 3 Taiyuan CZ-4 4 102.2 833 886 99.0 Atmospheric balloon
Meteor-2 20 1990 086A USSR Sep 28 Plesetsk F-2 1750 104.1 940 958 82.5 -
DMSP-5D F10 1990 105A USAF Dec 1 Vandenberg Atlas Ε 750 100.7 729 845 98.9 Operational even though broken nozzle
p r e v e n t e d it f r o m r e a c h i n g d e s i r e d o r b i t
Meteosat 5 1991 015B ESA Mar 2 Kourou Ariane 4 4 L P 681 1436.1 35787 35792 0.7 Meteosat Operational Program 2 ( M O P
2)
Almaz 1 1991 024A USSR Mar 31 Tyuratam SL-13 18550 90.6 293 305 72.7 R e m o t e Earth sensing; radar imaging
Meteor-3 4 1991 030A USSR Apr 2 4 Plesetsk SL-14 2215 109.5 1184 1210 82.5 -
N O A A 12 1991 032A NOAA May 14 Vandenberg Atlas Ε 1421 101.3 807 825 98.7 N O A A D , o n e of t h e o r i g i n a l T I R O S Ν
type. A V H R R / 1 , H I R S / 2 , M S U , D C S
Resurs-F 10 1991 035A USSR M a y 21 Plesetsk SL-4 6300 89.2 227 231 82.3 Earth resources; reentered 6/20/91
Okean 3 1991 039A USSR Jun 4 Plesetsk SL-14 1900 97.7 627 659 82.5 Oceanography
Resurs-F 11 1991 044A USSR Jun 28 Plesetsk SL-4 6300 89.8 253 268 82.3 Reentered 7/21/91
ERS1 1991 050A ESA Jul 17 Kourou Ariane 4 2384 100.5 774 775 98.5 E S A R e m o t e S e n s i n g satellite;
m i c r o w a v e , I R i m a g i n g of o c e a n s , i c e ,
and land m a s s e s
Meteor-3 5 1991 056A USSR Aug 15 Plesetsk SL-14 2215 109.4 1185 1203 82.6 Carries US-built Total O z o n e Mapping
Spectrometer (TOMS)
IRS 1B 1991 061A India Aug 29 Tyuratam SL-3 980 103.2 890 918 99.2 Indian R e m o t e S e n s i n g satellite;
l a u n c h e d commercially by USSR
UARS 1991 063B NASA S e p 15 STS-48 RMS 6795 96.2 573 579 57.0 U p p e r A t m o s p h e r e R e s e a r c h Satellite
D M S P - 5 D F11 1991 082A USAF Nov 28 Vandenberg Atlas Ε 830 101.9 835 855 98.9 -
JERS 1 1992 007A Japan F e b 11 Tanegashima H1 1340 96.1 566 570 97.7 J a p a n e s e Earth R e s o u r c e s Satellite
STS 45 1992 015A NASA Mar 24 Cape Canav. STS 105843 90.5 292 304 57.0 Carried A T L A S science payload.
Resurs-F 14 1992 024A CIS Apr 2 9 Plesetsk SL-4 6300 89.5 233 256 82.0 Remote sensing. Reentered 5/29/92
SROSS C 1992 028A India May 20 Sriharikota ASLV 106 91.3 256 436 46.0 S t r e c h e d Rohini Satellite Series.
Carried g a m m a ray detector,
ionospheric monitor. R e e n t e r e d 7 / 1 4 / 9 2
Insat 2 A 1992 041A India Jul 9 Kourou Ariane 44L 1906 1437.2 35741 35883 0.1 Indian communications, remote
sensing. Positioned above 74°E.
Topex/ 1992 052A US/France Aug 10 Kourou Ariane 4 2 Ρ 2402 112.4 1330 1342 66.0 O c e a n sensing,mapping mission
Poseidon sponsored by N A S A / C N E S .
Name Int'l I D Country LAUNCH DATA Payload ORBITAL DATA Comments
or Date Site Vehicle Mass Per- Peri- Apo- Incli-
Agency (kg) iod gee gee) nation
(min) (km) ( k m ) (decj)
Kitsat A 1992 052B Korea Aug 10 Kourou Ariane 4 2 P 50 112.0 1304 1325 66.0 First K o r e a n p a y l o a d . S t o r e / f o r w a r d
comm., Earth imaging.
Resurs-F 16 1992 056A CIS Aug 19 Plesetsk SL-4 6300 89.2 222 233 82.5 Remote sensing. Carried U.S. D o D
e x p e r i m e n t to m e a s u r e beryllium a t o m s
in l o w E a r t h o r b i t . R e e n t e r e d 9 / 4 / 9 2
Pion 5 1992 056C CIS Aug 19 Plesetsk SL-4 50 89.3 224 248 82.5 P a s s i v e subsatellite to d e t e r m i n e h o w
upper a t m o s p h e r e affects spacecraft
reentry projections. R e e n t e r e d 9 / 2 5 / 9 2 .
Pion 6 1992 056D CIS Aug 19 Plesetsk SL-4 50 89.3 225 247 82.5 S e e P i o n 5. R e e n t e r e d 9 / 2 4 / 9 2 .
Kosmos 2209 1992 059A CIS Sep 10 Tyuratam SL-12 2200 1435.9 35765 35808 1.1 R e m o t e sensing. Positioned above
24°E.
Freja 1992 064A Sweden Oct 6 Shuang CZ-2C 259 109.0 596 1759 63.0 Swedish ionospheric, magnetospheric
Cheng Tzu studies.
PRC 36 1992 064B PRC Oct 6 Shuang CZ-2C 2100 89.8 211 318 63.0 Fanhul Shi Weixing remote sensing,
Cheng Tzu microgravity experiments. Capsule
reentered 10/13/92; instrument module,
10/31/92.
Kosmos 2224 1992 087A CIS Dec 17 Tyuratam SL-12 2200 1436.0 35696 35878 2.1 Remote sensing. Located above 12°E.
Insat 2 B 1993 048B India Jul 2 2 Kourou Ariane 4 4 L 1931 1436.0 35774 35802 0.1 Telecommunications, weather
forecasting, search and rescue.
Located above 93.5°E.
N O A A 13 1993 050A NOAA Aug 9 Vandenberg Atlas Ε 1712 102.1 845 861 98.9 AVHRR/2, HIRS/2, MSU, SSU, SBUV/2,
S E M , D C S , S & R . Suffered fatal p o w e r
failure 8 / 2 1 / 9 3 .
Meteor-2 21 1993 055A CIS Aug 31 Plesetsk SL-14 2000 104.2 933 967 82.5 -
SPOT 3 1993 061A France Sep 26 Kourou Ariane 4 0 1907 101.5 821 823 98.7 Remote sensing
Kosmos 2265 1993 067A CIS Oct 26 Plesetsk SL-8 500 103.6 290 1558 82.9 Atmospheric density monitor
Meteosat 6 1993 073B ESA Nov 20 Kourou Ariane 4 4 L P 704 1432.5 35674 35757 1.2 Located a b o v e 0°
Meteor-3 6 1994 003A CIS Jan 25 Plesetsk SL-14 2215 109.4 1182 1207 82.6 -
GOES 8 1994 022A NOAA Apr 13 Cape Canav. Atlas 2100 1436.1 35769 35812 0.3 First of G O E S I - M s e r i e s .
DMSP-5D F12 1994 057A USAF Aug 29 Vandenberg Atlas Ε 830 102.0 839 856 98.9 -
N O A A 14 1994 089A NOAA Dec 30 Vandenberg Atlas Ε 1712 102.1 845 858 98.9 -
437
Β
APPENDIX
Abbreviations
ID One-dimensional
2D Two-dimensional
3D Three-dimensional
AATSR Advanced Along-Track Scanning Radiometer
ACRIM Active Cavity Radiometer Irradiance Monitor
ADEOS Advanced Earth Observing Satellite
ADM Angular dependence model
AEM Applications Explorer Mission
AIRS Atmospheric Infrared Sounder
ALT Altimeter
AMSR Advanced Microwave Scanning Radiometer
AMSU Advanced Microwave Sounding Unit
API Antecedent precipitation index
APT Automatic picture transmission
ASAR Advanced Synthetic Aperture Radar
ASTER Advanced Spaceborne Thermal Emission and Reflection Radi-
ometer
ATI Area time integral
439
44Q Appendix Β
Weather Facsimile
WFOV Wide field of view
WMO World Meteorlogical Organization
WSR Weather Service Radar
C
APPENDIX
Principal
Symbols
A Albedo, area
α Semimajor axis
B (T)k Planck function
C Matrix used to retrieve temperatures from radiances
c Speed of light
cx First radiation constant
c2 Second radiation constant
d sun Earth—sun distance
£ Irradiance
e Eccentric anomaly
ECT Equator crossing time
f Coriolis parameter
G Newtonian (or universal) gravitation constant
Η Radiant exposure
h Planck's constant, generalized height coordinate
hs Satellite height (above surface)
^TOA Height of top of atmosphere
i Inclination
447
448 Appendix C
sS E
Line strength
Radiance covariance matrix
ς Solar constant
"^sun
S T
Temperature covariance matrix
s Distance
Τ Period, temperature
f Anomalistic period
f Synodic or nodal period
T B
Brightness temperature
Equivalent blackbody temperature
Flux temperature
ί Time
ίο Epoch time
υ Gravitational potential, column-integrated mass
Geostrophic wind vector
y
g
Weighting function
Molecular weight
Principal Symbols 449
Systeme
International Units
Μ OST OF THE units in this book are Systeme International (SI) Units,
which is the internationally accepted form of the metric system. SI
Units are comprised of base units, supplementary units, and units derived from
the basic units.
Base Units
Length meter m
Mass kilogram kg
Time second s
Temperature kelvin Κ
Amount of substance mole mol
Electrical current ampere A
Luminous intensity candela cd
(continues)
451
Appendix D
Supplementary Units
N a m e d Derived Units
Force newton Ν kg m s~ 2
Area m 2
Volume m 3
Density kg m" 3
Wavelength m
Wavenumber rrT 1
Κ W π Γ s r Hz" 2 1 1
Li χ W irT sr" m 2 1
Irradiance Win" 2
Spectral irradiance Ελ
W rrT irT 2 1
Ev
W m Hz" 2 1
Εκ
Wmf m 2
Μ„ W irT H z " 2 1
Μ κ
WirT m 2
Radiant energy J
Radiant exposure Jm" 2
Radiant flux W
Radiant flux density Wm" 2
Bidirectional reflectance sr - 1
The basic, supplemental, and named derived units may be prefixed to indicate
order of magnitude as follows
Factor by which
Unit is multiplied Prefix Symbol
10 1 8
exa Ε
10 1 5
peta Ρ
10 1 2
tera Τ
10 9
giga G
10 6
mega Μ
10 3
kilo k
10 2
hecto h
10 1
deka da
ΚΓ 1
deci d
1(T 2
centi c
10~ 3
milli m
1(T 6
micro μ
10- 9
nano η
KT 1 2
pico Ρ
KT 1 5
femto f
-18
1 0
atto a
Several non-SI units are in use in satellite meteorology. Below are conversion
factors for some of them.
(continues)
454 Appendix D
langley joule/meter2
4.184 x 1 0 4
exact
liter meter 3
1 x 10~ 3
exact
micron meter 1 x 10" 6
exact
mile (nautical) meter 1.852 x 1 0 3
exact
mile (statute) meter 1.609344 x 1 0 3
exact
millibar pascal 1 x 10 2
exact
millimeter of mercury pascal 1.333224 x 1 0 2
Bibliography
Mechtly, E. A. (1969). The International System of Units: Physical Constants and Conversion Factors,
Revised. N A S A SP-7012, Washington, D C .
Standard Practice for the Use of the International System of Units (the Modernized Metric System)
(1993). American Society for Testing and Materials, ASTM E380-93, Philadelphia, PA, 35 pp.
Wallace, J. M., and P. V. Hobbs (1977). Atmospheric Science: An Introductory Survey. Academic
Press, N e w York.
Wertz, J. R., and W. J. Larson (ed.) (1991). Space Mission Analysis and Design. Kluwer Academic
Publishers, Dordrecht.
Ε
APPENDIX
Physical
Constants
Relative
Uncertainty a
Universal Constants
Speed of light in vacuum* 7
c 2 . 9 9 7 9 2 4 5 8 x 1 0 m s" 8 1
exact
Newtonian gravitation constant* 7
G 6 . 6 7 2 5 9 x ΙΟ" Ν m kg"
11 2 2
128
Planck's constant*7
h 6.6260755 x ΙΟ" J s 34
0.60
Boltzmann's constant* 7
k 1.380658 x 1 0 " J K" 23 1
8.5
First radiation constant* ' 7 0
C\ 1.1910439 x 1 0 " W m sr" 16 2 1
0.60
Second radiation constant* 7
Cl 1.438769 x 1 0 ~ m Κ 2
8.4
Stefan-Boltzmann constant* 7
σ 5 . 6 7 0 5 1 x 1 0 ~ W n T K~
8 2 4
34
Wein's displacement law constant* 7
2.897756 χ 10" m Κ 3
8.4
Avogadro's number* 7
N A
6 . 0 2 2 1 3 6 7 x 1 0 mol" 2 3 1
0.59
Molar gas constant* 7
R* 8 . 3 1 4 5 1 0 J mol" K" 1 1
8.4
Terrestrial Parameters
Standard atmosphere* 7
atm 1.01325 x 1 0 Pa 5
exact
Mean molecular weight of dry air^ 2 . 8 9 6 6 x 1 0 " kg m o l "
2 1
(continues)
455
Appendix Ε
Relative
Uncertainty"
Constant Symbol Value/Units (ppm)
Solar Parameters
Solar constant 5
^sun 1.368 x 1 0 W m - 3 2
2600
Radius of sun (visible disk)^ r 6.9595 x 1 0 m 8
1160
' sun
Mean Earth-sun distance^ ^sun 1.4956 x 1 0 m 1 1
470
a
One standard deviation uncertainty in parts per million.
b
From Cohen, E. R., and Β. N . Taylor (1986). The 1 9 8 6 Adjustment of the Fundamental Physical
Constants. CODATA Bulletin No. 63, Pergamon Press, Elmsford, N Y , and Oxford, U.K.
c
We have divided the value given by Cohen and Taylor (1986) by π so that the Planck function
yields radiance rather than radiant exitance.
d
Weast, R. C. (ed.) (1980). CRC Handbook of Chemistry and Physics, 61st ed. CRC Press, Inc.,
Boca Raton, FL.
* Defense Mapping Agency (1987). Supplement to Department of Defense World Geodetic System
1984 Technical Report: Part I—Methods, Techniques, and Data Used in WGS 84 Development.
D M A Tech. Rep. 8350.2-A, Building 5 6 , U.S. Naval Observatory, Washington, D C .
f Derived from the product Gm . t
8
See Section 10.1 of text. The uncertainty is estimated as half of the range in the mean satellite-
based measurements.
Index
457
458 Index