zotero-db/storage/C2ESSX4D/.zotero-ft-cache

STUDIES IN GEOPHYSICS

N NAS R NAE C IOM

The Earths Electrical Environment

REFERENCE COPY FOR LIBRARY USE ONLY
STUDIES IN GEOPHYSICS
The Earth's Electrical Environment

Geophysics Study Committee Geophysics Research Forum Commission on Physical Sciences,
Mathematics, and Resources National Research Council
NATIONAL ACADEMY PRESS Washington, D.C. 1986

NAS-NAc JUL 1 6 1986
LIBRARY

<2C Cq-z
C. /

NATIONAL ACADEMY PRESS 2101 Constitution Avenue, N.W. Washington, DC 20418
NOTICE: The project that is the subject of this report was approved by the Governing Board of the Na tional Research Council, whose members are drawn from the Councils of the National Academy of Sciences, the National Academy of Engineering, and the Institute of Medicine. The members of the committee respon sible for this report were chosen for their special competences and with regard for appropriate balance.
This report has been reviewed by a group other than the authors according to procedures approved by a Report Review Committee consisting of members of the National Academy of Sciences, the National Acad emy of Engineering, and the Institute of Medicine.
The National Research Council was established by the National Academy of Sciences in 1916 to associate the broad community of science and technology with the Academy's purposes of furthering knowledge and of advising the federal government. The Council operates in accordance with general policies determined by the Academy under the authority of its congressional charter of 1863, which establishes the Academy as a private, nonprofit, self-governing membership corporation. The Council has become the principal operating agency of both the National Academy of Sciences and the National Academy of Engineering in the conduct of their services to the government, the public, and the scientific and engineering communities. It is administered jointly by both Academies and the Institute of Medicine. The National Academy of Engineering and the Institute of Medicine were established in 1964 and 1970, respectively, under the charter of the National Academy of Sciences.
The Geophysics Study Committee is pleased to acknowledge the support of the National Science Founda tion (Grant EAR-8216205), the Defense Advanced Research Projects Agency, the National Aeronautics and Space Administration, the National Oceanic and Atmospheric Administration, the U.S. Geological Survey (Grant 14-08-001-G11 1 1), and the Department of Energy (Grant DE-FGO2-82ER12018) for the conduct of thisstudv.

Library of Congress Cataloging-in-Publication Data
The Earth's electrical environment.
(Studies in geophysics) Based on papers presented at the American Geophysical Union meetings in June 1983, Baltimore, MD. Includes bibliographies and index. 1 . Atmospheric electricity — Environmental aspects— Congresses. 2. Man— Influence of environment — Congresses. I. National Research Council (U.S.). Geophysics Study Committee. II. American Geophysical Union. III. Series. QC960.5.E27 1986 551. 5'6 86-8782 ISBN 0-309-03680-1

Printed in the United States of America

O.'def tfOfTl National Technical Information Service, Springfield, Va.

-2161

, . „a

[NAS1
NAE .NationaC Academy Tress
IOM
The National Academy Press was created by the National Academy of Sciences to publish the reports issued by the Academy and by the National Academy of Engineering, the Institute of Medicine, and the National Research Council, all operating under the charter granted to the National Academy of Sciences by the Congress of the United States.

Panel on the Earth's
Electrical Environment
E. PHILIP KRIDER, University of Arizona, Co-chairman RAYMOND G. ROBLE, National Center for Atmospheric Research, Co-chairman R. V. ANDERSON, Naval Research Laboratory KENNETH v. K. BEARD, University of Illinois at Urbana-Champaign WILLIAM L. CHAMEIDES, Georgia Institute of Technology ARTHUR A. FEW, JR., Rice University GIOVANNI P. GREGORI, Istituto di Fisica dell'Atmosfera, Rome WOLFGANG GRINGEL, Universitat Tubingen david j . HOFMANN, University of Wyoming WILLIAM A. HOPPEL, Naval Research Laboratory EDWIN KESSLER, NOAA Severe Storms Laboratory PAUL R. KREHBIEL, New Mexico Institute of Mining and Technology LOUIS J. LANZEROTTI, AT&T Bell Laboratories ZEV LEVIN, Tel Aviv University HARRY T. OCHS, Illinois State Water Survey RICHARD E. ORVILLE, State University of New York at Albany GEORGE C. REID, NOAA Aeronomy Laboratory ARTHUR D. RICHMOND, National Center for Atmospheric Research JAMES M. ROSEN, University of Wyoming W. DAVID RUST, NOAA Severe Storms Laboratory ISRAEL TZUR, National Center for Atmospheric Research MARTIN A. UMAN, University of Florida JOHN C. WILLETT, Naval Research Laboratory
Staff THOMAS M. USSELMAN
iii

Geophysics Study Committee
ARTHUR E. MAXWELL, The University of Texas at Austin, Chairman t ALLEN F. AGNEW, Geological Consultant, Corvallis, Oregon t RICHARD A. ANTHES, National Center for Atmospheric Research TD. JAMES BAKERrJoint Oceanographic Institutions, Inc. •COLIN BULL, Mercer Island, Washington GORDON P. EATON, Texas A&M University DEVRIE S. INTRILIGATOR, Carmel Research Center •NICHOLAS C. MATALAS, U.S. Geological Survey, Reston J. MURRAY MITCHELL, National Oceanic and Atmospheric Administration •V. RAMA MURTHY, University of Minnesota t RICHARD J. O CONNELL, Harvard University TMARTIN WALT, Lockheed Missiles and Space Company, Inc. FERRIS WEBSTER, University of Delaware Liaison Representatives RALPH ALEWINE, Defense Advanced Research Projects Agency BRUCE B. HANSHAW, U.S. Geological Survey, Reston GEORGE A. KOLSTAD, Department of Energy MICHAEL MAYHEW, National Science Foundation NED OSTENSO, National Oceanic and Atmospheric Administration SHELBY TILFORD, National Aeronautics and Space Administration Staff THOMAS M. USSELMAN
•Terms ended June 30, 1985. TTerms began July 1, 1985.
iv

Geophysics Research Forum
DON L. ANDERSON, California Institute of Technology, Chairman STANLEY I. AUERBACH, Oak Ridge National Laboratory JOHN J. BOLAND, The Johns Hopkins University THOMAS M. DONAHUE, University of Michigan CHARLES L. DRAKE, Dartmouth College PETER S. EAGLESON, Massachusetts Institute of Technology W. GARY ERNST, University of California, Los Angeles JOHN D. HAUN, Evergreen, Colorado WILLIAM w. HAY, University of Colorado CHARLES L. HOSLER, The Pennsylvania State University DEVRIE S. INTRILIGATOR, Carmel Research Center KEITH A. KVENVOLDEN, U.S. Geological Survey, Menlo Park C. GORDON LITTLE, National Oceanic and Atmospheric Administration CHARLES J. MANKIN, Oklahoma Geological Survey ARTHUR E. MAXWELL, The University of Texas at Austin FRANK B. McDONALD, National Aeronautics and Space Administration WALTER H. MUNK, University of California, San Diego JACK E. OLIVER, Cornell University EUGENE N. PARKER, The University of Chicago FRANK L. PARKER, Vanderbilt University HOWARD J. PINCUS, University of Wisconsin— Milwaukee PAUL w. POMEROY, Rondout Associates, Inc. RICHARD H. RAPP, The Ohio State University ROGER R. REVELLE. University of California, San Diego VERNER E. SUOMI, University of Wisconsin— Madison FERRIS WEBSTER, University of Delaware GUNTER E. WELLER, University of Alaska

Ex Officio JOHN D. BOSSLER, National Geodetic Survey ROBERT K. CRANE, Dartmouth College FRANK D. DRAKE, University of California, Santa Cruz ROBERT HOFSTADTER, Stanford University Staff PEMBROKE J. HART
vi

Commission on Physical Sciences. Mathematics, and Resources
HERBERT FRIEDMAN, National Research Council, Chairman CLARENCE R. ALLEN, California Institute of Technology THOMAS D. BARROW, Standard Oil Company, Ohio (Retired) ELKAN R. BLOUT, Harvard Medical School BERNARD F. BURKE, Massachusetts Institute of Technology GEORGE F. CARRIER, Harvard University CHARLES L. DRAKE, Dartmouth College MILDREDS. DRESSELHAUS, Massachusetts Institute of Technology JOSEPH L. FISHER, George Mason University JAMES C. FLETCHER, University of Pittsburgh WILLIAM A. FOWLER, California Institute of Technology GERHART FRIEDLANDER, Brookhaven National Laboratory EDWARD D. GOLDBERG, Scripps Institution of Oceanography MARY L. GOOD, Allied Signal Corporation J. ROSS MacDONALD, University of North Carolina, Chapel Hill THOMAS F. MALONE, Saint Joseph College CHARLES J. MANKIN, Oklahoma Geological Survey PERRY L. McCARTY, Stanford University WILLIAM D. PHILLIPS, Mallinckrodt, Inc. ROBERT E. SIEVERS, University of Colorado JOHN D. SPENGLER, Harvard School of Public Health GEORGE w. WETHERILL, Carnegie Institution of Washington IRVING WLADAWSKY-BERGER, IBM Corporation
RAPHAEL G. KASPER, Executive Director LAWRENCE E. McCRAY, Associate Executive Director
VII

Studies in Geophysics *
ENERGY AND CLIMATE Roger R. Revelle, panel chairman, 1977, 158 pp.
CLIMATE, CLIMATIC CHANGE, AND WATER SUPPLY James R. Wallis, panel chairman, 1977, 132 pp.
ESTUARIES, GEOPHYSICS, AND THE ENVIRONMENT Charles B. Officer, panel chairman, 1977, 127 pp.
THE UPPER ATMOSPHERE AND MAGNETOSPHERE Francis S. Johnson, panel chairman, 1977, 169 pp.
GEOPHYSICAL PREDICTIONS Helmut E. Landsberg, panel chairman, 1978, 215 pp.
IMPACT OF TECHNOLOGY ON GEOPHYSICS Homer E. Newell, panel chairman, 1979, 121 pp.
CONTINENTAL TECTONICS B. Clark Burchfiel, Jack E. Oliver, and Leon T. Silver, panel co-chairmen, 1980, 197 pp.
MINERAL RESOURCES: GENETIC UNDERSTANDING FOR PRACTICAL APPLICATIONS PaulB. Barton, Jr., panel chairman, 1981, 118 pp.
SCIENTIFIC BASIS OF WATER-RESOURCE MANAGEMENT Myron B. Fiering, panel chairman, 1982, 127 pp.
SOLAR VARIABILITY, WEATHER, AND CLIMATE John A. Eddy, panel chairman, 1982, 106 pp.
CLIMATE IN EARTH HISTORY Wolfgang H. Berger and John C. Crowell, panel co-chairmen, 1982, 197 pp.
•Published to date.
Vlll

FUNDAMENTAL RESEARCH ON ESTUARIES: THE IMPORTANCE OF AN INTERDISCIPLINARY APPROACH Charles B. Officer and L. Eugene Cronin, panel co-chairmen, 1983, 79 pp.
EXPLOSIVE VOLCANISM: INCEPTION, EVOLUTION, AND HAZARDS Francis R. Boyd, Jr., panel chairman, 1984, 176 pp.
GROUNDWATER CONTAMINATION John D. Bredehoeft, panel chairman, 1984, 179 pp.
ACTIVE TECTONICS Robert E. Wallace, panel chairman, 1986, 266 pp.
THE EARTH'S ELECTRICAL ENVIRONMENT E. Philip Krider and Raymond G. Roble, panel co-chairmen, 1986, 263 pp.
IX

Preface
This study is part of a series of Studies in Geophysics that have been undertaken for the Geophysics Research Forum by the Geophysics Study Committee. One purpose of each study is to provide assessments from the scientific community to aid policymakers in decisions on societal problems that involve geophysics. An important part of such assessments is an evaluation of the adequacy of current geophysical knowledge and the appropriateness of current research programs as a source of information required for those decisions.
The Earth's Electrical Environment was initiated by the Geophysics Study Commit tee and the Geophysics Research Forum with consultation of the liaison representatives of the agencies that support the Geophysics Study Committee, relevant committees and boards within the National Research Council, and members of the scientific commu nity.
How does atmospheric electricity affect man and his technological systems? Is our electrical environment changing as a result of air pollution, the release of radioactive materials, the construction of high-voltage power lines, and other activities? It is clear that modern technological advances can be seriously affected by various atmospheric electrical processes and that man is also beginning to affect the electrical environment in which he resides.
The study reviews the recent advances that have been made in independent research areas, examines the interrelations between them, and projects how new knowledge could be applied for benefits to mankind. The study also indicates needs for new re search and for the types of coordinated efforts that will provide significant new ad vances in basic understanding and in applications over the next few decades. It empha sizes a need to consider the interactions between various atmospheric, ionospheric, and telluric current systems that will be necessary to achieve an overall understanding of global electrical phenomena.
The preliminary scientific findings of the authored chapters were presented at an
xi

xii

PREFACE

American Geophysical Union symposium in Baltimore in June 1983. In completing their chapters, the authors had the benefit of discussion at this symposium as well as the comments of several scientific referees. Ultimate responsibility for the individual chap ters, however, rests with their authors.
The Overview of the study summarizes the highlights of the chapters and formulates conclusions and recommendations. In preparing the Overview, the panel co-chairmen and the Geophysics Study Committee had the benefit of meetings that took place at the symposium and the comments of the panel of authors and other referees. Responsibility for the Overview rests with the Geophysics Study Committee and the co-chairmen of the panel.

Contents

Overview and Recommendations

1

I. LIGHTNING

1. Lightning Phenomenology

23

Richard E. Orville

2. Physics of Lightning

30

E. Philip Krider

3. Positive Cloud-to- Ground Lightning

41

W. David Rust

4. Acoustic Radiations from Thunderstorms

46

Arthur A. Few, Jr.

5. Applications of Advances in Lightning Research to

Lightning Protection

61

Martin A. Uman

6. The Role of Lightning in the Chemistry of the

Atmosphere

70

William L. Chameides

II. CLOUD AND THUNDERSTORM ELECTRICITY

7. Thunderstorm Origins, Morphology, and Dynamics

81

Edwin Kessler

xiii

XIV

CONTENTS

8. The Electrical Structure of Thunderstorms

90

PaulR. Krehbiel

9. Charging Mechanisms in Clouds and Thunderstorms

114

Kenneth V. K. Beard and Harry T. Ochs

10. Models of the Development of the

Electrical Structure of Clouds

131

Zev Levin and Israel Tzur

III. GLOBAL AND REGIONAL ELECTRICAL PROCESSES

11. Atmospheric Electricity in the Planetary Boundary Layer ... 149 William A. Hoppel, R. V. Anderson, and John C. Willett

12. Electrical Structure from 0 to 30 Kilometers

166

Wolfgang Gringel, James M. Rosen, and David J. Hofmann

13. Electrical Structure of the Middle Atmosphere

183

George C. Reid

14. Upper-Atmosphere Electric-Field Sources

195

Arthur D. Richmond

15. The Global Atmospheric-Electrical Circuit

206

Raymond G. Roble and Israel Tzur

16. Telluric Currents: The Natural Environment and

Interactions with Man-made Systems

232

Louis J. Lanzerotti and Giovanni P. Gregori

Index

259

STUDIES IN GEOPHYSICS

Overview and Recommendations
INTRODUCTION How does atmospheric electricity affect man and his technological systems? Is our electrical environment changing as a result of air pollution, the release of radioactive materials, the construction of high-voltage power lines, and other activities? It is clear that modern technological advances can be seriously affected by various atmospheric electrical processes and that man is also beginning to affect the electrical environment in which he resides. Our need to assess these technological and environmental impacts requires a better understanding of electrical processes in the Earth's atmosphere than we now possess. Further research is needed to understand better the natural electrical environment and its variability and to predict its future evolution. We live in an environment that is permanently electrified. Certainly, the most spec tacular display of this state occurs during intense electrical storms. Lightning strikes the Earth 50 to 100 times each second and causes the death of hundreds of people each year. Lightning is also a major cause of electric power outages, forest fires, and damage to communications and computer equipment; and new sophisticated aircraft are becom ing increasingly vulnerable to possible lightning damage. Lightning contributes to the production of fixed nitrogen in the atmosphere, a gas that is essential for the growth of plants, and other trace gases. It is well known that the intense electric fields that are produced by thunderstorms can cause a person's hair to stand on end and produce corona discharges from antennas, trees, bushes, grasses, and sharp objects; these fields may also affect the development of precipitation in thunderstorms. Even in fair weather, there is an electric field of several hundred volts per meter near the ground that is maintained by worldwide thunderstorm activity. In the Earth's upper atmosphere near 100-km altitude, a current of a million am peres flows in the high-latitude auroral zones; changes in the upper atmosphere cur rents, through electromagnetic induction, cause telluric currents to flow within power
1

OVERVIEW AND RECOMMENDATIONS
and communication lines as well as within the Earth and oceans. The upper atmo spheric current systems are highly variable and are strongly related to solar-terrestrial disturbances. Power failures and communication disruptions have occurred during in tense geomagnetic storms. It also appears that the electromagnetic transients that are produced by lightning and man-made power systems can affect trapped particle popu lations in the magnetosphere and cause particle precipitation into the upper atmo sphere at low geomagnetic latitudes.
The practical needs for understanding many of the basic questions about atmo spheric electricity were brought into clear focus on November 14, 1969. Thirty-six seconds after lift-off from the NASA Kennedy Space Center, Apollo 12 was struck by lightning, and 16 seconds later it was struck again. The first discharge disconnected all the fuel cells from the spacecraft power busses, and the second caused the inertial platform in the spacecraft guidance system to tumble. Fortunately, the rocket was still under control of the Saturn V guidance system at the time of the strikes; and, as a result, the astronauts, who had never practiced for such a massive electrical disturbance, were able to reset their circuit breakers, reach Earth orbit, realign their inertial platform, and ultimately land on target on the Moon. Although permanent damage to Apollo 12 was minimal, the potential for disaster of this lightning incident called attention to the important unanswered questions regarding lightning and atmospheric electricity.
Research in atmospheric electricity traditionally has been divided into several broad areas: (1) ion physics and chemistry, (2) cloud electrification, (3) lightning, (4) fairweather electrical processes, (5) ionospheric and magnetospheric current systems, and (6) telluric current systems. Most of this research has been pursued independently by scientists and engineers in different disciplines such as meteorology, physics, chemistry, and electrical engineering.
This study reviews the recent advances that have been made in these independent research areas, examines the interrelations between them, and projects how new knowledge could be applied for benefits to mankind. The study also indicates needs for new research and for the types of coordinated efforts that will provide significant new advances in basic understanding and in applications over the next few decades. It em phasizes a need to consider the interactions between various atmospheric, ionospheric, and telluric current systems that will be necessary to achieve an overall understanding of global electrical phenomena.
LIGHTNING
Lightning is a large electric discharge that occurs in the atmosphere of the Earth and other planets and can have a total length of tens of kilometers or more. The continental United States receives about 40 million cloud-to-ground (CG) lightning strikes each year; on average, there are probably 50 to 100 discharges each second throughout the world (Chapter 1). Most lightning is produced by thunderclouds, and well over half of all discharges remain within the clouds. Most of our knowledge about the physics of lightning has come from the study of CG discharges. Most CG flashes effectively lower negative charge to ground, however recent evidence shows that positive charge can also be lowered (see Chapter 3 on positive lightning) . Cloud-to-ground lightning kills about a hundred people and causes hundreds of millions of dollars in property damage each year in the United States; it is clearly among the nation's most severe weather hazards.
Most CG discharges begin within the cloud where there are large concentrations of positive and negative space charge (see Chapter 8). After several tens of milliseconds, the preliminary cloud breakdown initiates an intermittent, highly branched discharge that propagates horizontally and downward and that is called the stepped-leader (Fig ure 1). When the tip of any branch of the stepped-leader gets close to the ground, the large electric field that is produced near the surface causes one or more upward propa

OVERVIEW AND RECOMMENDATIONS

3

/////////////////

FIGURE 1 The top figure shows the stepped-leader channel just before attachment; the bottom figure shows attachment and the development of the return stroke. The estimated time between the two figures is on the order of 0.001 sec.
gating discharges to form. When an upward discharge makes contact with the steppedleader, the first return stroke begins. The return stroke is an intense wave of ionization that starts at or just above the ground and that propagates up the leader channel at about one third the speed of light. The return stroke is typically the brightest phase of lightning. The peak currents in these return strokes can reach several hundred thou sand amperes; a typical value is about 40,000 A. The peak electric power that is dissi pated by a return stroke is on the order of 100 million watts per meter of channel: and the peak channel temperatures approach 30,000 K. A shock wave is produced by the rapid expansion of the hot, high-pressure channel, and this eventually becomes thun der with its own characteristics that depend on the nature of the discharge and the atmospheric environment (see Chapter 4) .
The currents in the return stroke carry the ground potential upward and effectively neutralize most of the leader channel. After a pause of 40 to 80 milliseconds, most CG flashes produce a new leader, the dart leader, that propagates down the previous re turn-stroke channel and initiates a subsequent return stroke. Most flashes contain two to four return strokes, with each affecting a different volume of cloud charge (see Chap

OVERVIEW AND RECOMMENDATIONS
ter 8). If a dart leader forges a different path to ground than the previous stroke, then the lightning will actually strike the ground in more than one place and will have a forked appearance. Chapters 2 and 3 discuss the physics of lightning in greater detail.
Among the more important advances that have been made in recent years has been the discovery that both in-cloud and CG discharges produce very fast-rising currents, i.e. , rise times of tens to hundreds of nanoseconds and rates of change of current (d//df) on the order of 10" A/sec. Most of the standard waveforms that are used to test the performance of lightning protectors and the integrity of lightning protection systems currently have current rise times and d//dts that are substantially slower than the above values; therefore, these standards may not be an adequate simulation of the true light ning threat to aircraft and other structures (see Chapter 5) .
A variety of nonequilibrium trace gases are produced within high-temperature light ning channels and by the shock wave that can affect tropospheric and stratospheric chemistry (see Chapter 6) .
Recent spacecraft observations have shown that lightning may be present in the atmospheres of Jupiter, Venus, and Saturn; the upcoming Galileo probe will carry a lightning detector to Jupiter. In the future a study of lightning in atmospheres that are radically different from the Earth's may lead to a better understanding of the forma tion and characteristics of lightning on Earth.
CLOUD ELECTRIFICATION
Although the vast majority of terrestrial clouds form and dissipate without ever pro ducing precipitation or lightning, they can be weakly electrified. In some clouds, the electrification intensifies as convective activity increases, and strong electrification usually begins when there is rapid vertical and horizontal growth of the cloud and the development of precipitation. Most lightning on Earth is produced by cumulonimbus clouds that are strongly convective (i.e. , they contain a vigorous system of updrafts and downdrafts) and that contain both supercooled water and ice. A small fraction of warm clouds are also reported to produce lightning.
The updrafts and downdrafts and the interactions between cloud and precipitation particles act in some still undetermined manner to separate positive and negative charges within the cloud. These processes usually transfer an excess of positive charge to the upper portion of the cloud and leave the lower portion with a net negative charge. Recent research has shown that the negative charge is usually concentrated at altitudes where the atmospheric temperature is between - 10°C and - 20°C (i.e., 6 to 8 km above sea level in summer thunderstorms and 1 to 3 km in winter storms) and that this altitude remains constant as the storm develops. This finding sets important criteria that must be met by any proposed thunderstorm charging mechanisms. The positive charge that is above the negative may be spread through deeper layers and does not exhibit as clear a relationship with temperature as does the negative charge. Positive charges are found at levels between - 25°C and - 60°C depending on the size of the storm, and this temperature range usually lies between 8 and 16 km above sea level.
Cloud electrification processes can be viewed as acting over two spatial scales: a microscale separation that ultimately leads to charged ice and water particles and then a larger-scale separation that produces large volumes of net positive and negative charge and eventually lightning. The microscale separation includes the creation of ion pairs, ion attachment, and charge that may be separated by collisions between individ ual cloud and precipitation particles. The larger cloud-scale separation may be due to precipitation or large-scale convection or some combination of the two.
Numerous mechanisms have been proposed for the electrification of clouds and thunderstorms, and several of these might be acting simultaneously. Feedback can occur through changes in the ion concentrations and electric field, and thus it is diffi

OVERVIEW AND RECOMMENDATIONS
cult to identify or evaluate the primary causes of electrification in a cloudy environ ment. Currently, there is a great need for more measurements to determine the loca tions, magnitudes, and movements of space charges within and near the cloud boundary. There is also a need to determine the charge-size relationship that is present on both cloud and precipitation particles, how these charges evolve as a function of time, and how these distributions are affected by lightning. Laboratory experiments have provided valuable information about the physics of selected microscale processes and are expected to continue to provide important data on the relative magnitude of various processes. Theory and numerical models also have played an important role in simulating and evaluating possible charging mechanisms on both the microscale and the cloud scale.
During the early nonprecipitating cloud stage, charging can occur by diffusion, drift, and selective capture of ions. Later, during the rain stage, there can be additional electrification due to drop breakup and other mechanisms based on electrostatic induc tion. Drift, selective ion capture, breakup, and induction are probably responsible for the charges and fields that are found in stratiform clouds; however, it is difficult to explain with just these mechanisms the stronger electrification that is found in convective clouds more than a few kilometers deep. For clouds in the hail stage, thermoelec tric and interface charging mechanisms can provide strong electrification on the microscale.
In thunderclouds, the charges that are generated on a microscale can be subse quently separated on a larger cloud scale by convection and/or gravitational settling. Particles near the boundary of the cloud will become electrified by ion attachment, and the convection of these charges may play an important role in the electrification. Con vection also plays a role in the formation and growth of cloud particles by forcing the condensation of water vapor until the particles are large enough to coalesce. Interac tions between cloud particles, particularly when there are rebounding collisions, may also produce charge separation. If the larger particles tend to carry charge of predomi nantly one sign, they will fall faster and farther with respect to the convected air and leave the oppositely charged, smaller particles at higher altitudes.
As the populations of charged particles increase, the mechanisms that discharge these particles become more effective. Two kinds of discharging are possible: (1) dis charge by ionic conduction, point discharge, or lightning and (2) discharge by collision and/or coalescence with cloud particles of opposite polarity. The attachment of ions to cloud particles will be a function of the particle charge and the electric field of the cloud, and strong fields may also produce corona discharges from large water drops and the corners of ice crystals. Corona ions and lightning will increase the local electri cal conductivity, and this, in turn, may prevent or reduce any further buildup of space charge in this region of the cloud.
Collisional discharge will take place at all stages of cloud particle growth. These mechanisms are enhanced if the interacting particles are highly charged and of oppo site polarity; therefore, if a charging mechanism is to be effective, it must separate charge at a rate that is sufficiently high to overcome the discharging processes. It is worth noting that the electric forces on charged elements of precipitation can be several times larger than gravity; therefore, the terminal velocities and frequency of collisions of these particles will be a function of the electric field. More detailed discussions of the various processes involved in cloud electrification are given in Chapters 8, 9, and 10.
Recently, there have been attempts to analyze the patterns of the Maxwell current density that thunderclouds produce at the ground in order to define better the charac teristics of the cloud as an electrical generator. The Maxwell, or total current density, contains components due to ohmic and non-ohmic (corona) ion conduction, convec tion, precipitation, displacement, and the charge-separating currents within the cloud. Under some conditions, there is evidence that the total current density may be

OVERVIEW AND RECOMMENDATIONS
coupled directly to the meteorological structure of the storm and/or the storm dy namics, but the lack of simultaneous Maxwell current measurements both on the ground and aloft has not allowed the details of this relationship to be determined. Such measurements will also be complicated by the complexity of the meteorological and electrical environment outside the storm.
ELECTRICAL STRUCTURE OF THE ATMOSPHERE It has been known for over two centuries that the solid and liquid Earth and its atmosphere are almost permanently electrified. The surface has a net negative charge, and there is an equal and opposite positive charge distributed throughout the atmo sphere above the surface. The fair-weather electric field is typically 100 to 300 V/m at the surface; there are diurnal, seasonal, and other time variations in this field that are caused by many factors. The atmosphere has a finite conductivity that increases with altitude; this conductivity is maintained primarily by galactic cosmic-ray ionization. Near the Earth's surface, the conductivity is large enough to dissipate any field in just 5 to 40 minutes (depending on the amount of pollution); therefore, the local electric field must be maintained by some almost continuous current source. Ever since the 1920s, thunderstorms have been identified to be the dominant genera tor in the global circuit. Most cloud-to-ground lightning transfers negative charge to the ground, and the point discharge currents under a storm transfer positive space charge to the atmosphere. In addition, there are precipitation and other forms of con vection currents and both linear and nonlinear conduction currents that must be con sidered when attempting to understand the charge transfer to the earth by a thunder storm. The electrical structure of a thunderstorm is complex (see Chapter 8), but it is often approximated simply as a vertical electric dipole. The conductivity of the fair-weather atmosphere near the surface is on the order of 10" H mho/m, and it increases nearly exponentially with altitude to 60 km with a scale
MAGNETOSPHERE
GLOBAL ELECTRICAL CIRCUIT FIGURE 2 Schematic of various electrical processes in the global electrical circuit.

OVERVIEW AND RECOMMENDATIONS

ELECTRON DENSITY (rrf3)

1000 F

I010 I012 1 ft I I I I I I I II I I II I I Magnetosphere

o
UJ X

Planetary Boundary Layer

300

1000 2000

l'■''■''l■ '''■

TEMPERATURE (K) ELECTRICAL CONDUCTIVITY (sm'1)

FIGURE 3 Nomenclature of atmospheric regions based on profiles of electrical conductivity (a), neutral temperature, and electron number density.

height of about 6 km. The main charge carriers below about 60 km are small positive and negative ions that are produced primarily by galactic cosmic rays. Above 60 km, free electrons become more important as charge carriers and their high mobility pro duce an abrupt increase in conductivity throughout the mesosphere. Above 80 km, the conductivity becomes anisotropic because of the influence of the geomagnetic field, and there are diurnal variations due to solar photoionization processes.
The atmospheric region above about 60 km is known as the equalization layer and is usually assumed to be an isopotential surface and the upper conducting boundary of the global circuit. Currents flow upward from the tops of thunderstorms to this layer where they are rapidly distributed throughout the world. Worldwide thunderstorms maintain a potential difference of 200 to 600 kV between the equalization layer and the surface—the Earth-ionosphere potential. This potential difference, in turn, drives a downward conduction current that is on the order of 2 x 10 " 12 A/m2 in fair-weather regions and constant with altitude.
Today, there are still many details that need to be clarified about the role of thunder storms as the generators in the global circuit (Figure 2) . Upward currents have been detected above thunderclouds, but how these currents depend on storm dynamics, stage of development, lightning frequency, precipitation intensity, and cloud height, for example, is still not known. There is a need for further measurements to quantify the relationships between diurnal variations of the ionospheric potential, the electric field or air-earth current, and worldwide estimates of thunderstorm frequency.
Many electrical processes interact within the global circuit, and the following sub sections will describe selected processes that occur within certain atmospheric regions (Figure 3) . It should be recognized that the global circuit includes mutual electrical interactions between all atmospheric regions.

OVERVIEW AND RECOMMENDATIONS
This report will not present an encyclopedic review of all the electrical phenomena that occur in the atmosphere but will simply give some examples that illustrate a few of the basic processes and some of the important interrelationships. Notably absent, but important to an overall understanding of atmospheric electricity, are discussions deal ing with electromagnetic phenomena such as sferics, Schumann resonances, and whis tlers. These and many other subjects, however, have been reviewed in the two- volume Handbook of Atmospherics, edited by Volland (1982), and recent proceedings of inter national conferences on atmospheric electricity (Dolezalek and Reiter, 1977; Orville, 1985). Human and biological effects of atmospheric electricity are also important re search areas that are not considered in this study.
The Planetary Boundary Layer
The planetary boundary layer (PBL) is the lowest few kilometers of the atmosphere where interactions with the surface, man, and the biosphere are the most pronounced. Galactic cosmic rays are the main source of ionization in the PBL; however, near land surfaces, ionization is also produced by decays of natural radioactive gases emanating from the soil surface and by radiations emitted directly from the surface. Ionization from radioactive sources depends on soil type and surface structure and on the meteoro logical dispersal rate; this ionization normally decreases rapidly with altitude, and at about 1 km its contribution to the total ionization is less than that from cosmic rays. Other sources of ions in the PBL include lightning; electrification due to waterfalls, ocean surf, and man-made sprays; a variety of combustion processes; point discharge or corona currents that are produced whenever the ambient electric field exceeds break down; and frictional processes associated with blowing dust, snow, or volcanic ejecta.
In the troposphere, atmospheric trace gases are numerous and variable, and the ion chemistry is complicated by clustering processes and the relatively long lifetime of the terminal ions. As a further complication, clouds and other aerosols play an important role as sinks for small ions and thereby alter the ion distribution (see Chapter 11). Over continental areas, the loss of ions by attachment to aerosols can be larger than the loss by recombination. Some atmospheric aerosols are hygroscopic, and the particle size increases with relative humidity. At large humidities, fog and cloud droplets form and produce a large decrease in the electrical conductivity of the atmosphere. Since a de crease in conductivity can be a precursor of fog, it might be possible to improve fore casts of the onset of fog by electrical measurements.
Turbulent transport and convection within the PBL are important processes that govern the momentum, heat, and moisture exchanges between the atmosphere, geosphere, and biosphere. These processes influence the mean wind profile, the vertical distribution of temperature, water vapor, trace gases, aerosols, and the ion distribution throughout the troposphere. Turbulent mixing and convection can prevent the buildup of radioactive emanations near the ground and can also disperse aerosols to a greater altitude in the troposphere.
Electrical processes in the PBL are complex, highly variable, and span a tremendous range of space and time scales. The electrical variables respond to many of the lower atmospheric processes but usually have little influence on the phenomena to which they respond. Within the PBL, local turbulent fluctuations of space-charge density impose a time-varying electric field that is comparable in magnitude with or even greater than the electric field maintained by global thunderstorm activity. Since the PBL is the region of the atmosphere with the greatest resistance, it is this layer as well as the generators that control the currents in the global circuit. Electrical processes in the PBL are discussed in greater detail in Chapter 1 1 .

OVERVIEW AND RECOMMENDATIONS
Mid- Troposphere-Stratosphere
The main source of ionization in the mid-troposphere-stratosphere ( ~ 2-50-km) re gion is cosmic radiation; the ionization rate depends on magnetic latitude and on solar activity. At about 50° geomagnetic latitude and at 20-km altitude, the ion-production rate during sunspot maximum is about 30 percent smaller than during sunspot mini mum; at 30-km altitude (same latitude) it is about 50 percent smaller. Following solar flares that produce energetic charged particles (solar proton events), the ion-produc tion rate in the stratosphere may increase by orders of magnitude for periods of hours to days, but deeper in the atmosphere the effect is much smaller. Solar flares are usually followed by a reduction (Forbush decreases) in the ion-production rate for periods of hours to weeks that is caused by a temporary reduction of the incoming cosmic-ray flux.
The composition and chemistry of the ions that establish the bulk electrical proper ties in the mid-troposphere-stratosphere are relatively unknown. The ion concentra tions are also affected by aerosols whose distributions are quite variable in both space and time. Aerosols tend to accumulate at temperature inversion boundaries and can cause a general loss of visibility that can be seen by airline passengers as they pass through such layers. Such a buildup of aerosols causes a general decrease in the smallion concentration, and, thus, the electrical conductivity is also reduced— resulting in an increase in the local electric field.
The concentration of particles with a radius greater than about 0. 1 micrometer de creases with altitude above the PBL, and a relative minimum occurs in the upper tro posphere. The particle concentrations increase within the lower stratosphere, peak near 20 km, and then decrease again with altitude. This persistent structure is fre quently referred to as the 20-km sulfate layer; the character of this layer is controlled largely by gases emitted during volcanic eruptions, as discussed in Chapter 12.
Aerosol particles that have radii on the order of 0.01 micrometer are referred to as condensation nuclei (CN) and are uniformly mixed throughout the troposphere above the PBL. Near the surface, the CN concentration may be large owing to local sources; above the tropopause the concentration decreases with altitude. In recent years, a CN layer has frequently been observed near 30 km. As a result of the El Chichon volcanic eruption in 1982, the normal CN concentration at 30-km altitude increased by at least two orders of magnitude and measurably affected the ion concentration and electrical conductivity.
Under steady-state conditions, the air-earth current density is constant with altitude if there is large-scale horizontal homogeneity and if no thunderstorms or other localized electrical disturbances are in the vicinity. The air-earth current varies with magnetic latitude because of the magnetic variations in cosmic-ray fluxes. The current is gener ally enhanced over orographic features such as mountain ranges because of the de creased columnar resistance (mountains are closer to the ionosphere than the near-sealevel surface) . Estimates have been made that indicate that as much as 20 to 30 percent of the total global current flows into the high mountain peaks.
Mesosphere
Inthemesosphere (50-85 km altitude), the major daytime source of ionization is solar Lyman-alpha photoionization of nitric oxide (NO). The major source of NO for this region is the thermosphere, where NO is produced by extreme ultraviolet (EUV) radia tion (wavelengths less than 100 nm) and auroral particle precipitation. Meteorological processes in the upper atmosphere transport NO from the thermosphere to the meso sphere, where its distribution is variable. Somewhat smaller sources of ionization in the upper mesosphere include solar x-ray ionization and the photoionization of oxygen in a metastable state. At high latitudes, energetic electrons, protons, and bremsstrahlung

10

OVERVIEW AND RECOMMENDATIONS

ultraviolet radiation associated with auroral particle precipitation are variable sources

related to geomagnetic activity.

Solar protons are a sporadic and intense source of ionization at high latitude follow

ing intense solar flares. These solar proton events can increase the electrical conductiv

ity of the magnetic polar-cap mesosphere by several orders of magnitude (at altitudes

down to about 50 km) during such events. In addition, the current carried by the

bombarding solar protons can often exceed the local air-earth conduction current flow

ing in the circuit.

The principal primary positive ions produced in the mesosphere are N \ , 0*2, and

NO + , but all participate in a wide range of reactions that lead to a rich spectrum of

ambient positive ions. An equally rich range of negative ions is generated by reactions

initiated by the attachment of electrons to form the main primary species, O "2 and O " .

Rocketborne mass-spectrometer measurements have shown that below the mesopause

the positive ions are proton hydrates with as many as 20 water molecules clustering to

individual ions. The positive-ion chemistry of the mesosphere is better understood than

is the negative-ion chemistry (see Chapter 13).

The interaction of the terminal ions with aerosol particles is probably a significant

sink for ions in the polar aerosol layers near the summer mesopause where noctilucent

clouds are commonly observed.

The electrical conductivity of the mesosphere is important because it governs the

electrical properties of the equalization layer in the global circuit. Below about 60 km,

the terminal small ions are the main charge carriers; but above 60 km, free electrons

can exist and their high mobility is responsible for the abrupt increase in electrical

conductivity observed in the mesosphere. Furthermore, above 70 km, collisions be

tween electrons and air molecules become infrequent enough so that electrons are con

fined to spiral about a magnetic field line and the motion perpendicular to the field

becomes more difficult than motion along the field. The electrical conductivity be

comes anisotropic, and this anisotropy has a dominant influence on the electrical prop

erties of the global circuit above 70 km .

4^

Rocketborne measurements of the upper atmosphere conductivity and electric field

indicate some puzzling features. There appear to be regions in the upper stratosphere

and mesosphere that have abrupt increases and decreases in vertical conductivity pro

files. The decreases are probably associated with aerosol layers, but the increases are

difficult to interpret. On occasion, the electric field near 50- to 70-km altitude has been

observed to increase enormously from what is expected if the mesosphere is a passive

element in the global circuit. The mesosphere may not be electrically passive but may,

in fact, contain active electrical generators that are not currently known.

Ionosphere and Magnetosphere
The major sources of ionization above about 85 km are extreme-ultraviolet (EUV) radiation and auroral particle precipitation (see Chapter 14). The ionizing portion of the solar spectrum (i.e., wavelengths below 102.7 nm) is absorbed in the thermosphere and creates an ionosphere that consists of positive molecular and atomic ions (e. g. , N \ , NO + , O I , O + ) and negative electrons. The solar EUV radiation and the electron and ion densities throughout the ionosphere are highly dependent on solar activity; there are known variations with the 11-yr sunspot cycle, the 27-day rotation of the Sun, and solar flares.
Auroral particle precipitation is responsible for large variations in ion and electron densities at high latitudes. The bulk of the precipitation occurs within the auroral oval that encircles the geomagnetic pole in magnetic conjugate polar caps. Observations over many years show that there is always auroral activity within the oval. The activity varies considerably over the day and even from hour to hour owing to interactions of

OVER VIEW AND RECOMMENDATIONS

11

the solar-wind plasma with the Earth's magnetic field. The total power dissipated by particles bombarding the upper atmosphere is typically 109 W, but during large geo magnetic storms it can approach 1012 W.
The sources and composition of the ions that maintain the bulk electrical properties of the upper atmosphere are generally known on the dayside of the Earth, but at night there are still uncertainties with regard to the ionization sources.
In the classical view of the global circuit (see Chapter 15), the ionosphere is assumed to be at a uniform potential with respect to the surface; however, the known upperatmosphere generators are not included. The two major generators that operate in the ionosphere above about 100 km are the ionospheric wind dynamo and the solar-wind/ magnetosphere dynamo (see Chapter 14) . Atmospheric winds have the effect of moving the weakly ionized ionospheric plasma through the geomagnetic field. This movement produces an electromotive force and generates electric currents and fields. This process is complicated by the variability of the ionospheric winds and the anisotropic electrical conductivity in the ionosphere.
The magnitude of the horizontal electric field associated with the wind-driven dy namo is on the order of 1 mV/m. A total current of about 100,000 A flows horizontally in the ionosphere because of the combined action of the wind and electric field, mainly on the sunlit side of the Earth. This current flows in two counterrotating vortices on opposite sides of the equator, and these patterns dominate at low latitudes and midlatitudes. Global-scale horizontal potential differences of about 5 to 10 kV are gener ated by the ionospheric wind dynamo.
The ionospheric winds that drive the dynamo are mainly caused by upward propa gating tides from the lower atmosphere that have large day-to-day fluctuations. Dur ing geomagnetic storms, however, thermospheric winds increase in response to highlatitude auroral heating and cause disturbances at low latitudes to the fields and currents of the ionospheric wind dynamo.
The solar-wind/magnetosphere dynamo results from the flow of the solar wind around and perhaps partly into and within the Earth's magnetosphere. The motion of this plasma through the geomagnetic field produces an electromotive force and cur rents at high latitudes that result in an antisunward flow of plasma over the magnetic polar cap and a sunward flow of ions in the vicinity of the dawn and dusk auroral zones. This motion is described by a two-cell counterrotating ion circulation with one cell on the dawn side and the other on the dusk side of the magnetic polar caps. The polar-cap electric field is typically 20 mV/m, with an ionospheric convection velocity of 300 m/ sec. Larger fields of about 50 to 100 mV/m occur in the vicinity of the auroral ovals. The large-scale potential difference that is associated with this horizontal ion flow over the polar caps has a total dawn-to-dusk drop of about 50 kV. This potential drop and the configuration of the two-cell pattern are highly variable. The potential drop has values of 20 to 30 kV during geomagnetic quiet conditions that increase to 100 to 200 kV during geomagnetic storms. These fields are mainly confined to the polar caps because of the shielding from currents within the magnetosphere. During geomagnetic storms, however, the shielding currents can be altered and electric fields have been observed to propagate all the way from the polar caps to the equator.
Currents are an integral part of the complex electrical circuit associated with the solar-wind/magnetosphere dynamo. Currents flowing along the direction of the mag netic field couple the auroral oval and high-latitude ionosphere with outer portions of the magnetosphere. Typically about a million amperes of current flow in the solarwind/magnetosphere dynamo. The dynamo currents and fields with this high-latitude system are extremely complex and highly variable (see Chapter 14) .
The large-scale horizontal fields (scale sizes 100 to 1000 km) within the ionosphere can propagate or map downward in the direction of decreasing electrical conductivity. Horizontal fields of a smaller scale (1 to 10 km) on the other hand are rapidly attenu

12

OVERVIEW AND RECOMMENDATIONS

ated. The larger-scale horizontal electric fields that do map to the surface become vertical at the surface because of the high surface conductivity. The solar-wind/magnetosphere field can alter the surface fields at high latitudes by 20 to 50 V/m depending on the level of geomagnetic activity and the magnitude of the dawn-to-dusk potential drop across the magnetic polar cap.

Telluric Currents
Telluric currents consist of both natural and man-induced electric currents flowing in the solid earth and oceans. The fundamental causes of the natural currents are elec tromagnetic induction resulting from a time-varying, external geomagnetic field or the motion of a conducting body (such as seawater) across the Earth's internal magnetic field. These telluric currents, in turn, produce magnetic fields of their own that add to the external geomagnetic field and that produce a feedback on the ionospheric current system. The complexities associated with telluric currents arise from the complexities in the external current sources and the conductivity structure of the Earth (see Chapter 16).
The external inducing field also has various scale sizes that contribute to the complex ities in the telluric current systems. The ionospheric dynamo currents that are associ ated with the solar diurnal and lunar tides have a planetary-scale size. The ionospheric current variations, however, also have smaller-scale features that are associated with auroral and equatorial electrojets. At low frequencies, the external inducing sources can be approximated by a planetary-scale field that is occasionally altered by strong spatial gradients during geomagnetically disturbed conditions. At higher frequencies (magnetic storms, substorms, or geomagnetic pulsations) , the source can often be quite localized and highly time dependent.
Electromagnetic induction caused by ionospheric and magnetospheric current vari ations has a pronounced effect on telluric currents and on man-made systems. These effects have been detected by a number of investigators, and it is now well recognized that there is a direct electromagnetic coupling from the ionosphere to the telluric cur rents. The large variation in conductivity of the solid earth can give rise to various channeling effects within the Earth, thereby considerably complicating the flow pat terns of the telluric currents. The current patterns are different for different frequen cies of external induction. The longer the period of the time-varying field, the deeper into the Earth the induced currents are expected to flow. For example, a signal with a period of about 24 hours is generally believed to have a skin depth of 600 to 800 km . The distribution of sediments, the degree of hydration, differences in porosity, and other properties of the Earth all have an influence on the signal response. Properly inter preted, telluric currents can be a tool to study both shallow and deep structures within the Earth.

TECHNIQUES FOR EVALUATING THE ELECTRICAL PROCESSES AND STRUCTURE
The cornerstone of our understanding of the Earth's electrical environment is an integration of measurements, theory, and modeling. The new instruments and tech niques that have been developed in recent years are diverse, and various chapters in this volume contain details on techniques beyond those illustrated in this Overview.
Remote measurements of electric and magnetic fields can now be used to infer many properties of lightning and lightning currents. Also, the amplitude and time character istics of thunder and various radio-frequency (rf) noise emissions can be used to trace the geometrical development of lightning channels within clouds as a function of both space and time. There are now large networks of ground-based lightning detectors that

OVERVIEW AND RECOMMENDATIONS

13

can discriminate between in-cloud and cloud-to- ground lightning and accurately de termine the locations of ground-strike points. With such a detection capability, it should now be possible to determine whether and how the characteristics of individual cloud-to- ground discharges depend on their geographic location, the local terrain, and/or the meteorological structure of the storm.
The rf noise that is generated by lightning in the hf and vhf bands appears in the form of discrete bursts, and within a burst there are hundreds to thousands of separate pulses. If the difference in the time of arrival of each pulse is carefully measured at widely separated stations, the location of the source of each pulse can be computed, and the geometrical development of the rf bursts can be mapped as a function of time.
Satellite observations of lightning have provided rough estimates of the global flash ing rates and the geographic distribution of lightning as a function of season. Optical detectors, such as those now in orbit on the Defense Meteorological Satellite Program (DMSP) satellites, are limited in their temporal and spatial coverage, but they have provided data that show a progression of lightning activity toward the summer hemi sphere and notable absences of lightning over the ocean during the observing intervals (see Chapter 1) . The data to date are only for local midnight, dawn, and dusk; there is a need to obtain data at other times. Measurements of hf radio noise by the Ionosphere Sounding Satellite-B have also been used to estimate a global lightning flash rate.
Global detection of lightning is necessary to determine the global flashing rate and how this rate relates to other parameters in the global circuit. In recent years, the National Aeronautics and Space Administration has developed new optical sensors that could be used to detect and locate lightning in the daytime or at night and with continu ous coverage by using satellites in geosynchronous orbits. These sensors are capable of measuring the spatial and temporal distribution of lightning over extended periods with good spatial resolution and offer significant new opportunities for research — without the inherent sampling biases of low-altitude orbiting satellites— and for many applications.
Artificial triggering of lightning now provides the capability of studying both the physics of the discharge process and the interactions of lightning with structures and other objects in a partially controlled environment (see Chapter 2) . When a thunder storm is overhead and the surface electric field is large, a small rocket is launched to carry a grounded wire rapidly upward. When lightning is triggered by the wire, the first stroke is not like natural lightning, but subsequent return strokes appear to be almost identical to their natural counterparts.
Triggered lightning is now being used to investigate the luminous development of lightning channels, the characteristics of lightning currents, the velocities of return strokes, the relationships between currents and electromagnetic fields, the mechanisms of lightning damage, the performance of lightning protection systems, and many other problems. The main benefit of this triggering technique is that it can be used to cause lightning to strike a known place at a known time, thus enabling controlled experi ments to be performed. Although lightning cannot be reproduced in full in the labora tory, several lightning simulators have been developed and have provided some quanti tative information on the generation of thunder.
Cloud electrification and charge-separation processes are closely coupled to the cloud microphysics and the storm dynamics. The natural storm environment is ex tremely complicated, and its quantification involves a host of electrical and meteoro logical parameters. Many of these parameters and their measurements are treated in Chapters 7 and 8 and the three-volume publication, Thunderstorms: A Social. Scien tific, and Technological Documentary, edited by Kessler (1982). One of the greatest needs is for an in-cloud instrument that can measure in a thunderstorm environment the charge on the smaller cloud particles as a function of particle size and type (see Chapter 8).

14

OVERVIEW AND RECOMMENDATIONS

In recent years, cooperative field programs, such as the Thunderstorm Research International Program (TRIP), have improved our knowledge about the overall elec trical structure of thunderstorms. These programs have also provided a framework wherein a number of different investigators using different techniques can study the same thunderstorms at the same location at the same time. For instance, in situ and remote electrical measurements have been made in New Mexico in conjunction with Doppler radar studies of the cloud precipitation and dynamics and in-cloud sampling of the larger cloud particles. Laboratory experiments also continue to provide informa tion on the effects of electricity on cloud microphysics and charge separation mecha nisms that are critical to the interpretation of data collected by such field programs. Finally, numerical models of the electrical development and structure of thunder storms have provided an important framework in which to interpret the cloud mea surements and laboratory experiments.
The techniques that are used to determine the electrical structure of the fair-weather atmosphere are diverse. Usually, vertical profiles of one or more atmospheric-electrical variables— typically electric field, conductivity, and current density— are measured over a relatively short time span. Sensors are carried aloft on aircraft, balloons, or rockets, and the data are presented both as profiles and as numerically integrated results. The vertical profiles represent almost instantaneously measured parameters rather than time averages. Aircraft or constant-pressure balloons, however, do have the capability of measuring temporal variations at a given level. Profiles have been measured over land because of the convenience and to study specific terrain effects and over water in attempts to eliminate distortions caused by land. Profiles have led to the detection of convection currents in the planetary boundary layer that are comparable in magnitude with the total current, the electrode effect over water under stable condi tions, the response of columnar resistance to pollution, and the diurnal variation in ionospheric potential.
In addition to the standard electrical parameters, determinations of ion and aerosol contents and compositions, mobility, and chemistry are all critical to an understanding of conductivity. Again, these quantities are usually displayed as vertical profiles and, to a certain extent, are characteristic of the type of platform on which the instrumentation is carried aloft (e.g., some airplane measurements provide horizontal profiles but are limited in their vertical extent). Vertical measurements are probably a good first ap proximation to the global electrical structure; however, horizontal variations should also be measured as a function of time to complement the profile data. Tethered bal loons can provide the time variations of selected electrical properties at a few locations, and this has been attempted on an experimental basis. For understanding the global circuit, it would be valuable to have a number of vertical profiles taken at the same time and, in addition, to repeat certain profiles to obtain the time variations.
Knowledge of the upper-atmosphere current systems is important for understanding the interactions among the ionosphere, magnetosphere, and the solar wind. Some of these current systems were studied during the International Magnetospheric Study, 1976-1980, and during the NASA Dynamics Explorer satellite program. The goals of these programs have been to investigate the coupling of the solar-wind energy through the magnetosphere and into the ionosphere; but little effort was made to couple these current systems into the global electrical circuit. Additional measurements of magneto sphere- ionosphere currents are planned for the International Solar-Terrestrial Physics program.

SOCIETAL IMPACT
Severe weather phenomena that disrupt our lives include tornadoes, hail, high winds, hurricanes, floods, snowstorms, and lightning. Among them, lightning ranks as

OVERVIEW AND RECOMMENDATIONS

15

the number one killer, followed closely by tornadoes. Lightning is much less dramatic than a tornado passing through an area or a severe snowstorm that paralyzes a city, but lightning can strike quickly and kill with little or no warning.
Lightning is a leading cause of outages in electrical power systems and was the initial cause of the massive power blackout in New York City on July 13, 1977. The possible effects of lightning on advanced aircraft, nuclear power stations, and sophisticated military systems are problems of increasing concern.
The detailed physics of how lightning strikes a structure, a power line, or an aircraft and its effects are still not known. The approaching leader is not influenced by the object that is about to be struck until it is perhaps a few tens of meters away. At that time, an upward-moving streamer leaves the object and similar discharges may also leave other objects nearby. When the upward-moving streamer attaches to the down ward-moving leader, the return stroke begins. When the details of this attachment are better understood we should be able to predict with higher probability what will and what will not be struck under various conditions and thereby provide better lightning protection. For example, the positioning of overhead ground wires above power trans mission lines and the protection of complex structures could be optimized (see Chapter 5).
The current rise time is an important parameter for lightning protection because if the current interacts with an inductive load, the voltage on that load is proportional to the rate of increase of the current. Most of the standard surge waveforms that are used to verify the performance of protectors on power and telecommunications circuits spec ify that open-circuit voltage should have a rise time of 0.5, 1.2, or 10 microseconds and that the short-circuit current should have a rise time of 8 or 10 microseconds. These values are substantially slower than recently measured lightning current rise times, which are in the range of tens to hundreds of nanoseconds; therefore, it is probable that the degree of protection that is provided by devices tested to present-day standards will not be adequate for protection against direct lightning surges.
The unusually destructive nature of lightning that lowers positive charge to ground is only partially documented and is poorly understood (see Chapter 3) . Because of the large and long continuing currents, positive lightning may ignite a disproportionately large number of fires, especially in grasslands and forests. The apparent pattern is for positive lightning to strike preferentially outside areas of rainfall, and this further en hances the likelihood of its starting a fire. Positive lightning may be correlated with storm severity and tornado occurrence, and its detection could enhance our present severe-storm detection and warning systems.
Newly developed lightning-detection equipment now makes it possible to make real time decisions on the preparations for repairs of utility systems, early warning and detection of lightning-caused forest fires, and a variety of other warning functions in situations that allow protective action to be taken, such as launches at the NASA Ken nedy Space Center and outdoor recreational activities. Among the main users of light ning location data at present are the Bureau of Land Management (BLM) in the west ern United States and Alaska and the Electric Power Research Institute in the eastern United States. The BLM and the Forest Services of most Canadian provinces utilize the time and location of lightning storms to determine when and where to look for forest fires. Early detection of these fires provides considerable savings both in the natural resources and in the cost of fighting the fires. In the eastern United States the lightning data are being used to accumulate statistics on lightning occurrence and for real-time applications by electric power utilities. For warnings of lightning-intensive storms, these data are also disseminated in real time to many National Weather Service offices and to a growing number of television stations.
Although cloud electrification processes are ultimately responsible for producing lightning, these processes can also electrify an aircraft flying in a cloud. It is quite

OVERVIEW AND RECOMMENDATIONS
common for the potential of an aircraft to be raised by several million volts, and most planes have discharger wicks to control the interference in radio communications when the aircraft goes into point discharge. Lightning will not usually present a hazard to commercial aircraft as long as the present design practices are continued and the stan dard practice of avoiding large thunderclouds is maintained. However, since manynew aircraft are being developed with composite materials instead of aluminum and with the increased use of computers and microcircuit technology, the decreased electri cal shielding on the outside and increased sensitivity inside means that there will be an increasing vulnerability to lightning disturbances (see Chapter 5) . Thus, cloud electric ity and lightning will have to be considered carefully in the design and operation of future aircraft systems.
Natural telluric currents can significantly disturb man-made systems such as com munication cables, power lines, pipelines, railways, and buried metal structures. The largest natural disturbances are associated with the intense auroral current systems that flow at high latitudes during geomagnetic storms. There have been frequent reports of these disturbances, inducing currents on long telephone and telegraph wires that are large enough to generate sparks and even permanent arcs. When this occurs, there can be outages and shutdowns in both land and sea cables and fires can be started by over heating the electrical systems. Currents of up to about 100 A are sometimes induced in power transformers at northern latitudes and cause power blackouts and system fail ures. During the large geomagnetic storm of February 11, 1958, the Toronto area suffered from an induced power blackout.
Long pipelines are also affected by telluric current disturbances. The Alaskan pipe line has been the subject of careful investigation, principally because of its location across the auroral zone. One of the concerns has been the rate of corrosion of the pipe line, which is enhanced by telluric currents. However, telluric currents appear to affect electronic equipment related to operational monitoring and corrosion control rather than to produce specific serious corrosion problems. A relationship between the ex pected current flow and geomagnetic activity has been derived and suggests that the pipeline is a large man-made conductor that is capable of significantly affecting the local natural regime of telluric currents.
There is also a concern that the long, power-transmission lines planned for future arctic development will be subject to larger induced currents by auroral activity than was previously considered. This would require new protection equipment develop ment for high-latitude applications.
Telluric currents have also been used in the search for natural resources with two different approaches— magnetotellurics and geomagnetic depth sounding. Telluric currents can also be used to study long-period tidal phenomena and water flows and the Earth's astronomical motion and as possible precursors for earthquakes and volcanic eruptions. It has also been suggested that a natural waveguide for telluric currents in the Earth's crust, consisting of an insulating layer of dry rocks sandwiched between an upper hydrated conducting layer and an underlying conducting hot layer, could be used for communications. There are also investigations to determine the feasibility of using the natural resonances in the earth-ionosphere waveguide, Schumann reso nances, as a means for long-distance communications.
RECOMMENDATIONS
An increased interest in understanding the Earth's electrical environment has re sulted from recent advances in different disciplines, along with the recognition that many of man's modern technological systems can be adversely affected by this environ ment. This understanding appears to be on the threshold of rapid progress.

OVERVIEW AND RECOMMENDATIONS

17

There should be a concerted effort of coordinated measurement campaigns, supported by critical laboratory experiments, theory, and numerical modeling of processes, to improve our understanding of the Earth's electrical environment.
Because the study of the electrical environment is commonly divided into three ma jor components — lightning, cloud electricity, and the global circuit (including ion chemistry and physics and ionospheric, magnetospheric, and telluric currents)— the specific needs in these areas are detailed below. However, it should not be forgotten that there are interactions among these components and that the understanding of these interactions may be fundamental to an understanding of an individual compo nent.
1 . More needs to be known about the basic physics of the lightning discharge and its effects on structures in order to design proper protection systems.
Most lightning begins within a thundercloud, but the initiation and subsequent de velopment of a flash within the cloud are poorly understood. The physics of electrical breakdown over distances of 10-10,000 m is not understood, nor is the relationship between the channel geometry and the fields and charges that existed before the dis charge. The fields and currents that are produced by most of the important lightning processes have large submicrosecond variations, but how the discharge currents de velop as a function of space and time and what the ranges of variability of the maxi mum / and d//dt parameters are need to be determined. The power and energy bal ances within the lightning channel and many other important lightning parameters also need to be determined.
There should be a comprehensive and carefully coordinated effort to understand the basic physics of intracloud and cloud-to-ground lightning discharges and their effects on our geophysical environment. This new knowledge should be applied to the devel opment of improved lightning-protection methods.
Several new techniques are now capable of providing much insight into the complex and varied physical processes that occur during a lightning flash. For example, radio interferometry and time-of-arrival methods can be used to trace the three-dimensional development of lightning channels with microsecond resolution. Rockets can be used to trigger lightning under a thunderstorm, so that many of the physical properties of the discharge and its interactions with structures can be studied in a partially controlled environment.
2. The question of how thunderclouds generate electricity has been a fascinating scientific problem for over two centuries, but only in the past decade have cooperative experiments using new experimental techniques provided valuable insights into the complex and varied electrical processes that occur within clouds. Unfortunately, there are no sensors that can determine the charge-size relationship on the smaller cloud particles inside a thunderstorm; thus the data are not adequate to determine which of the many possible mechanisms dominate the generation and separation of charge. In addition to not knowing the charge-size relationship for various cloud particles, it is not known how this relationship evolves with time when there is lightning. The electrical forces on individual elements of precipitation can be several times larger than gravity, but further research is needed to determine whether (and how) these and other electri cal effects play a significant role in the formation of precipitation.
In view of the successes of recent research, significant new understanding of cloud electricity and lightning can be made by continuing to develop new instruments and by making coordinated in situ and remote measurements of selected thunderclouds. These studies should be complemented by measurements of cloud microphysics and dy namics, by comprehensive laboratory studies, and by theory and numerical modeling.
The complexity of the processes that produce both precipitation and lightning makes

18

OVERVIEW AND RECOMMENDATIONS

it impossible to construct or validate theories of cloud electrification from simple field experiments. It is only through the complementary efforts of comprehensive field ob servations, laboratory experimentation, and numerical modeling that we can hope to understand the physical processes that are important in thunderstorms. An improved understanding of the major processes that create strong electric fields and their interac tions with cloud particles and precipitation might lead to better forecasting of electrical hazards to aviation, forestry, and other outdoor activities.
The first goal of the in-cloud measurements should be to determine the charge-size relationship for various cloud and precipitation particles and the role of screening lay ers in the upper and lower regions of the storm. The electric current densities that flow above and below the cloud should also be monitored as a function of time. Since the natural storm environment is complicated, laboratory experiments should focus on the detailed physics of mechanisms that appear to be important on the basis of both the incloud measurements and the numerical models. Laboratory experiments should also determine the effects of electric fields on drop coalescence efficiencies and the ability of electrified drops to scavenge charged constituents of atmospheric aerosols. Analyses of the in-cloud and laboratory data could be accelerated through the establishment of a common data base, particularly for theory and numerical modeling efforts.
3. Even in fair weather the solid earth and atmosphere are electrified. Thunder storms have been identified as the dominant generator in the global electric circuit, but many details remain concerning storms as electrical generators and their electrical in teractions with their neighboring environment. Lightning and the steady currents above and below thunderclouds play an important role in maintaining an electrical potential between the upper atmosphere and the surface, but the amount and type of lightning and the values of cloud currents that flow to the surface and the upper atmo sphere are not well known. The lightning phenomenology and cloud currents may depend on many factors, such as the geographical location of the storm, the season, and the meteorological environment; these dependencies have yet to be determined. The charge transports to the surface under a storm are due to linear and nonlinear fielddependent currents, precipitation and other forms of convection currents, and light ning. Unfortunately, the values of each of these current components and their depen dence on the stage of the storm, the lightning-flash frequency, or the local terrain are poorly known. The charge transports to land and ocean surfaces that occur in fair weather, and also to mountainous terrain, need to be determined.
With recent progress in the development of satellite lightning sensors and the tech nology for measuring the electrical effects of storms with rocket-, balloon-, aircraft-, and ground-based sensors, a new attack on this fascinating problem of atmospheric electricity is needed.
There should be an effort made to quantify further the electrical variables that are acting in the global electric circuit and to determine their relationship to the various current components that flow within and near thunderstorms. There is also an impor tant need for theoretical and numerical studies to quantify further the role of thunder storms as generators in the global circuit.
The establishment of the ionospheric potential, or some other globally representative parameter, as a geoelectrical index that gives an indication of the state of the global circuit would be extremely useful. This index would be the electrical equivalent of the geomagnetic index that has been used for many years to characterize geomagnetic phe nomena.
The effects of stratiform clouds and large-scale cyclones on the global circuit also need to be quantified. Once a globally representative parameter that describes the state of the global circuit has been obtained, it can then be related to other remotely observed

OVERVIEW AND RECOMMENDATIONS

19

quantities such as the global lightning flashing rate or directly observed quantities such as the air-earth current or surface electric field.
4. Electromagnetic and optical sensors, both on the ground and on satellites, can be used to (1) detect and map lightning on a regional, national, and global scale and (2) determine, for the first time, how much lightning actually occurs and its geographic distribution as a function of time. With ground-based sensors, it should be possible to determine whether and how the characteristics of individual lightning flashes depend on their geographical location and the storm structure. If a global detection capability were implemented, it would be possible to map and monitor the intensity of lightning storms and to examine the effects of lightning on the global circuit, the ionosphere, and the magnetosphere. When combined with simultaneous spectroscopic measurements, the satellite data could also be used to determine when and where lightning produces significant concentrations of trace gases in the atmosphere.
A lightning sensor, capable of measuring lightning flashes during both day and night, should be flown on a geosynchronous satellite at the earliest possible date. The resulting data when combined with those from other sensors and data from groundbased detection networks will provide information that could be used to relate light ning to storm size, intensity, location, rainfall, and other important meteorological parameters.
5. Electrical processes in the lower atmosphere and, in particular, within the plane tary boundary layer, are important because these, together with global variations, determine the electrical environment of man and the biosphere. Galactic cosmic rays and various radioactive decays produce atmospheric ions that undergo a complex and still only partially understood series of ion-chemical reactions. The composition of the ions is poorly known between the surface and about 50 km, and profile measurements are needed. How the ion characteristics relate to atmospheric aerosols and various trace gases needs to be determined before the bulk electrical properties of the atmosphere can be understood. A significant fraction of the ions attach to atmospheric particles; there fore, smoke and other forms of particulates can significantly affect the electrical prop erties of the lower atmosphere.
Turbulence and convection in the planetary boundary layer play an important role in establishing the vertical distributions of ions, trace gases, and particles. These pro cesses also transport space charge and drive convection currents that alter the electrical properties of the planetary boundary layer.
The clarification of the chemistry of atmospheric ions, their mobilities, and the phys ics of electrical processes in the troposphere and stratosphere will require further mea surements, particularly in determining how these processes are affected by man s activ ities and natural events. There is also a need for further laboratory measurements and modeling to determine the important chemical reactions and ion composition in the atmosphere.
A number of meteorological research stations in a variety of geographic locations should begin to measure electrical parameters routinely to determine the relationships between electrical and meteorological processes. Vertical profile measurements of elec trical properties should be continued in an attempt to determine their relationships to aerosols and trace-gas chemistry. Provisions should be made for the expansion of such synoptic measurements during planned international programs (e.g., the Global Change Programme of the International Council of Scientific Unions, which is cur rently in the planning stages).
6. Recent research has indicated that the mesosphere may not be electrically passive but may, in fact, contain active electrical generators that are not understood. In addi

20

OVERVIEW AND RECOMMENDATIONS

tion, ground-based and balloonborne measurements have indicated that there is a global electrical response to cosmic-ray and solar variations that is also not understood. The bulk electrical properties of the middle atmosphere are poorly defined, and there is a need to determine the ion composition and chemistry both for quiet conditions and during solar-terrestrial events. There is also new evidence that the electric fields pro duced by thunderstorms and lightning can produce significant disturbances in the elec trical structure of the upper atmosphere and magnetosphere.
The horizontal electric fields that are generated by the ionospheric-wind dynamo and the solar-wind/magnetospheric dynamo propagate downward to the Earth's sur face where they can locally perturb the fair-weather electric field by about 1-2 percent and 20-50 percent, respectively. Horizontal currents in the middle atmosphere and the characteristics of the equalization layer need to be determined in order to understand better the electrical interactions that occur between the upper and lower atmosphere.
To determine (1) the electrical properties of the middle atmosphere, (2) the effects of thunderstorms on ionospheric and magnetospheric processes, and (3) the effects of time variations in the cosmic-ray and energetic solar-particle fluxes on the properties of the global circuit, additional measurements are required. Theoretical investigations and modeling are also important components of such investigations.
Lightning has long been known to be a source of whistlers in the Earth's magnetosphere, and recent spacecraft observations suggest that lightning also generates whis tler-mode signals on Jupiter. The questions of just how lightning fields couple to a whistler duct and whether these fields have effects on the ionosphere or magnetosphere are important and need further investigation.

REFERENCES
Dolezalek. H., and R. Reiter, eds. (1977). Electrical Processes in Atmospheres, Steinkopff, Darmstadt, Ger many.
Kessler, E.,i'd. (1982). Thunderstorms: A Social. Scientific, and Technological Documentary , Univ. of Okla homa Press. Norman, Okla. .
Orville, R. E., ed. (1985). Proceedings of the Vllth International Atmospheric Electricity Conference, spe cial issue of /. Geophys. Res. 90 (June 30, 1985).
Volland. H., ed. (1982). Handbook of Atmospherics, Vol. 1 and II, CRC Press, Inc., Boca Raton, Fla.

I LIGHTNING

Lightning Phenomenology
1
RICHARD E. ORVILLE State University ofNew York at Albany

INTRODUCTION
Severe weather phenomena that disrupt our lives in clude tornadoes, hail, high winds, hurricanes, snow storms, and lightning. It is not well known that in most years, lightning ranks as the number one killer, followed closely by tornadoes. Much less dramatic than a tornado passing through an area or a severe snowstorm that par alyzes a city, a lightning ground strike can quickly kill one or two people in less than a second with little or no warning. Annually in the United States about 100 peo ple are killed by lightning strikes, and reliable estimates for the world would be in the thousands. Lightning on a global and regional scale is an area of science that brings together the interests of the atmospheric physicist, chemist, and meteorologist in an effort to learn its char acteristics.
The phenomenology of lightning involves the fre quency of lightning observed over large spatial and time scales. It involves the maximum and average flashing rate per unit area and the variation of flash characteris tics with location and storm type. Studies of lightning phenomenology can now be discussed in terms of both satellite and ground-based observations. With the use of satellites, we obtain data on the global lightning flash rates and the distribution of lightning with respect to the continents and oceans. With the extensive use of
23

ground-based observations, we can determine the flash ing rates and flash characteristics of individual storms. In addition, we can monitor the variations of the ground flashes as a function of location and storm type.
SATELLITE OBSERVATIONS OF LIGHTNING
Optical Detectors
Significant advances in obtaining a better estimate of global flash rates and distribution have occurred as the result of satellite lightning observations in the last dec ade. Turman (1978, 1979), Turman et al. (1978), and Turman and Edgar (1982), using optical detectors on the Defense Meteorological Satellite Program (DMSP) satellites, showed the distribution of lightning at dawn and dusk for a period of 1 year. One example of this recent result is shown in Figure 1.1, where the dusk lightning distribution for November-December 1977 demonstrates the spatial distribution and the rate. Note that the lightning is found mostly in the southern hemi sphere, but significant activity still occurs in the north ern hemisphere.
The latitudinal and seasonal variation of the light ning activity is best shown by examining Figure 1 . 2 (Kowalczyk, 1981; Turman and Edgar, 1982). In this histo gram, the lightning rate has been summed over

24

»>P joniu*!

U"» O O «=>

"\ ,

_r" cp v *H ?' o

w

™ ^ 1 ^ =\

1

JS

::: ' ' 1

"

1 N lf> O O ^,

— d d -' n a

1^

,>::■ v£ -i*|

jj ." ;>4..:,.:

■giHJc /

^ s St HI |(R' i

i

at* i»3

J

S WK'v ~— ..'. l

1

wtn

L

- -.

45»- l -^v'.^

'( - J-

i

A

-<i

JC

1

' Y z> *w -||

4

miliaria iiiiiiii !'■ f ^ a :-a v / ,:.:.: ^pH$

CO

53l 11

,^"is j

1 n :,:,:.: ^^|

\

_pl " .

I.

3

1

£v§h|

]M:.JI^r-

*o
B

| K $ /' « ■ _J jjf
'Isij '^ MVr
' Bf
« *^ /? :':':':
^-
r

^*- -

i

?:,:.:.: £^
IE '

a
*

'

-c

"V=

H

X-

1

I

4

Ji =c
c
8 '5

Qa

;r (3

u

i-

R >.

£

.c r,

>s

00

B-
1

s. u F
J*=" <
"c

.™
B u0
3 cd a. =
t/3
8.
DX
>. s
h
CM
S*3 ,—-
a
— u 51
'J-
> -7, B
.7. r:
E
c c£
-C
st r-
-E

*>t ionium

O£

LIGHTNING PHENOMENOLOGY
DAWN AUG-SEP

DUSK

25
November. The progression of the lightning activity to ward the southern hemisphere as summer approaches in that hemisphere is evident. The striking absence of lightning over the ocean is apparent in all three months and clearly shows the importance of land in the produc tion of thunderstorms.

rfm-m-rn-L-r
6Cfe ¥fS 2(ft 0° 20°N 4CfN 6(fN 60° 4<fS 20°S 0° 2CPN 4CfN 6CfN LATITUDE
FIGURE 1.2 The latitudinal variation of the dawn and dusk global lightning activity as a function of season. Adapted from Kowalczyk and Bauer (1981 j and Turman and Edgar (1982).
longitude at both dawn and dusk and presented as a function of latitude. The dusk distributions show a smooth change as the seasons change. But note the sec ondary peak at 30-40° that persists through all the dusk distributions except August-September. The dawn dis tributions do not change as smoothly, but the seasonal shift is apparent. The enhanced polar-front activity at dawn is quite evident in the November-December southern hemisphere and the April-June northern hemi sphere. At dawn there appears to be an overall mini mum for the period January-March. Analysis of the lightning rates for an entire year show a variation of 10 percent from a global value that is estimated to range from 40 to 120 flashes per second.
Other studies using the DMSP data from a different sensor provide a glimpse at the global midnight light ning activity with a spatial resolution of approximately 100 km (Orville and Spencer, 1979; Orville 1981). A study of the global midnight lightning activity yields a lightning rate of 96 flashes per second, but this could be in error by a factor of 2 (Orville and Spencer, 1979). Orville plotted a series of monthly maps reproduced in Figure 1.3 for the months of September, October, and

Radio Detectors
Recent measurements of high-frequency radio noise by the Ionosphere Sounding Satellite-B have been used by Kotaki et al. (1981) to estimate a global lightning flash rate of approximately 300 per second, in contrast to the optical measurements discussed previously. The radio measurements may overestimate the lightning fre quency since it is assumed that all the emissions are pro duced by lightning. But the satellite optical measure ments are uncertain in estimating the lightning rate since the fraction of lightning that is actually detected depends on a calibration factor that represents a best estimate.
Despite the availability of satellites to estimate the global lightning activity, we have made little progress in obtaining a flash rate with small error bars owing to the present experimental limitations of sensor sensitivity, area coverage, and the number of satellite platforms. Nevertheless, the satellite observation provides us with the first reliable estimate of the distribution of global lightning. The resolution of the varying global flash rate estimates may depend on the close coordination of satel lite-based and ground-based observations of lightning and the availability of larger-coverage-area platforms, such as geosynchronous satellites.
GROUND OBSERVATIONS OF LIGHTNING
Most of our information on the characteristics of lightning has come, and will continue to come, from ground-based observations of the lightning flash. Many of these studies have focused on the ratio of intracloud to cloud-to-ground flashes, the lightning ground-flash density, and the flashing rate of different types of thun derstorms.
Intracloud Versus Cloud-to-Ground Lightning
The ratio of intracloud to cloud-to-ground lightning is of fundamental importance. How does this ratio vary with latitude and longitude, and how does this ratio vary in the lifetime of a storm? Are there storms that have nearly all intracloud flashes and consequently are less damaging, and are there storms that have almost all ground flashes and consequently are of greater concern?

RICHARD E. ORVILLE
FIGUHE 1.3 Three maps showing the progression of monthly lightning for (a) September, (b) October, and (c) November. From Orville (1981), reproduced with permission of the American Meteorological Society.

LIGHTNING PHENOMENOLOGY

27

Studies by Prentice and Mackerras (1977) have sum marized much of the available data on the ratio of intracloud to cloud-to-ground flashes (NrINK). From an anal ysis of 29 data sets from 13 countries, they obtain the following relationship for an average thunderstorm:

NtJNK = (4.16 + 2. 16 cos 3X)

0.4T

[ 0.6 +

(1.1)

72 - 0.98X

where T, the number of thunder days per year, is less than or equal to 84 and X, the latitude in the northern and southern hemispheres, is less than 60°. If the num ber of thunder days is unknown, then the ratio can be estimated from the relation

NJN, = (4.16 + 2. 16 cos 3X).

(1.2)

This result is plotted in Figure 1.4. Note that the ratio has the highest value in the tropics where most of the lightning was shown to occur by the satellite data. Re call, however, that the satellite data were composed of both intracloud and cloud-to- ground lightning flashes. There is at the moment no way to distinguish between a ground flash and an intracloud flash from a satellite.

Lightning Flash Density
The number of lightning strikes per unit time per unit area, or the flash density, is a fundamental quantity of interest. Most of the available information has been ob tained with lightning flash counters.
Prentice (1977) summarized the values for several ge ographical areas and reported 5 flashes per km2 per year in Queensland, Australia; 0.2 to 3 flashes per km2 per year in Norway, Sweden, and Finland; and 0.05 to 15 flashes per km2 per year in South Africa depending on the location.
Piepgrass et al. (1982) reported the results of studying 79 summer storms at the Kennedy Space Center, Flor ida, which produced 10 or more discharges, during the years 1974-1980. Using field mill sites covering an effec tive area of 625 km2, they observed an area flash density for all discharges during June, July, and August to range from 4 to 27 discharges per km2 per month, with a sys tematic uncertainty of perhaps a factor of 2 in the sam ple area. The mean and the standard deviation of the monthly area density over the above years was 12 + 8 discharges per km2. Approximately 38 percent of the dis charges were ground flashes. Therefore, they were able to estimate the ground flash density to be 4.6 ±3.1 flashes per km2 per month.
The most recent estimate of the ground flash density

in the United States has been made by Maier and Piotrowicz (1983) using thunderstorm hour statistics and is re produced as Figure 1.5. They used thunderstorm dura tion data from approximately 450 aviation weather reporting stations, each with an uninterrupted 30 years of records. The station density available is twice that of any previous thunderstorm frequency analysis of the United States. The maximum annual ground flash den sities of 18 per km2 are found in the western interior of Florida. High flash densities greater than 12 per km2 are found over much of Florida and westward to eastern Texas. Flash densities greater than 8 per km2 are found in most of Oklahoma, Kansas, Missouri, Arkansas, Lou isiana, Mississippi, and Tennessee. Most western and northeastern states have flash densities that are less than 4 per km2.
Lightning Flash and Related Characteristics
Data from two summers at the Kennedy Space Cen ter, Florida, have been used to estimate the flashing rates in thunderstorms (Livingston and Krider, 1978). It was observed that large storms evolve through an ini tial, an active, and a final phase of activity. Most of the lightning activity was observed to occur in the active phase with 71 percent of the lightning, although this phase of the storm occupied only 27 percent of the total storm duration. During the active phase, 42-52 percent of all lightning was to ground, while during the final storm period, only about 20 percent of the lightning was to ground. The discharge rate for all storms observed in 1975 was approximately 4 flashes per minute with a maximum flashing rate of 26 discharges per minute dur ing any 5-minute period. The highest flashing rate aver aged over an entire storm was about 9 discharges per minute for over 200 minutes. More recent data from a 4year interval indicates that the mean rate of flashes is about 2.4 discharges per minute per storm (Piepgrass et al., 1982).
The relationship of rainfall to lightning flash rates has been investigated by Piepgrass et al. (1982). They re ported that when the meteorological conditions favor the production of lightning, there is almost a direct pro portionality between the total rain volume and the total number of flashes. Maier et al. (1978) noted in an earlier paper that the lightning counts were proportional to the total storm rainfall and that the proportionality in creased with the rain volume until the rainfall reached about 1.2 to 2.7 X 104 m3 per flash. Beyond these vol umes, storms that produced more rainfall tended to pro duce proportionally less lightning. Piepgrass et al. (1982) point out that, "Clearly, these problems warrant further study."

28

RICHARD E. ORVILLE

30 40 50 X , DEGREES
FIGURE 1 .4 The ratio of intracloud to cloud-to-ground lightning as a function of latitude. From Prentice and Mackerras (1977).
Lightning Location Networks
The study of lightning phenomenology has made a major advance in the last decade with the introduction of new magnetic direction-finding techniques (Krider et al., 1980) that provide the means to monitor ground strikes over areas exceeding 106 km2. Extensive networks of lightning direction finders have been established for forest-fire detection in the western United States, Can ada, and Alaska. Figure 1.6 shows the coverage as of the summer of 1984, and it can be predicted that within the next few years the entire United States will be covered.
One expanding lightning detection network covers

the East Coast and is approaching the Mississippi River to provide coverage of the eastern part of the United States (Orville et al., 1983) . This network is operated by the State University of New York at Albany in a multi drop communication network that links all the direction finders to one computer. Data are now retrieved on the time, location, number of strokes in the flash, polarity of the charge lowered to ground, and amplitude of the peak magnetic radiation field that can be related to the maximum current in the first stroke. These data, in turn, can be analyzed and related to the meteorological patterns producing the observed phenomena.
To report all initial results would far exceed the space available in this brief paper; nevertheless, it is interest ing to note a few observations. The highest ground flash rate recorded by the East Coast Network occurred on June 13, 1984, when 50,836 flashes were detected in a 24-hour period over an area of approximately 250,000 km2. The highest hourly summary was 7800 flashes with the highest 5-minute rate exceeding 10,000 ground flashes per hour. These results are remarkable when it is realized that these flash rates were from storms in only three states— Pennsylvania, New Jersey, and part of New York— and at the time were producing approxi mately 3 percent of the entire global lightning activity.
Other results indicate that lightning is recorded in every week of the year along the East Coast and that the polarity of the lightning ground strikes shows a change from negative to positive in the fall and a shift back to negative in the spring. A discussion of positive lightning and its characteristics is presented by Rust (Chapter 3, this volume) .

FIGURE 1.5 Lightning flash density esti mates on an annual basis. Adapted from Maier and Piotrowicz (1983) and MacGorman etal. (1984).

LIGHTNING PHENOMENOLOGY

29

FIGURE 1.6 Coverage of North America by time of arrival (dashed lines) and wide-band magnetic direction finders (solid lines) as of the summer of 1985.
CONCLUSIONS
The past decade has been a period of significant ad vances in lightning knowledge. Satellite studies have provided the first confirmation of the early estimates of the global lightning flash rates and added new informa tion on the distribution of lightning over the land and over the ocean. The development of widely distributed ground-based lightning networks provides for the first time the ability to monitor and calculate lightning char acteristics in near real time. Relating these parameters to the meteorological observations of visible and infra red images from space and to radar observations from the ground poses a major challenge in the near future.

REFERENCES
Kotaki, M., I. Kuriki, C. Katoh, and H. Sugiuchij (1981). Global dis tribution of thunderstorm activity, /. Radio Res. Labs. Japan 66.
Kowalczyk, M., and E. Bauer (1981). Lightning as a source of NO, in the troposphere, final report FAA-EE-82-4.
Krider, E. P., A. E. Pifer, and D. L. Vance (1980). Lightning direc tion-finding systems for forest fire detection, Bull. Am. Meteorol. Soc.61, 980-986.
Livingston, J. M., and E. P. Krider (1978). Electric fields produced by Florida thunderstorm, /. Geophys. Res. 83, 385-401.
MacGorman, D. R., M. W. Maier, and W. D. Rust (1984). Lightning strike density for the contiguous United States from thunderstorm duration records, prepared for Division of Health, Siting and Waste Management, Office of Nuclear Regulatory Research, U.S. Nuclear Regulatory Commission, Washington, D.C., 44 pp.
Maier, M. W., and J. M. Piotrowicz(1983). Improved estimates of the area density of cloud-to-ground lightning over the United States, presented at 8th International Aerospace and Ground Conference on Lightning and Static Electricity, June 21-23, 1983, Forth Worth, Texas.
Maier, M. W., A. G. Boulanger, and J. Sarlet (1978). Cloud-toground lightning frequency over south Florida, preprint. Confer ence on Cloud Physics and Atmospheric Electricity (Issaquah, Wash.), American Meteorological Society, Boston, Mass., pp. 605610.
Orville, R. E. (1981). Global distribution of midnight lightning Sep tember to November 1977, Mon. Weather Rev. 109, 391-395.
Orville, R. E., and D. W. Spencer (1979). Global lightning flash fre quency JVfon. Weather Rev. 107, 934-943.
Orville, R. E., R. W. Henderson, and L. F. Bosart (1983). An East Coast lightning detection network, Bull. Am. Meteorol. Soc. 64, 1029-1037.
Piepgrass, M. V., E. P. Krider, andC. B. Moore(1982). Lightning and surface rainfall during Florida thunderstorms, /. Geophys. Res. 87, 11193-11201.
Prentice, S. A. (1977). Frequencies of lightning discharges, in Physics of Lightning, R. H. Golde, ed., Academic Press, New York, pp. 465-496.
Prentice, S. A., and D. Mackerras(1977). The ratio of cloud to cloudground lightning flashes in thunderstorms, /. Appl. Meteorol. 16, 545-549.
Turman, B. N. (1978). Analysis of lightning data from the DMSP satel lite,/. Geophys. Res. 83, 5019-5024.
Turman, B. N. (1979). Lightning detection from space, Am. Scientist 67,321-329.
Turman, B. N., and B. C. Edgar (1982). Global lightning distribu tions at dawn and dusk, /. Geophys. Res. 87, 1191-1206.
Turman, B. N., B. C. Edgar, and L. N. Friesen (1978). Global light ning distribution at dawn and dusk for August-September 1977, EOS 59, 285.

Physics of Lightning
2
E. PHILIP KRIDER University of Arizona

INTRODUCTION
Lightning is a transient, high-current electric dis charge that occurs in the atmospheres of the Earth and other planets and that has a total path length on the or der of kilometers. Most lightning is produced by thun derclouds, and well over half of all discharges occur within the cloud. Cloud-to-ground flashes (Figure 2.1), although not so frequent as intracloud flashes, are, of course, the primary lightning hazard to people or struc tures on the ground. The continental United States re ceives an estimated 40 million cloud-to-ground strikes each year, and lightning is among the nation's most damaging weather hazards (see Chapter 1, this vol ume) . The peak power and total energy in lightning are very large. Thus far, it has not been possible to simulate in the laboratory either the geometrical development of a lightning channel or the full extent of lightning dam age. Lightning is a leading cause of outages in electric power and telecommunications systems, and it also is a major source of interference in many types of radio com munications. The possible effects of lightning on ad vanced aircraft, nuclear power plants, and sophisti cated military systems are problems of increasing concern.
Besides its many deleterious effects, lightning also has some unique benefits. The chemical effects of lightning
30

may have played an important role in the prebiotic syn thesis of amino acids, and today lightning is still an im portant source of fixed nitrogen, a natural fertilizer, and other nonequilibrium trace gases in the atmosphere (see Chapter 6, this volume). Also, lightning-caused fires have long dominated the dynamics of forest ecosystems throughout the world. The electromagnetic fields that are radiated by lightning can be used to study the phys ics of radio propagation and have been used for many years in geophysical prospecting. Also, lightning-caused "whistlers" are still being employed to study the charac teristics of the ionosphere and magnetosphere. Light ning plays an important role in maintaining an electric charge on the earth and is therefore an important com ponent of the global electric circuit (see Chapter 15, this volume). It is clear, therefore, that an understanding of the physics of lightning is important to further insight into our geophysical environment as well as for the de velopment of optimum protection from the lightning's hazards.
In recent years, new experimental techniques have enabled researchers to obtain a better understanding of the physics of lightning. Among these techniques have been applications of optical, acoustic, and electromag netic sensors to measure the properties of various dis charge processes on time scales ranging from tens of nanoseconds to several seconds. These measurements

PHYSICS OF LIGHTNING

31

FIGURE 2.1 Cloud-to-ground lightning over Tucson, Arizona.

have also been used to infer properties of the thunder cloud charge distribution that is affected by lightning (see Chapter 8, this volume) .
Within the category of cloud-to-ground lightning, there are flashes that effectively lower negative charge to ground and those that lower positive charge. Most ground flashes are negative, but there is recent evidence that positive discharges are often unusually deleterious, and, therefore, this type is discussed separately in Chap ter 3 (this volume).
Before describing some of the recent advances in lightning research in more detail, the processes that oc cur during a typical negative lightning flash to ground will be reviewed briefly. There are still many open ques tions about basic lightning phenomena, such as whether or how the characteristics of individual flashes depend on the type of thunderstorm, the season, and the type of terrain that is struck.
CLOUD-TO-GROUND LIGHTNING
Simplified sketches of the luminous processes that oc cur during a typical cloud-to-ground discharge are given in Figures 2.2, 2.3, and 2.4. For more detailed discussions of these phenomena, the reader is referred to Schonland (1964), Uman (1969), and Salanave (1980). Figure 2.2a shows an assumed cloud charge-distribu tion just before the lightning begins. Concentrations of negative charge are shown at altitudes where the ambi ent air temperature is - 10 to - 20°C, typically 6 to 8 km above mean sea level. The positive charge is more diffuse than the negative and most of it is at higher alti tudes (see Chapter 8, this volume).
Cloud-to-ground lightning almost always starts

within the cloud with a process that is called the prelimi nary breakdown . The location of the preliminary break down is not well understood, but it may begin in the high-field region between the positive and negative charge regions, as shown in Figure 2.2b. After several tens of milliseconds, the preliminary breakdown initi ates an intermittent, highly branched discharge that propagates horizontally and downward and that is called the stepped-leader. The stepped-leader is sketched in Figures 2.2c and 2. 2d, and this process effec tively lowers negative charge toward ground. The indi vidual steps in the stepped-leader have lengths of 30 to 90 m and occur at intervals of 20 to 100 /tsec. The direc tion of the branches in a photograph indicates the direc tion of stepped-leader propagation; for example, in Fig ure 2. 1 each stepped-leader propagated downward.
When the tip of any branch of the stepped-leader gets close to the ground, the electric field just above the sur face becomes very large, and this causes one or more upward discharges to begin at the ground and initiate the attachment process (see Figure 2.3). The upward propagating discharges rise until one or more attach to the leader channel at a junction point that is usually a few tens of meters above the surface. When contact oc curs, the first return stroke begins. The return stroke is basically an intense, positive wave of ionization that starts at or just above the ground and propagates up the leader channel at about one third the speed of light (Fig ure 2.2e). The peak currents in return strokes range from several to hundreds of kiloamperes, with a typical value being about 40 kA. These currents carry the ground potential upward and effectively neutralize most of the leader channel and a portion of the cloud charge. The peak power dissipated by the return stroke

32
km 10- , km 10- , kmlOkm 10

E. PHILIP KRIDER
is probably on the order of 10s watts per meter of chan
nel (Guo and Krider, 1982), and the peak channel tem perature is at least 30,000 K (Orville, 1968).
The last few steps of the stepped-leader, the onset of a connecting discharge, and the beginning of a return stroke are illustrated in Figure 2.3. Here, the distance between the object that is about to be struck and the tip of the leader when the connecting discharge begins is called the striking distance (SD) and is an important concept in lightning protection. The distance to the ac tual junction (J) between the leader and the connecting discharge is often assumed to be about half the striking distance.
After a pause of 40 to 80 milliseconds, most cloud-toground flashes produce a new leader, the dart leader, which propagates without stepping down the previous return-stroke channel and initiates a subsequent return stroke. Most flashes contain two to four return strokes, and each of these affects a different volume of cloud charge (see also Chapter 8, this volume). Figure 2.4 shows a sketch of a dart leader and the subsequent re turn stroke. Visually, lightning often appears to flicker because the human eye can just resolve the time inter vals between different strokes. In 20 to 40 percent of all cloud-to-ground flashes, the dart leader propagates down just a portion of the previous return-stroke chan nel and then forges a different path to ground. In these cases, the flash actually strikes the ground in two places, and the channel has a characteristic forked appearance that can be seen in many photographs. (See the left two flashes in Figure 2.1.)

km 10- .
km 10 5km
FIGURE 2.2 Sketch of the luminous processes that form the steppedleader and the first return stroke in a cloud-to-ground lightning flash.

IMPORTANT RESULTS OF RECENT RESEARCH
Three types of research have recently provided new information about the physics of lightning. These are described briefly, and their importance is indicated. First, we discuss how remote measurements of electric and magnetic fields can be used to infer properties of lightning currents, including some implications of re cent measurements. Next, we describe how the sources of radio-frequency (rf) noise can be used to trace the geometrical development of lightning channels as a function of time and to determine other properties of lightning. Finally, we discuss how small rockets can be used to trigger lightning artificially and give some appli cations of this technique.
Time-Domain Fields and Lightning Currents
Recently, it has become clear that the electric and magnetic fields that are radiated by different lightning

PHYSICS OF LiGHTNISG

33

i Leader

(a)

(b)

(c)

(d)

(e)

h 15Ms +— 15/is + 5,is f—0.5/is H

FIGURE 2.3 Sketch of the luminous processes that occur during at tachment of a lightning stepped-leader to an object on the ground.

processes have different but characteristic signatures that are reproduced from flash to flash. For example, Figure 2.5 shows three of the many impulses that were radiated by a typical cloud-to-ground flash at a distance of about 50 km. These particular signatures were re corded using a broadband antenna system and an oscil loscope that covered all frequencies from about 1 kHz to 2 MHz. Trace (a) shows a cloud impulse that was radi ated during the preliminary breakdown; trace (b) shows the waveform that was radiated by the first return stroke; and trace (c) shows a subsequent return stroke. The small pulses that precede the first return stroke in trace (b) were radiated by individual steps of the stepped-leader just before the attachment occurred (see Figure 2.3). The characteristics of these newly mea sured signatures have been put to use in the detection and location of cloud-to-ground lightning. For exam ple, there are now large networks of magnetic direction finders that can discriminate between the shapes of the return-stroke fields and other processes and that can provide accurate locations of the ground-strike points (see Chapters 1 and 5, this volume; Kridereta/., 1980).
In a series of recent papers, Uman and co-workers have developed a theoretical model that describes the shapes of the electric and magnetic fields that are pro duced by return strokes at various distances (Uman et al., 1975; Masters al, 1981; Uman and Krider, 1982). One particularly important result of this work is the pre diction that during the first few microseconds of the stroke, i.e., just after the attachment process has been

—i 5km
FIGURE 2.4 The development of a lightning dart-leader and a re turn stroke subsequent to the first in a cloud-to-ground lightning flash.

34

E. PHILIP KRIDER

20 T
01

(a)
Ar
H111111111

LU

(b)

h-

LU

\
CO

l

1

1

1

1

1

1

1

1

1

1

o

jK^. (c)

-H

1

I

t

0 40 80 120 160 200

MICROSECONDS

FIGURE 2.5 Examples of electric- field impulses that were produced by a cloud-to-ground flash at a distance of about 50 km. Trace (a) was radiated during the preliminary breakdown, trace (b) is due to the first return stroke, and trace (c) is due to a subsequent return stroke.

completed, the waveform of the distant or radiation field is proportional to the channel current,

~ 'RAD (*)=

I(t - Die),

2rD

(2.1)

where £ is the vertical electric field that is measured at the ground at time t, no the permeability of free space, v the return stroke velocity, c the speed of light, and D the horizontal distance to the flash. A typical first return stroke will produce a peak field of about 8 V/m at a dis tance of 100 km (Lin et al., 1979). The stroke velocity near the ground is typically on the order of 108 m/sec (Idone and Orville, 1982). With these values, Eq. (2.1) predicts a peak current of about 40 kA, a value that is in good agreement with currents that have been measured in direct strikes to instrumented towers (Berger et al., 1975; Garbagnati etal., 1981).
Recently, Weidman and Krider (1978, 1980) exam ined the microsecond and submicrosecond structure of return-stroke E-field and field derivative, dEldt, signa tures. These investigators found that a typical first stroke produces an electric field "front" that rises in 2-8 /tsec to about half of the peak-field amplitude. This

front is followed by a fast transition to peak whose mean 10-90 percent rise time is about 90 nsec (see Figure 2.6). Subsequent stroke fields have fast transitions similar to first strokes, but fronts that last only 0.5 to 1 /tsec and that rise to only about 20 percent of the peak field. This fine structure in the initial return-stroke field is illus trated by the waveform shown in Figure 2.7.
Unfortunately, the origin of fronts in return-stroke fields is still not well understood, particularly for first strokes (Weidman and Krider, 1978). If fronts are pro duced by upward connecting discharges, then these dis charges must have lengths in excess of 100 m and peak currents of 10 kA or more. A front may be produced by a slow surge of current in the leader channel prior to the fast transition, but then this surge must contain currents on the order of 10 kA or more, and the associated chan nel length must be at least 1 km. To date, the available optical data are not adequate to determine whether ei ther of these processes (or both) does actually occur. Clearly, more research will be needed before we shall understand the physics of the important striking pro cess.

Peak 90% 10% Breakpoint
40 -\
30-
2 20-|
E 10-

10% -90%
N=I25 Mean = 90ns (j= 40ns

100

200

Nanoseconds

300

FIGURE 2.6 Histogram of the 10 to 90 percent rise times of the fast portions of return stroke fields over seawater.

PHYSICS OF LIGHTNING FIRST RETURN STROKE RANGE 19km
i—i i—i i i i i—i i i 4 ps/ d i v

35
The actual current rise time is important for the de sign of lightning protection systems (see Chapter 5, this volume). For example, if a 100-nsec current interacts with a resistive load, the voltage rise time on that load will be 100 nsec. Most of the standard surge waveforms that are used to verify the performance of protectors on power and telecommunications circuits specify that open-circuit voltage should have a rise time of 0.5, 1.2, or 10 /xsec and that the short-circuit current should have a rise time of 8 or 10 nsee (see IEEE Standard 587-80 and FCC Docket 19528, Part 68). These values are substan tially slower than those shown in Figure 2.6; therefore, it is probable that the degree of protection that is pro vided by devices that have been tested to the above stan dards will not be adequate for direct lightning surges.
Measurements of the maximum d£/drs that are radi ated by return strokes striking seawater are summarized in Figure 2.8. Here, the average maximum dEldt is about 33 Vimlnsee when the values are range-normal ized to 100 km using an inverse distance relation. To

200 ns/div
FIGURE 2.7 The initial portion of a return-stroke field recorded with fast time resolution. The same signal is shown on both traces, and the peaks coincide in time.

The rise times of the fast field components that are radiated by return strokes are summarized in Figure 2.6. Note that the mean 10 to 90 percent value is about 90 nsec. These submicrosecond components in the field must be caused by submicrosecond components in the current, but few of the currents that have been mea sured during direct strikes to instrumented towers show components as fast as 90 nsec (Berger et al., 1975; Garbagnati et al., 1981). It is possible that an upward dis charge from a tall tower or the electrical characteristics of the tower itself will reduce the rise time that is mea sured from the value that would actually be present in a strike to normal terrain. Therefore, new measurements of lightning currents and fields with fast time- resolution will be necessary before we can understand the true cur rent rise times.

V/ m //u s
FIGURE 2.8 Histogram of the maximum d£/df during the initial portion of return-stroke fields. All data have been range-normalized to 100 km using an inverse distance relation.

36

E. PHILIP KRIDER

measure a dE/dt signature on a 10-nsec time scale, it is essential that the field propagation from the lightning source to the measuring station be entirely over saltwa ter: otherwise there can be a significant degradation in the high-frequency content of the signal due to propaga tion over the relatively poorly conducting earth. It is possible that strikes to saltwater contain inherently faster rise times than lightning strikes to ground, but Weidman and Krider (1978) argued that this is proba bly not the case.
If Eq. (2.1) is valid, then the dE/dt values in Figure 2.8 can be used to infer the maximum dlldt in the light ning channel, i.e.,

2ir D

d//df =

— dE/dt.

(2.2)

ti0v

If a typical dE/dt is 33 V/m/^sec at 100 km and v is 108 m/sec, then Eq. (2.2) implies that the maximum dlldt is typically 1.5 X 10" A/sec during a return stroke, a value that is about a factor of 20 higher than the tower mea surements. At this point, it should be noted that, if re turn strokes do contain large current components with dlldt on the order of 1.5 X 10" A/sec, then, if these currents interact with an inductive load, the overvoltage will be substantially larger and faster than has pre viously been assumed and the lightning hazard will be greater. (See Chapter 5, this volume, for further discus sions of lightning protection . )
As a final point, we note that the electromagnetic fields that are produced by lightning are now known to be large and to change rapidly; hence, these fields them selves can be deleterious. For example, Uman et al. (1982) showed that, if an object is struck directly by lightning, then the associated electromagnetic distur bance may be substantially more severe than the electro magnetic pulse (EMP) produced by an exoatmospheric nuclear burst at all frequencies below about 10 MHz. Also, Krider and Guo (1983) showed that the typical peak field from a return-stroke at 100 km corresponds to a peak electromagnetic power at the source of at least 20,000 megawatts.

Locations of Lightning Radio Sources
The radio-frequency noise that is generated by light ning in the HF and VHF bands appears in the form of discrete bursts, and within each of these bursts there are hundreds to thousands of separate pulses. If the differ ence in the time of arrival of each pulse is carefully mea sured at four widely separated stations, the location of the source of each pulse can be computed, and the geo metrical development of the rf bursts can be mapped as

a function of time (Proctor, 1971). Unfortunately, the physical processes that produce HF and VHF radiation in lightning are not well understood. Proctor (1981) re ported that the pulses in most bursts are produced by a regular progression of source points, and, therefore, he suggests that bursts are produced by new ionization pro cesses and extensions of old channels.
If the source location of each rf pulse within a burst is plotted, the width of the associated "radio image" of the channel ranges from about 100 m to more than 1 km (Proctor, 1981, 1983; Rustan et al., 1980). Figure 2.9 shows the paths of the central cores of six successive lightning discharges that were reconstructed by Proctor (1983) . By combining reconstructions such as these with measurements of the associated changes in the electric field at the ground, Proctor inferred that in-cloud chan nels usually have a net negative charge and that the av erage line charge density is about 0.9 C/km. Also, by dividing the length of a channel segment by the time required for that portion to develop, Proctor deter mined that the average velocity of streamer formation ranges from 4 X 104 m/sec to 8 X 105 m/sec with a mean of (1.4 ± 1.2) X 105 m/sec. If Proctor's average charge density is multiplied by the velocity of channel forma tion, then the average current in developing channels would appear to be on the order of 100 A, a value that is in reasonable agreement with other estimates (Brook and Ogawa, 1977).
In an analysis of the locations of the first rf sources in 26 lightning flashes, Proctor (1983) found that most dis charges begin within or near precipitation, i.e., those regions of the cloud that produce a radar reflectivity greater than 25 dBZ. He also reports that all steppedleaders begin in a narrow range of altitudes where the ambient air temperature is - 5 to - 16°C. The average altitude of the initial rf sources in the flashes studied by Proctor was about 4 km above ground level ( - 10°C), and the standard deviation was only 440 m.
The geometrical forms of intracloud discharges range from concentrated "knots" or "stars" a few kilometers in diameter to extensive branched patterns up to 90 km in length. Proctor (1983) reports that successive discharges in a storm often form an interconnected system and that some flashes seem to extend the paths of earlier dis charges. In one case, two flashes that were separated by just 1 .6 sec produced tortuous channels that ran parallel to each other for almost 2 km, but the channels re mained about 300 m apart.
Although more data will be required before Proctor's results can be generalized, it is clear that time-of-arrival methods offer great promise for future research, partic ularly for those phases of lightning that occur within a cloud (see also Taylor, 1978; Rustan era/., 1980). Radio

PHYSICS OF LIGHTNING

37

FIGURE 2.9 Geometrical reconstructions of six successive lightning discharges in South Africa (Proctor, 1983). The top panel shows a plane view of the channels from above, and the bottom panel shows an elevation view that is a projection of the same channels on a vertical plane parallel to the x axis.

interferometer observations of lightning have also pro vided interesting results (Warwick et al., 1979; Hayenga and Warwick, 1981), and perhaps in the future interferometric methods will be developed to the point that they can provide unambiguous three-dimensional reconstructions of the discharge processes.
Artificial Triggering of Lightning
The last development that we describe is the artificial triggering of lightning by small rockets. This technique is particularly important because it provides, for the first time, the capability of studying both the physics of the discharge process and the interactions of lightning with structures and other objects in a partially con trolled environment. Although rockets were first used to study atmospheric electricity in the eighteenth century, the first artificial initiation of lightning was clearly demonstrated by Newman et al. (1967). The technique has subsequently been improved by researchers in France (Fieux et al., 1975; Fieux and Hubert, 1976; St. Privat d'Allier Research Group, 1982) and is now being used to investigate a variety of lightning problems in France, Japan, and the United States.
When a thunderstorm is overhead and the electrical conditions are favorable, a small rocket is launched and carries a grounded wire aloft. If the rocket is fired when

the surface electric field is 3 to 5 kV/m, then about two thirds of all launches will trigger a lightning discharge (Fieux et al. , 1978) . Most triggers occur when the rocket is at an altitude of only 100 to 300 m, and the first stroke in the flash usually propagates upward into the cloud. The majority of the subsequent strokes follow the first stroke and the wire to ground; but in about one third of the cases, the subsequent strokes actually forge a differ ent path to ground. These latter events are called "anomalous triggers" (Fieux et al., 1978). The first stroke in a triggered discharge is not like natural light ning, but subsequent strokes appear to be almost identi cal to their natural counterparts.
An example of lightning that was triggered by Hubert and co-workers is shown in Figure 2.10. The upward branching in this photograph was produced by a leader that propagated upward from the wire, and the bright, straight section of channel near the ground shows the path of the wire just before it exploded as a result of the lightning current.
Triggered lightning is now being used to investigate the luminous development of lightning channels, the characteristics of lightning currents, the velocities of re turn strokes, the relation between currents and fields, the mechanisms of lightning damage, the performance of lightning protection systems, and many other prob lems (Fieux et al., 1978; Hubert and Fieux, 1981; Horii,

38

E. PHILIP KRIDER

FIGURE 2.10 An example of a rocket-trig gered lightning flash in New Mexico.

1982; Miyachi and Horii, 1982; St. Privat d'Allier Re search Group, 1982; Hubert et al., 1984).
Among the more important results to date have been a direct experimental verification of the existence of submicrosecond fields and currents during return strokes and the general validity of Eq. (2.1) (Fieux et al. , 1978; Djibari et al, 1981; Hubert and Fieux, 1981). Waldteufel et al. (1980) also reported a curious case in which a triggered discharge originated everywhere in clear air.
The main benefit of the rocket triggering technique is that it can be used to cause lightning to strike a known place at a known time, thus enabling controlled experi

ments to be performed. The triggering wire guides the lightning current to a point where a variety of sensors can measure the physical properties of the discharge and its deleterious effects directly. All cameras and data-re cording equipment can be turned on and be fully opera tional just before the rocket is fired. In most locations, the total number of triggers is limited to a few tens of events per year by the frequency of overhead storms, but the number and quality of the measurements can be made quite high to compensate for the relatively few events.

PHYSICS OF LIGHTNING

39

CONCLUSION
We have seen that new experimental techniques now provide an opportunity to investigate many of the im portant questions that remain unanswered about the physics of lightning. Among the more important un knowns are the following:
How is lightning initiated within a cloud? Can the initiation of lightning be suppressed or con
trolled? What are the mechanisms of stepped- and dart-leader
propagation? What factors control the geometrical development of
lightning? What is the physics of the attachment process? What are the currents in return strokes? How rapidly do return-stroke currents change with
time? What physical processes control the propagation of
return strokes? What is the energy balance of the various lightning
processes? What physical phenomena occur during a cloud dis
charge? What are the characteristics of the currents in cloud
discharge processes? What processes generate HF and VHF radio noise in
lightning?
Recent spacecraft observations have shown that light ning may be present in the atmospheres of Jupiter, Ve nus, and Saturn, and the upcoming Galileo probe will carry a lightning detector to Jupiter (Lanzerotti et ah, 1983). Perhaps a study of lightning in atmospheres that are radically different from that of Earth will help us to better understand lightning on Earth and offer even more challenging questions for future work.
REFERENCES
Berger, K., R. B. Anderson, and H. Kroninger (1975). Parameters of lightning flashes, Electro 80, 23-37.
Brook, M., and T. Ogawa (1977). The cloud discharge, in Lightning, Vol. 1, R. H. Golde, ed., Academic, New York, pp. 191-230.
Djibari, B., J. Hamelin, C. Leteinturier, and J. Fontain (1981). Com parison between experimental measurements of the electromag netic field emitted by lightning and different theoretical models— Influence of the upward velocity of the return stroke, Proc. Int. Con}, on EMC. March 1981, Zurich.
Fieux, R., and P. Hubert (1976). Triggered lightning hazards, Nature 260, 188.
Fieux, R., C. Gary, and P. Hubert (1975). Artificially triggered light ning above land. Nature 257. 212-214.
Fieux, R., C. H. Gary, B. P. Hutzler, A. R. Eybert-Berard, P. L. Hubert, A. C. Meesters, P. H. Perroud, J. H. Hamelin, and J. M.

Person (1978). Research on artificially triggered lightning in France, IEEE Trans. Power Appar. Syst. PAS-97, 725-733. Garbagnati, E., F. Marinoni, and G. P. Lo Pipero (1981). Parameters of lightning currents— Interpretation of the results obtained in Italy, Proc. 16th Int. Con}, on Lightning Protection, July 1981, Szeged, Hungary. Guo, C., and E. P. Krider (1982). The optical and radiation field sig natures produced by lightning return strokes, J. Geophys. Res. 87, 8913-8922. Hayenga, C. O., and J. A. Warwick (1981). Two-dimensional interferometric positions of VHF lightning sources, J. Geophys. Res. 86, 7451-7462. Horii, K. (1982). Experiment of artificial lightning triggered with rockets, Mem. Fac. Eng. (Nagoya U.)34, 77-112. Hubert, P., and R. Fieux (1981). 2— Mesure des courants de foudre a la station d'etude de la foudre de St.-Privat-d"Allier, Rev. Gen. £/ecfr. 5, 344-349. Hubert, P., P. Laroche, A. Eybert-Berard, and L. Barret (1984). Trig gered lightning in New Mexico, /. Geophys. Res. 89, 2511-2521. Idone, V. P., and H. E. Orville (1982). Lightning return stroke veloci ties in the Thunderstorm Research International Program (TRIP), 7. Geophys. Res. 87, 4903-4915. Krider, E. P., and C. Guo (1983). The peak electromagnetic power radiated by lightning return strokes, /. Geophys. Res. 88, 84718474. Krider, E. P., R. C. Noggle, A. E. Pifer, and D. L. Vance (1980). Lightning direction-finding systems for forest-fire detection, Rull. Am. Meteorol. Soc. 61, 980-986. Lanzerotti, L. J., K. Rinnert, E. P. Krider, M. A. Uman, G. Dehmel, F. O. Gliem, and W. I. Axford (1983). Planetary lightning and lightning measurements on the Galileo Probe to Jupiter's atmo sphere, in Proceeding in Atmospheric Electricity,]. Latham and L. Ruhnke, eds., Deepak Publ. Co., Hampton, Va., pp. 408-410. Lin, Y. T., M. A. Uman,J. A. Tiller, R. D. Brantley, E. P. Krider, and C. D. Weidman (1979). Characterization of lightning return stroke electric and magnetic fields from simultaneous two-station mea surements,/. Geophys. Res. 84, 6307-6314. Master, M. J., M. A. Uman, Y. T. Lin, and R. B. Standler (1981). Calculations of lightning return stroke electric and magnetic fields above ground, /. Geophys. Res. 86, 12127-12132. Miyachi, I. and K. Horii (1982). Five years' experiences on artificially triggered lightning in Japan, 7th International Conference on Gas Discharges and Applications, London, pp. 468-471. Newman, M. M., J. R. Stahman, J. D. Robb, E. A. Lewis, S. G. Martin, and S. V. Zinn (1967). Triggered lightning strokes at very close range, /. Geophys. Res. 72, 4761-4764. Orville, R. E. (1968). A high-speed time-resolved spectroscopic study of the lightning return stroke, Pts. 1, 2, 3, /. Atmos. Sci. 25, 827-856. Proctor, D. E. (1971). A hyperbolic system for obtaining VHF radio pictures of lightning,/. Geophys. Res. 76, 1478-1489. Proctor, D. E. (1981). VHF radio pictures of cloud flashes, /. Geophys. Res. 86,4041-4071. Proctor, D. E. (1983). Lightning and precipitation in asmall multicel lular thunderstorm,/. Geophys. Res. 88, 5421-5440. Rustan, P. L., M. A. Uman, D. G. Childers, W. H. Beasley, and C. L. Lennon, (1980). Lightning source locations from VHF radiation data for a flash at Kennedy Space Center, /. Geophys. Res. 85, 48934903. Salanave, L. E. (1980). Lightning and Its Spectrum , Univ. of Arizona Press, Tucson. Schonland, B. F. J. (1964). The Flight of Thunderbolts, Clarendon Press, Oxford. St. Privat d'Allier Research Group (1982). Eight years of lightning ex periments at St. Privat d'Allier, Rev. Gen. Electr. 9. 561-582.

40

E. PHILIP KRIDER

Taylor, W. L. (1978). A VHF technique for space-time mapping of lightning discharge processes, /. Geophys. Res. 83,3575-3583.
Uman, M. A. (1969). Lightning, McGraw-Hill, New York. Uman, M. A, and E. P. Krider (1982). A review of natural lightning:
Experimental data and modeling, IEEE Trans. Electromagn. Compat.,EMC-24, 79-112. Uman, M. A., D. K. McLain, and E. P. Krider (1975). The electro magnetic radiation from a finite antenna, Am. J. Phys. 43, 33-38. Uman, M. A., M. J. Master, and E. P. Krider (1982). A comparison of lightning electromagnetic fields with the nuclear electromagnetic pulse in the frequency range lOMO7 Hz, IEEE Trans. Electromagn. Compat.,EMC-24, 410-416.

Waldteufel, P., P. Metzger, J. L. Aouley, P. Laroche, and P. Hubert (1980). Triggered lightning strokes originating in clear air, /. Geophys. Res. 83, 2861-2868.
Warwick, J. W., C. O. Hayenga, and J. W. Brosnahan (1979). Interferometric directions of lightning sources at 34 MHz, J. Geophys. Res. 84, 2457-2468.
VVeidman, C. D., and E. P. Krider (1978). The fine structure of light ning return stroke wave forms, J. Geophys. Res. 83, 6239-6247.
Weidman, C. D., and E. P. Krider (1980). Submicrosecond risetimes in lightning return-stroke fields, Geophys. Res. Lett. 7, 955-958. [See also C. D. VVeidman and E. P. Krider (1982), Correction, /. Geophys. Res. 87, 7351.]

Positive Cloud-to-Ground Lightning
3
W. DAVID RUST NOAA National Severe Storms Laboratory

INTRODUCTION
Of the two common types of lightning flash, cloud-toground (CG) and intracloud (IC), the CG flashes have historically received more attention and study. This is undoubtedly because they not only have more visible channels that lend themselves to quantitative observa tions but also because they are responsible for most death and damage caused by lightning. There are vari ous common names for several visually different fea tures of CG lightning, e.g., streak, forked, and ribbon. However, the parameters often used in the scientific lit erature to categorize CG lightning are the polarity of charge lowered, the magnitude of the current, and the direction of propagation of the initiating leader and/or ensuing return stroke. While most CG flashes transfer negative charge from the cloud to the ground, early doc umentation of flashes that lower positive charge from cloud to ground ( + CG) is found in Berger's classical study (1967) of lightning at Mount San Salvatore above Lake Lugano in Switzerland. Berger used instruments and photography to document strikes to towers on the mountain and found a minority of + CG flashes. All but one of the + CG flashes had an upward-propagating leader followed by a downward-moving return stroke, altogether opposite to the common negative CG flashes, consisting of downward-propagating leaders followed
41

by upward return strokes. Because these + CG flashes originated from tall towers, and apparently not even from the mountain peaks without tall structures, they have been termed "triggered" lightning, in contrast to those occurring naturally. They nearly always had only one return stroke. Berger's documentation of + CG flashes to Mount San Salvatore established their devas tating nature; they had larger peak currents, charge transfer, action integral, and duration than most nega tive flashes.
In general the instrumentation used to observe + CG flashes is the same as that for other lightning. Typical measurements include the electrostatic-field change for the entire flash, fast electric- and magnetic-radiationfield wave forms for the return stroke, optical tran sients, and thunder. Photographic and television re cordings are also often made. In addition, one type of modern automatic lightning-strike locating system is be ing used experimentally in several parts of the world to locate + CG flashes. However, its ability to distinguish between intracloud and + CG flashes and its resulting detection efficiency are still unknown.
RECENT AND ONGOING STUDIES
The recent and renewed interest in + CG flashes has been stimulated by observations of winter storms in Ja

42
pan, where there is a large percentage (although low total number) of CG flashes that are positive but not triggered by tall structures. These storms have low cloud bases and cloud tops at about 5 km and are much colder in their lower region than typical summer thunder storms. Although they produce fewer flashes, there have been several reports of strikes to aircraft. Whether the strikes to aircraft were from + CG flashes is unknown, but the prevalence of + CG flashes in these storms has led to questions concerning their correlation to storm structure and their relevance to aviation safety.
The existence of + CG flashes in Japan was first re ported by Japanese scientists, who have since conducted collaborative research programs with scientists from the United States and Sweden. The recently published results by Brook et al. (1982) include the best quantita tive information on charge transfer for naturally occur ring + CG flashes. From their multistation network of electrostatic field-change sensors, they calculated charge transfer and the magnitude of continuing cur rent. As in Berger's earlier work on + CG flashes trig gered by tall towers, the naturally occurring + CG flashes in Japan have large currents. While the largest negative flashes are comparable with + CG flashes, it appears that + CG flashes as a group tend to have signif icantly greater charge transfer and currents.
The combined data from seven Japanese winter storms show a remarkably good correlation between the percentage of + CG flashes per storm and the vertical shear in the horizontal wind in the cloud layer. This sug gests that if a storm has its upper positive charge dis placed horizontally from its lower negative charge, the production of + CG flashes is facilitated. This may well hold true for such storms that are relatively shallow and only mildly convective; preliminary results from other investigations indicate that the correlation may not be universal. Studies of this are in progress elsewhere.
The finding of naturally occurring + CG flashes in Japan inspired the search for these flashes in the highly sheared, large, and often severe springtime storms over the Great Plains of the United States. The occurrence of + CG flashes during the mature and later stages of se vere storms has been verified in the observational pro gram at the National Severe Storms Laboratory (NSSL) in Oklahoma (Rust et al. , 1981a) . Shown in Figure 3. 1 is a sketch of an isolated supercell thunderstorm, which contains an intense updraft, wind shear, turbulence, a large anvil, and a mesocyclone (rotation that can pro duce a tornado). Flashes to ground in the regions of heavy precipitation have always been observed to lower negative charge, while those from the upshear and downshear anvil near the main storm tower can lower either polarity of charge. Only a relatively few CG

W. DAVID RUST
flashes have been observed to emerge from beneath the mesocyclone wall cloud (i.e., the visible manifestation of a mesocyclone; see Figure 3.1), and two have been documented as positive. Rust et al. (1981b) also ob served and obtained electric-field change records for 16 + CG flashes in 30 minutes from a downshear anvil, well away from the main tower of a severe storm shortly after it produced a wall cloud. No other flashes to ground were seen from the anvil during that time. In most cases, + CG flashes appear visually to emanate from high in the storm. The + CG flashes can cluster in time and dominate CG activity for certain periods.
Positive CG flashes are observed both in isolated supercells and in squall-line storms. Preliminary analysis of two squall lines indicates that + CG flashes often oc cur on the back side (relative to squall-line movement) after the squall line has been in existence for several hours. Acoustic recordings of thunder have been ana lyzed for two of six + CG flashes that were detected within an 8-minute period in the back side of such a squall line. They show a significant number of acoustic sources, and thus channels, throughout a depth of 15 km (the freezing level was about 4 km above ground). Doppler radar data show that these + CG flashes were im bedded in the low radar reflectivities (less than 17 dBZ) associated with very light precipitation behind the squall line. Visual observations indicate that + CG flashes may propagate horizontally through tens of kilo meters along the back side of the squall line before com ing to ground. Relatively large horizontal extent is also apparent in observations of a few confirmed + CG flashes during the final stage in some of the smaller thunderstorms over the Rocky Mountains where usually there are no or only a few + CG flashes (Fuquay , 1982) .
FIGURE 3.1 Sketch of observed locations and polarities of CG flashes from severe thunderstorms. The spiral denotes the updraft re gion and rotation (mesocyclone). Only negative CG flashes have been observed in the intense precipitation. The horizontal scale of this storm sketch is greatly compressed. After Rust etal. (1981a).

POSITIVE CLOUD-TO-GROUND LIGHTNING

43

The apparent correlation between + CG flashes and storm development and severity suggested above are quite tentative. This is due to the paucity of + CG flash data, caused in part by the relative infrequency of + CG flashes and in part by the need for corrobative verifica tion of the occurrence of each + CG flash until simpler detection techniques are proven.
Nowhere is this more obvious than in the large data bases recently acquired with a few of the automatic strike-locating systems that have been modified but as yet unproved (as stated by the manufacturer, Lightning Location and Protection, Inc.) for detection of positive as well as negative ground flashes. If we had confidence in the data bases obtained with these systems, much could readily be learned, owing to the large numbers of supposed + CG flashes that have and continue to be re corded. Samples of results from these as yet unproven systems include the following:
1 . Orville et al. (1983) reported a case study of a cy clone that produced several convective cells that moved through their East Coast strike-locating network. They found that while only 4 percent of all flashes to ground in the storm are identified as + CG flashes, the percent age increased to 37 percent in the last hour of significant CG flash activity. The also reported observations of a higher percentage of + CG flashes in the later stages of other storms.
2. The NSSL strike-locating system has been used to study the diurnal variation of + CG flashes for summer time storms in Oklahoma, which tend to be less severe than storms in the spring. The fraction of CG flashes that were positive, averaged for 1 month, peaks about 2 hours later than the total CG flash activity.
3. Attempts have been made to ascertain if the + CG flashes observed in severe storms are related to storm severity or tornado occurrence. Two tornadic storms have been analyzed using the strike-locating system in Oklahoma. The ratio of + CG to all CG flashes appears to be greater before and during tornadoes than after ward. This result is preliminary, not only because of un certainties in the performance of the strike locating sys tem but also because only two tornadic storms have been analyzed. It also remains to be shown just what severe storm parameters are related to the production of + CG flashes. Some possibilities that are being examined are mesocyclone strength, updraft speed, shear, and precip itation structure.
PHYSICAL CHARACTERISTICS
Although a few of the characteristics of + CG flashes have been described above, it is worth considering what

we know in total about their characteristics, irrespective of the storm conditions in which they occur. There are two characteristics that appear consistently in all the re ported observations of + CG flashes: (1) the vast major ity have only a single return stroke, and (2) the return stroke is often followed by continuing current. A repre sentative electrostatic field change for a + CG flash is shown in Figure 3.2. Before the return stroke, there is usually lengthy preliminary activity, which averages about 0.25 sec but can be as long as about 0.8 sec. If incloud channels for + CG flashes are primarily horizon tal, as suggested by observations in several locations, and if progression speeds are 105 m/sec as typically ob served, then there may be large horizontal extent to many + CG flashes. Indeed, horizontal movement of + CG flashes before they come to ground has been de termined from analysis of multistation field-change data in Florida (Brook et al. , 1983) and is indicated also by visual observations of squall lines in Oklahoma.
The field change for the leader to ground has not yet been extensively studied; however, both multistation analysis for several flashes in Japan and photographic evidence for a few flashes in the Rocky Mountains and Oklahoma show that the leader propagates down from the cloud to the ground, in contrast to the initially up ward-moving, triggered flashes to Mount San Salvatore. Recent studies in Japan indicate that + CG flashes can be preceded by either a stepped or a nonstepped leader.
The return-stroke wave form (Figure 3.3) is similar in shape to that for negative flashes (Rust et al., 1981b), with a relatively slow initial ramp followed by a faster
R
I.Ot
2115:37.27
FIGURE 3.2 Typical electric- field change for + CG flash, recorded at 2115:37.270 CST on May 30. 1982. The distance to the flash was about 4 km. Time at the center of the scale (at 0.5 sec) is 2115:37.300 CST. Interval a is the preliminary breakdown and leader: b is the time from the return stroke through the end of the continuing current seen on the streak photograph in Figure 3.4: and c is a larger interval of possible continuing current or additional intracloud breakdown. The return stroke (see Figure 3.3) is labeled R.

44

W. DAVID RUST

i

i

i

i

i

0

25

50

79

100/ii

FIGURE 3.3 Electric-field wave form for return stroke of the + CG flash in Figure 3.2. The slow ramp, interval x, is followed in interval y by a faster transition to peak, which is typical of both negative and positive return strokes.

transition to peak. Thus far, the wave forms obtained in several widely separated locations are essentially the same. For 15 visually confirmed + CG flashes in Okla homa, the average zero-to-peak rise time is 6.9 nsee. In Florida, the average for three visually observed + CG flashes at distances of 20-40 km is about 4 /xsec, a value comparable with negative flashes in the same storm. In both locations, some + CG flashes have been observed with fast transition portions of the wave form having rise times of less than 1 /xsec. While there was no visual or photographic documentation, apparent + CG flashes in Sweden have yielded zero-to-peak times of 525 nsee for flashes at ranges of approximately 100 km; the mean zero-to-peak times for + CG flashes was re-

ported to be about twice that for negatives. The peak amplitude of the electrostatic-field change due to the return stroke itself averages about one tenth of the total change for the entire flash. This appears due to the large preliminary breakdown and continuing current, which dominate the field change.
Analysis of multistation electric-field change mea surements by Brook et al. (1982) reveal + CG flashes with continuing currents up to 10s A and positive charge
transfer to ground of up to several hundred coulombs. Nakahori et al. (1982) made direct measurements of currents in + CG flashes and found peak stroke currents of 31 kAand total charge transfer of 164 coulombs in one flash. The largest magnitudes of charge transfer are of ten more than 10 times greater than these for negative flashes to ground in summer storms.
The duration of continuing currents in + CG flashes has been reported to vary from a few milliseconds to about 250 msec. However, the longer durations were obtained from single-station field-change measure ments. The few streak-film and TV recordings of con tinuing current obtained thus far indicate that later por tions of the slow field change may not always be from continuing current in the channel to ground but may be additional intracloud activity. For example, the streak photograph in Figure 3.4 (for the field change in Figure 3.2) indicates a continuing current duration of 60 msec, but Figure 3.2 alone could be interpreted as indicating at least 200 msec of current flow. Either the luminosity decreased below the threshold for the film, or the cur-

FIGURE 3.4 Streak-film photograph of + CG flash recorded on May 30, 1982 (see Figures 3.2 and 3.3). Continuing current is evident from the smearing of luminosity. It is visible in the photograph for about 60 msec and occurs during interval b in Figure 3.2.

2115:37.270

POSITIVE CLOUD-TO-GROUND LIGHTNING

45

rent in the channel ceased and the remaining activity was intracloud.
PRACTICAL IMPLICATIONS
The destructive nature and practical importance of + CG flashes to the electrical power industry are at least partially documented. The multiline grounding and wire-strand fusing in high-voltage transmission lines cannot be explained by normal negative flashes to ground in Japan.
Because of the usual occurrence of continuing cur rent, + CG flashes may ignite a disproportionate num ber of fires, especially in grasslands and forests. Of the 75 + CG flashes reported in the Rocky Mountain study, all had field changes indicative of continuing current. The apparent pattern is for + CG flashes to strike out side the rainfall, further enhancing the likelihood of their starting a fire. One third of all storm days in a 3year period had + CG flashes within a 30-km radius of the U.S. Forest Service observing site at Missoula, Mon tana. Thus on any given day the fire-starting probability from + CG flashes in mountain thunderstorms appears significant.
The results to date of the observations in severe storms suggest that + CG flashes may be correlated with storm severity and tornado occurrence. If this hypothesis were shown to be true, it would enhance existing severestorm detection and warning capabilities. Additional observations to test this hypothesis are in progress.
An area of increasing importance and study is the ef fect of lightning on new-generation aircraft (see Chap ter 5, this volume) . The relevance of + CG flashes to this problem is currently unknown; however, the reports of strikes to aircraft in the winter storms in Japan suggest that + CG flashes were involved. There are several as pects of + CG flashes needing additional clarification to ascertain whether these flashes pose an unusual threat to aircraft. They include the presence of fast return-stroke rise times and large peak currents; the frequent occur rence of continuing current; and the apparent ten dency, especially in squall lines, for + CG flashes to propagate through large horizontal and vertical extents and to be in low radar reflectivity regions, which can appear "innocent" on radar.
Because + CG flashes are a small percentage of the total flashes that most storms produce, appear to have large spatial extent, and do not seem to have a unique electric-field change, it is difficult to verify + CG flashes. However, because of their potentially devastat ing nature and their possible link to storm severity, + CG flashes have become an important research topic in several parts of the world.

AREAS NEEDING ADDITIONAL RESEARCH
It is worth noting those general areas of research that are needed to increase our understanding of + CG flashes and our ability to determine and cope with their effects on important technologies. These research areas are listed below but not in order of priority.
1 . Measure electric- field changes and wave forms for a large number of confirmed + CG flashes to establish their typical characteristics with greater certainty.
2. Determine what storm types and environmental conditions are conducive to + CG flashes for storms throughout the year.
3. Relate the production of + CG flashes to storm evolution and structure, including also flash initiation and propagation characteristics.
4. Determine typical and extreme peak currents. 5. Evaluate the capabilities of automatic strike-lo cating systems in identifying + CG flashes, especially their detection efficiency and false identification rate. 6. Determine the significance of + CG flashes to avi ation, especially to new-generation aircraft (typified by composite structures and computer-controlled flight) . 7. Determine the importance of the threat posed by + CG flashes to power distribution systems, including whether they are the cause of the unexpected large num ber of faults on power lines in various parts of the United States.
REFERENCES
Bcrger, K. (1967). Novel observations of lightning discharges: Results of research on Mount San Salvatore, /. Franklin Inst. 283, 478-525.
Brook, M., M. Nikano, P. Krehbiel, and T. Takeuti (1982). The elec trical structure of the Hokuriku winter thunderstorms, /. Geophys. Res. 87, 1207-1215.
Brook, M., P. Krehbiel, D. MacLaughlan, T. Takeuti, and M. Nakano (1983). Positive ground stroke observations in Japanese and Florida storms, in Proceedings in Atmospheric Electricity, L. H. Ruhnke and J. Latham, eds., A. Deepak Publishing, Hampton, Va., pp. 365-369.
Fuquay, D. M. (1982). Positive cloud-to-ground lightning in summer thunderstorms,/. Geophys. Res. 87, 7131-7140.
Nakahori, K., T. Egawa, and H. Mitani (1982). Characteristics of winter lightning currents in Hokuriku district, manuscript 82WM205-3, IEEE Power Engineering Society 1982 Winter Meet ing (Jan. 31-Feb. 5, 1982, New York).
Orville, R. E., R. W. Henderson, and L. F. Bosart (1983). An East Coast lightning detection network, Bull. Am. Meteorol. Soc. 64, 1029-1037.
Rust, W. D., W. L. Taylor, D. R. MacGorman, and R. T. Arnold ( 1981 a) . Research on electrical properties of severe thunderstorms in the Great Plains, Bull. Am. Meteorol. Soc. 62, 1286-1293.
Rust, W. D., D. R. MacGorman, and R. T. Arnold (1981b). Positive cloud-to-ground lightning flashes in severe storms, Geophys. Res. Lett. 8, 791-794.

Acoustic Radiations from Lightning
4
ARTHUR A. FEW, JR. Rice University

ACOUSTIC SOURCES IN THUNDERSTORMS
Electric storms produce a variety of acoustic emis sions. The acoustic emissions can be broadly divided into two categories—those that are related to electric processes (i.e., they correlate with lightning) and those that either do not depend on cloud electricity or for which no correlations with electric changes have been observed. Only the first group will be discussed here (see Few, 1982; Georges, 1982, for reviews of nonelectrical acoustics).
Two types of acoustic emissions correlated with elec tric processes are thunder, which is produced by the rap idly heated lightning-discharge channel, and infrasonic emissions produced by electrostatic fields throughout the charged regions of the cloud. Thunder is probably the most common of all loud natural sounds, while other acoustic emissions are not ordinarily observed without special devices.
THUNDER— THE RADIATION FROM HOT CHANNELS
Spectrographic studies of lightning return strokes (Orville, 1968) show that this electric-discharge process heats the channel gases to a temperature in the 24,000 K range. At high temperatures the expansion speed of the
46

shock wave is roughly 3 x 103 m/sec and decreases rap idly as the shock wave expands; in comparison the mea sured speeds for various lightning-breakdown processes range from 104 to 108 m/sec (Uman, 1969; Weber et al. , 1982). Therefore, the electric breakdown process in a discharge event is completed in a given length of the channel before the hydrodynamic responses are fully or ganized. Other electric processes occur over longer pe riods (e.g., continuing currents) , but the energy input to the hot channel is strongly weighted toward the early breakdown processes when channel resistance is higher (Hill, 1971).
Shock-Wave Formation and Expansion
The starting point for developing a theory for a shockwave expansion to form thunder is the hot ( - 24,000 K), high-pressure ( > 106 Pa) channel left by the electric dis charge. Hill's (1971) computer simulation indicated that approximately 95 percent of the total channel en ergy is deposited within the first 20 /tsec with the peak electric power dissipation occurring at 2 jtsec; during the 20-/tsec period of electric energy input, the shock wave can only move approximately 5 cm. This simula tion may actually be slower than real lightning because Hill used a slower current rise time than indicated bv

ACOUSTIC RADIATIONS FROM LICHTNING

47

more recent measurements (Weidman and Krider, 1978).
The time-resolved spectra of return strokes (Orville, 1968) show the effective temperature dropping from - 30,000 to ~ 10,000 K in a period of 40 /tsec and the pressure of the luminous channel dropping to atmo spheric in this same time frame. During this period the shock wave can expand roughly 0.1 m. Even though channel luminosities and currents can continue for pe riods exceeding 100 ^sec, the processes that are impor tant to the generation of thunder occur very quickly (< 10 usee) and in a very confined volume (radius ~5 cm). The strong shock wave propagates outward be yond the luminous channel, which returns to atmo sphere pressure within 40 usee. The channel remnant cools slowly by conduction and radiation and becomes nonconducting at temperatures between 2000 and 4000 K perhaps 100 msec later (Uman and Voshall, 1968) .
Turning our attention now to the shock wave itself we can divide its history into three periods—strong shock, weak shock, and acoustic. The division between strong and weak can be related physically to the energy input to the channel; the weak-shock transition to acoustic is somewhat arbitrary. Calculations and measurements have shown that the radiated energy is on the order of 1 percent of the total channel energy (e.g., Uman, 1969; Krider and Guo, 1983), hence most of the available en ergy is in the form of internal heat energy behind the shock wave.
As the strong shock wave expands it must do thermo dynamic work (Pd V) on the surrounding fluid. The ex pected distance though which the strong shock wave can expand will be the distance at which all the internal thermal energy has been expended in doing the work of expansion. Few (1969) proposed that this distance, which he called the "relaxation radius," would be the appropriate scaling parameter for comparing different sources and different geometries. The expressions for the spherical, R„ and cylindrical, Rc, relaxation radii are

R, = (3£(/4irP0)1/3

(4.1)

and

Rc = (£y/TTPo)1'2,

(4.2)

where £, is the total energy for the spherical shock wave, Ei is the energy per unit length for the cylindrical shock wave, and P0 is the environmental atmospheric pres sure. Table 4.1 gives Rc over a range of values that have been suggested in the literature for £/. Nondimensional distances denoted by X may be defined for spherical problems by dividing by Rs and for cylindrical problems by dividing by flr.
Figure 4.1 shows the propagation of the strong shock

TABLE 4.1 Relaxation Radii (/?,) (in meters) for Different Energies per Unit Length (£/) of Cylindrical Shock Waves

Ei 104
2 x 104 5 x 104 105
2 x IP 5 x HP 10"

P„ = 100 kPa ( - surface)
0.18 0.25 0.40 0.56 0.80 1.26 1.78

Pa = 60 kPa P„ = 30 kPa ( - 4-km height) ( - 9-km height)

0.23

0.33

0.32

0.46

0.52

0.73

0.72

1.02

1.03

1.46

1.63

2.30

2.30

3.25

into the transition region (X ~ 1) and beyond into the weak-shock region. As the shock front passes the relaxa tion radius (X = 1) the central pressure falls below am bient pressure as postulated in the definition of the re laxation radius. The momentum gained by the gas during the expansion carries it beyond X = 1 and forces the central pressure to go momentarily below atmo spheric. At this point the now weak-shock pulse decou ples from the hot-channel remnant and propagates out ward. Figure 4.2 shows on a linear coordinate system the final output from Brode's (1955) numerical solution, the weak-shock pulse at a radius of X = 10.5.
Figure 4.3 shows Plooster's (1968) cylindrical shock wave near X = 1 with Brode's (1955) spherical shock wave. The effects of channel tortuosity will be discussed

Q.
a5

SPHERICAL SHOCK WAVE . CYLINDRICAL SHOCK WAVE

FIGURE 4. 1 The expansion of spherical and cylindrical shock waves from the strong-shock region into the weak-shock region. The radii of both spherical and cylindrical geometries have been nondimensionalized using the relaxation radii defined in Eqs. (4.1) and (4.2). The spherical shock wave is that of Brode (1956), and the cylindrical shock wave is from a similarity solution by Sakurai (1954).

48

ARTHUR A. FEW, JR.

P/Pn
FIGURE 4.2 The weak shock wave formed from the spherical strong shock wave. This is the final pressure profile computed by Brode (1956). For an energy input of 10» J/m (R, = 0.56 m for P„ = W Pa) this weak shock wave would be approximately 6 m from the lightning channel.
in greater detail later; for now we note that owing to tortuosity we cannot expect the shock wave to continue to perform as a cylindrical wave once it has propagated beyond a distance equal to the effective straight section of the channel that generated it. If the transition from cylindrical to spherical occurs near X = 1 as suggested by Few (1969), then the spherical weak-shock solutions of Brode provide a good means of estimating the wave shapes of lightning-caused acoustic pulses.
Figure 4.4 presents a graphical summary of the vari ous transitions that are thought to take place. The initial strong shock will behave cylindrically following the dashed line based on Plooster's (1968) computations; this must be the case for the line source regardless of the tortuosity because the high-speed internal waves (3 x 103 m/sec) will hydrodynamically adjust the shape of the channel during this phase. The transition from strong shock to weak shock occurs near X = 1, and the transi tions from cylindrical divergence to spherical diver gence will occur somewhere beyond X = 0.3 and proba bly beyond X = 1 depending on the particular geometry of the channel at this point. The family of lines labeled \ in Figure 4.4 represent transitions occurring at different points. \ is the effective length, L, of the cylindrical source divided by Rc (\ = LlRc); it is approximately equal to the value of X at which the transition to spheri cal divergence takes place.
Comparisons with Numerical Simulations and Experiments In the numerical solutions of Plooster (1971a, b) and Hill (1971) the energy inputs to the cylindrical problem

were computed as a function of time for specified cur rent wave shapes and channel resistance obtained from the computations in the numerical model. These model results predicted that the energy input to the lightning channel was an order of magnitude or more below the values obtained from electrostatic estimates or from other indirect measurements of lightning energy (Few, 1982). The major differences might be due to the as sumed current wave forms used in the models. The re cent data obtained with fast-response-time equipment yields current rise times for natural cloud-to-ground lightning in the 35-50 kA//xsec range (Weidman and Krider, 1978). These values are considered as represen tative of normal strokes; extraordinary strokes have been measured with current rise times in the 100-200 kA//xsec range. By way of comparison, Hill's (1971) cur rent rise time was 2.5 kA//xsec.
Laboratory simulations of lightning have been suc cessfully performed in a series of experiments conducted at Westinghouse Research Laboratories; these results provide us with our best quantitative information on thunder generation. In these tests a 6.4 x 106 V impulse generator was used to produce 4-m spark discharges in air (Uman et al., 1970). Circuit instrumentation al lowed the measurement of the spark-gap voltage and current from which the power deposition can be com puted. Calibrated microphones were used to measure the shock wave from the spark as a function of distance. The results of the research (Uman et al. , 1970) have been compared with the theory of Few (1969) and with other

2.2

/>

//

1.8

//

:

a.

•^Spherical Pressure Wave

////

o

|

Cylindrical Pressure Wave /j

i

o

//

1.4

// ✓ /

1

/s

If)

e i.o

i

.2

.8

LO 1.2

FIGURE 4.3 Comparison of spherical and cylindrical shock-wave shapes near X = 1 . These profiles are for the point-source, ideal-gas solutions of Brode (1955) and Plooster (1968). In the transition region of strong shock to weak shock, these wave shapes are nearly identical. From Few (1969) with permission of the American Geophysical Union.

ACOUSTIC RADIATIONS FROM LIGHTNING

49

possible interpretations (Plooster, 1971a). The data were found to be consistent with the theory developed by Few.
Figure 4.5 compares a measured spark-pressure pulse with the profile that is predicted from the theory; both represent conditions in the plane perpendicular to the spark channel . Figures 4 . 6 and 4 . 7 summarize the exten sive series of spark measurements. Figure 4.6 is in the same format as Figure 4.4. The center line passing through the scattered points and labeled L = 0.5 m cor responds (using the measured energy input of 5 x 103 J/m, which gives Rc = 0. 126 m) to x = 4 in Figure 4.4. The two boundary lines L = 6.25 cm and L = 4.0 m would correspond to x values 0.5 and 32. The lower bound is very close to the lower limit value of one third indicated in Figure 4.4. The upper bound of Figure 4.6 (x = 32) is too large to be depicted in Figure 4.4, where x = 4 is the last line shown.
The data points of Figure 4.6 corresponding to the larger x or L values could represent situations where the shock-wave expansion was following the cylindrical be havior over a long distance, hence large x- However, if the expansions were truly cylindrical to that extent, then

1000

-I—I—I I I I 1 1

— Spherical Divergence — Cylindrical Divergence

FIGURE 4.4 Line-source shock-wave expansion. The overpressure of the shock front is given for spherical (Brode, 1956) and cylindrical (Plooster. 1968) shock waves. Line sources must initially follow cylin drical behavior, but on expanding to distances of the same size as line irregularities they change to spherical expansion following curves simi lar to the depicted curves. From Few (1969) with permission of the American Geophysical Union.

Pressure Profile Spark(5XI03j/m) at 3meters
Predicted profile
TIME
FIGURE 4.5 Comparison of theory with a pressure wave from a long spark. The measured pressure wave from a long spark (Uman et ah, 1970) is compared with the predicted pressure from a section of a mesotortuous channel having the same energy per unit length, x is assumed to be 4/3. From Few (1969) with permission of the American Geophysical Union.
the length of the pulse would be longer, as required by the cylindrical-wave predictions. The data of Figure 4.7 indicate that this cannot be the case. The lengths of the positive- pressure pulses shown in Figure 4.7 are clearly not in the cylindrical regime; if anything, they tend to be even shorter than predicted by the spherical expan sion. (See also Figure 4.5.)
It is obvious from both the spark photographs and wave forms in Uman et al. (1970) that the spark is tortu ous and produces multiple pulses. They found that the wave shapes, more distant from the spark where pulsetransit times were most similar, showed evidence of an in-phase superposition of pulses; at closer range the pulses exhibited greater relative phase shifts and more multiplicity aspects. The in-phase superposition of spherical waves would reproduce the distributions shown in Figures 4.6 and 4.7. The pressure amplitude would be increased relative to a single pulse, but the wavelength would not be substantially affected.
The measured spark wave forms (Uman et ah, 1970) were systematically shorter than predicted by the the ory. As shown in Figure 4.5, the tail of the wave was compressed, and the data of Figure 4.6 indicate that the positive pulse was similarly shortened. This shortening could be due simply to an inadequacy in the numerical shock-wave model; we think, instead, that the differ ence results from the energy input being instantaneous in the one case (Brode, 1956) and of longer duration for the spark case. If energy, even in small quantities, con tinues to be input into the low-density channel core after the shock front has moved outward then the core will be kept at temperatures much higher than predicted by the theories, having an instantaneous energy input followed by expansion. Owing to the elevated sound speed associ ated with the higher core temperature the part of the

50

ARTHUR A. FEW, JR.

1. V Distance, meters FIGURE 4 .6 Shock-front overpressure as a function of distance from the spark. The dots represent data obtained with a piezoelectric micro phone; the crosses data obtained with a capacitor microphone. The total electric energy per unit length computed from measurement of the spark voltage and current is 5 x 103 J/m. Also shown are theoreti cal values for cylindrical and spherical shock waves. From Uman et al. (1970).
0. 7 I ' 1 1—I—l—r-i-r] 1 ' 1—I—- - - - |
FIGURE 4.7 Duration of positive part of the shock wave from the long spark. For the same data of Figure 4.6, we see here the length of the positive pressure pulse for the 5 x 10' J/m sparks at various dis tances. From Umanef al. (1970).

wave following the shock front will form and propagate outward faster than predicted by theory. We expect, therefore, that the elevated core temperature associated with sparks and lightning can reasonably produce the shortened wave forms.
The wave shape produced by the shock wave is re lated to the energy per unit length of the lightning flash; thunder is superposition of many such pulses from the lightning channel; hence, the power spectrum of the thunder, with simplifying assumptions, can be related to the energy per unit length of the channel (Few, 1969) . Other properties (tortuosity and attenuation) that influ ence the spectrum of thunder are discussed later.
The assumptions in this theory all affect the thunder spectrum in the same sense; the peak of the theoretical spectrum will occur at higher frequencies than the peak of the real thunder spectrum (Few, 1982). The light ning-channel energy that one estimates from the peak will therefore be an overestimate of the actual lightningchannel energy. Holmes et al. (1971) provided the most complete published thunder spectra to date; these spec tra show a lot of variation. Most of the spectra are con sistent with the qualitative expectations of thunder pro duced by multiple-stroke lightning, but a few of them exhibit very-low-frequency ( < 1 Hz) components that are dominant during portions of the record and appear to be totally inconsistent with the thunder-generation theory from the hot explosive channel. Dessler (1973), Bohannon et al. (1978), and Balachandran (1979) sug gested that these lower- frequency components might be electrostatic in origin; Holmes et al. (1971) also consid ered that this was a possible explanation.
Tortuosity and the Thunder Signature
With respect to the effects of lightning-channel tortu osity on the thunder signal there is almost unanimous agreement among researchers. Lightning channels are undeniably tortuous and are tortuous apparently on all scales (Few et al., 1970). For convenience in discussing channel tortuosity Few (1969) employed the terms microtortuosity, mesotortuosity, and macrotortuosity rel ative to the relaxtion radius of the lightning shock wave.
For a lightning channel having an internal energy of 10s J/m (see Table 4.1), Rc = 1/2 m. The microtortuous
features smaller than Rc, although optically resolvable, are probably not important to the shock wave as mea sured at a distance because the high-speed internal waves (3 x 103 m/sec) are capable of rearranging the distribution of internal energy along the channel while the shock remains in the strong-shock regime. At the mesotortuous scale ( — Rc) the outward propagating shock wave decouples from the irregular line source because

ACOUSTIC RADIATIONS FROM LIGHTNING

51

the acoustic waves from the extended line source can no longer catch up with the shock wave. Somewhere in this mesotortuous range the divergence of the shock waves makes the transition from cylindrical to spherical.
Whereas the mesotortuous channel segments are im portant in the formation and shaping of the individual pulses being emitted by the channel the macrotortuous segments are fundamental to the overall organization of the pulses and the amplitude modulation of the resulting thunder signature. Few (1974a) computed that 80 per cent of the acoustic energy from a short spark was con fined to within ± 30° of the plane perpendicular to the short line source. A macrotortuous segment of a light ning channel will direct the acoustic radiations from its constituent mesotortuous, pulse-emitting segments into a limited annular zone. An observer located in this zone (near the perpendicular plane bisecting the macrotortu ous segment) will perceive the group of pulses as a loud clap of thunder, whereas another observer outside the zone will perceive this same source as a lower-amplitude rumbling thunder. This relationship between claps, rumbles, and channel macrotortuosity has been con firmed by experiment (Few, 1970) and in computer sim ulations (Ribner and Roy, 1982).
Loud claps of thunder are produced, as mentioned above, near the perpendicular plane of macrotortuous channel segments; there are three contributary effects (Few, 1974a, 1975) to the formation of the thunder claps. The directed acoustic radiation pattern described above is one of the contributing factors, and this effect is distributed roughly between ± 30° of the plane.
A second effect, which occurs only very close to the plane, is the juxtaposition of several pulses in phase, which increases the pulse amplitude to a greater extent than would a random arrival of the same pulses.
The third effect contributing to thunder clap forma tion is simply the bunching in time of the pulses. In a given period of time more pulses will be received from a nearly perpendicular macrotortuous segment of chan nel than from an equally long segment that is perceived at a greater angle owing to the overall difference in the travel times of the composite pulses.
In this section we have examined the complex nature of the formation of individual pulses from hot lightning channels and how a tortuous line source arranges and directs the pulses to form a thunder signature. The re sulting thunder signature depends on (1) the number and energy of each rapid channel heating event (leaders and return strokes); (2) the tortuous and branched con figuration of the individual lightning channel; and (3) the relative position of the observer with respect to the lightning channel.
Perhaps the most convincing discussion of thunder

generation as described above comes not from analytical evidence but from research using sophisticated com puter models of thunder. Ribner and Roy (1982) synthe sized thunderlike acoustic signals utilizing computergenerated waves formed by the superposition of N wavelets from tortuous geometric sources. The resulting "thunder" is highly similar to natural thunder (see Fig ures 4.8 and 4.9). Where the computer models are used to simulate laboratory experiments, there is also close agreement.
PROPAGATION EFFECTS
Once generated, the acoustic pulses from the light ning channel must propagate for long distances through the atmosphere, which is a nonhomogeneous, aniso tropic, turbulent medium. Some of the propagation ef fects can be estimated by modeling the propagation us ing appropriate simplifying assumptions; however, other effects are too unpredictable to be reasonably modeled and must be considered in individual situa tions.
Three of the largest propagation effects—finite-am plitude propagation, attenuation by air, and thermal refraction—can be treated with appropriate models to account for average atmosphere effects. Reflections from the flat ground can also be easily treated. Once the horizontal wind structure between the source and the receiver are measured, the refractive effects of wind shear and improved transient times may also be calcu lated. Beyond these effects, elements such as vertical

TIME, l

41

M

FIGURE 4.8 Schematic depiction of the synthetic generation of thunder by computer by the superimposition (upper trace) of N wave lets from a tortuous line source (Ribner and Roy, 1982); the summed signal is shown on the lower trace.

52

ARTHUR A. FEW, JR.

(1982) used the Otterman theory to develop an expres sion for the lengthening of acoustic pulses generated by mesotortuous lightning-channel elements. The result for the length of the positive-pressure pulse at the ground, LK, is given by

-f-(v2 - V2) = -^-floLo"2]1o

FIGURE 4.9 Comparison of synthetic (upper trace) and real (lower trace) thunder signals (Ribnerand Roy, 1982).
winds, nonsteady storm-related horizontal winds, tur bulence, aerosol effects, and reflections from irregular terrain produce complications that must be either ig nored or examined on a case-by-case basis.
Finite-Amplitude Propagation
As large-amplitude acoustic waves propagate through air, theory predicts that the shape of the wave must evolve with time. A single pulse will evolve to the shape of an N wave (see, for example, the spark wave in Figure 4.5); further propagation of the wave produces a lengthening of this N wave. The best theoretical treat ment of this process for application to the thunder prob lem is the one developed by Otterman (1959). His for mulation addressed the lengthening of a Brode-type pulse, such as Figure 4.2, from an initial length (L0) at an initial altitude (H0) down to the surface; his treat ment differs from many others that do not include the change of ambient pressure (P0) with altitude. Few

(4.3) VRocosfl/ 2Hg
R0 is the distance from the channel to the front of the pulse at the initial state where the fractional overpres sure at the pulse front is Ilo = SP0l Pq. The angle 0 is measured between the acoustic ray path and vertical; y is the ratio of specific heats; and He is the atmosphere scale height.
Equation (4.3) provides the finite-amplitude stretch ing that should be applied to the waves predicted by strong-shock theory. Uman et al. (1970) demonstrated that pulse stretching occurred beyond Brode's final pres sure profile shown in Figure 4.2; we see this clearly in Figure 4.7. Few (1969) used linear propagation beyond the profile of Figure 4.2 to estimate the power spectrum of thunder but commented that nonlinear effects may be important. The need for application of nonlinear or finite-amplitude theory to the thunder signal has been voiced in a number of papers in addition to these men tioned above (e.g., Holmes et al., 1971; Few, 1975, 1982; Hill, 1977; Bass, 1980).
If the Brode pressure pulse (shown in Figure 4.4) is used as the initial condition for the finite-amplitude propagation effect, the following values for input to Eq. (4.3) areflo = 10.46fif, L0 = 0.53flc, and Ilo = 0.03. In addition, if y = 1.4 and HR = 8 x 103 m are used in Eq. (4.3), the following equation is obtained:
L.R = Re)I0.386 + 0.147 LIn ( V l—O.—46^Rccos0-)/
23 (4.4)
16 x 103
Equation (4.4) has been used to generate the values in Table 4.2. The relaxation radii (flc) cover the entire range of values for fic in Table 4. 1 . Three values for 6 are represented, as are three heights for the source. In gen eral, the finite-amplitude propagation causes a dou bling in the length of the positive pulse within the first kilometer, but beyond this range the wavelength re mains approximately constant. The theory developed by Otterman did not include attenuation of the signal; because attenuation reduces wave energy, which in turn

ACOUSTIC RADIATIONS FROM LIGHTNING

53

TABLE 4.2 Finite-Amplitude Stretching of a Positive Pulse (Length, L0) for a Range of Cylindrical Relaxation Radii (Rc),

Source Heights (tfp), and Angles (9), See Eg. (4.4)

«,(m)

0.20 0.40 0.60 0.80 1.00 1.50 2.00 2.50 3.00 3.50 0.11 0.21 0.32 0.42 0.53 0.80 1.06 1.33 1.59 1.86

9=0° Z,„(m), H() = 1 km L, (m), Ho = 4 km Lt(m),H„ = 8km

0.24

0.45

0.65

0.84

1.03

1.49

1.93

2.35

2.77

3.17

0.26 0.49 0.71 0.93 1.14 1.66 2.16 2.65 3.12 3.59

0.26 0.51 0.74 0.96 1.18 1.72 2.24 2.75 3.25 3.74

9 = 45° Lc (m), H„ = 1 km L,(m), H„ = 4 km L,(m),H„ = 8 km

0.24

0.46

0.67

0.87

1.06

1.54

1.99

2.44

2.87

3.30

0.26 0.50 0.73 0.96 1.18 1.71 2.22 2.73 3.22 3.71

0.27

0.52

0.76

0.99

1.22

1.77

2.31

2.83

3.35

3.85

9=60° L,, (m), H„ = 1 km L„(m), Hi, = 4 km Z,„(m),H„ = 8 km

0.25 0.47 0.69 0.89 1.10 1.59 2.06 2.52 2.98 3.42

0.27

0.51

0.75

0.98

1.21

1.76

2.29

2.81

3.32

3.82

0.28 0.53 0.77

1.01

1.25

1.81

2.37

2.91

3.44

3.97

reduces the wave stretching, this theory should be viewed as a maximum estimator of the pulse length.
The finite-amplitude propagation effect does, how ever, help to resolve the overestimate of lightning-chan nel energy made by acoustic power-spectra measure ments. Few (1969) noted that the thunder-spectrum method yielded a value for £/ that was an order of mag nitude greater than an optical measurement by Krider et al. (1968). By assuming a doubling in wavelength by the finite-amplitude propagation, the energy estimate is reduced by a factor of 4 , bringing the two measurements into a range of natural variations and measurement pre cision.
Attenuation
There are three processes on the molecular scale that attenuate the signal by actual energy dissipation; the wave energy is transferred to heat. Viscosity and heat conduction, called classical attenuation, represent the molecular diffusion of wave momentum and wave in ternal energy from the condensation to the rarifaction parts of the wave. The so-called molecular attenuation results from the transfer of part of the wave energy from the translational motion of molecules to their internal molecular rotational and vibrational energy during the condensation part of the wave and back out during the rarifaction part of the wave. The phase lag of the energy transfer relative to the wave causes some of the internal energy being retrieved from the molecules to appear at an inappropriate phase; thus it goes into heat rather than the wave. These three processes can be treated the oretically within a common framework (Kinsler and Frey, 1962; Pierce, 1981). The amplitude of a plane wave, 6 P, as a function of the distance, x, from the coor dinate origin is given by

bP = 6P0e-

(4.5)

where 5P0 is the wave amplitude at the origin. The coef ficient of attenuation, a, can be shown in the low-fre quency regime to be

(4.6) 2c
In Eq. (4.6), u is the angular frequency and t is the re laxation time (or e-folding time) for the molecular pro cess being considered; c is the speed of sound. The lowfrequency condition above assumes that wr < 1. The expressions for depend on the particular molecular pro cesses under consideration; it is important to note, how ever, that a is proportional to w2 for the assumed condi tions; hence, attenuation alters the spectral shape of the propagating signals.
For thunder at frequencies below 100 Hz it can be shown (Few, 1982) that the total attenuation is insignifi cant. However, for the many small branches having much lower energy than the main channel, the frequen cies will be much higher and attenuation is important. Because of lower initial acoustic energies, spherical di vergence, and attenuation it is unlikely that acoustics emitted by the smaller branches and channels can be easily detected over longer distances (see also Bass, 1980; Arnold, 1982).

Scattering and Aerosol Effects
The scattering of acoustic waves from the cloud parti cles is similar to the scattering of radar waves from the particles; both are strongly dependent on wavelength. The intensity of the scattered sound waves from a plane acoustic wave of wavelength, \, incident on a hard sta

54

ARTHUR A. FEW, JR.

tionary sphere of radius a is proportional to (ira2)(a/X)4; this is the same relationship that appears in the radar cross-section expression for these parameters. For thun der wavelengths ( ~ 1 m) and cloud particles ( ~ 10 "3 m) the ratio (a/X)4 is 10" 12. The cloud is, therefore, trans parent to low-frequency thunder just as it is to meterwavelength electromagnetic radiation, although insig nificant fractions of the radiation do get scattered.
There are, however, eddies in the same size range as low-frequency thunder wavelengths, and these fea tures, owing to small thermal changes and flow shears, produce a distortion of wave fronts and scattering-type effects. For the part of the turbulent spectrum having wavelengths smaller than the acoustic wavelengths of interest, the turbulence can be treated statistically by scattering theory. Larger-scale turbulence must be de scribed with geometric acoustics. For the low-frequency thunder, turbulent scattering will attenuate the highamplitude beamed parts of the thunder signal; this in creases the rumbles at the expense of claps.
In the first part of this subsection we discussed the cloud particles as sources of acoustic scattering; there are other and probably more important ways in which these aerosol components interact with the acoustic waves. First, the surface area of the cloud particles within a volume provides preferred sites for enhanced viscosity and heat conduction; hence, the presence of particles increases the classical attenuation coefficient. Another totally different process produces attenuation by changing the thermodynamic parameters associated with the acoustic wave over the surfaces of cloud parti cles; this changes the local vapor-to-liquid or vapor-tosolid conversion rates. For example, during the compressional part of the wave the air temperature is increased and the relative humidity is decreased relative to equilibrium; the droplets partially evaporate in re sponse and withdraw some energy from the wave to ac complish it. The opposite situation occurs during the ex pansion part of the wave. Because the phase-change energy is ideally 180 out of phase with the acoustic-wave energy' this process produces attenuation. Landau and Lifshitz (1959) included this effect in their "second vis cosity" term. This attenuation process differs from the other microscopic processes in that it can be effective at the lower frequencies. The magnitude of this effect plus the enhanced attenuation by viscous and heat conduc tion at the surface exceed that of particle-free air by a factor of 10 or greater depending on the type, size, and concentration of the cloud particles (Kinsler and Frey, 1962).
Finally, there is a mass-loading effect with respect to the cloud particles that must be considered. The ampli tude of the fluid displacement, f, produced by an acous

tic wave of pressure amplitude bp and angular fre quency 03 is (Kinsler and Frey, 1962)

bp
f=
p0co>

(4.7)

Using 50 Pa as a representative value of bp for thunder inside a cloud we find for a 100-Hz frequency that f = 100 nm . The part of the cloud particle population whose diameter is much smaller than this, say 10 /tm, should, owing to viscous drag, come into dynamic equilibrium with the wave flow. [Dessler (1973) computed the re sponse time for a ~ 10-/tm droplet to re-establish dy namic equilibrium with drag forces; only 10 " 3 sec is re quired.] These cloud particles, which participate in the wave motion, add their mass to the effective mass of the air; this effects both the speed of sound and the impedence of the medium. For higher-frequency waves, fewer cloud particles participate, so the effect is re duced; whereas lower-frequency waves include greater percentages of the population and are more strongly af fected. Clouds are, therefore, dispersive with respect to low-frequency waves. Also, the cloud boundary acts as a partial reflector of the low-frequency acoustic signals because of the impedence change at the boundary. As suming a total water content of order 5 g/m3, we esti mate that the order of magnitude of the effect on sound speed and impedence is 10 " 3; this is not large, but it may be detectable.
The cloud aerosols interact with the acoustic waves in three different ways depending on their size relative to the amplitude of air motion of the sound. The smallest fraction "ride with the wave"— altering the wave-prop agation parameters. The largest particles are stationary and act as scatterers of the acoustic waves. The particles in the middle range provide a transition scale for the above effects but are primarily responsible for enhanced viscous attenuation.
In summary, there are several processes that can ef fectively attenuate higher-frequency components of thunder; this is in support of the conclusions of the pre vious section. We have, in addition, found three pro cesses that affect the low-frequency components. Low frequencies can be attenuated by turbulent scattering and, in the cloud, by coupling wave energy to phase changes. We have also found that low frequencies inter act with the cloud population dynamically; as a result, cloud boundaries may act as partial reflectors and incloud propagation may be dispersive.

Refraction
There is a wide range of refractive effects in the envi ronment of thunderstorms. In the preceding section we

ACOUSTIC RADIATIONS FROM LIGHTNING

55

found that turbulence on the scale of the acoustic wave length and smaller could be treated with scattering the ory. Turbulence larger than acoustic wavelengths, up to and including storm-scale motions, should be describable by geometric acoustics or ray theory. To actually do this is impractical because it requires detailed informa tion (down to the turbulent scale) of temperature and velocity of the air everywhere along the path between the source and the observer. Since the thunder sources are widely distributed we would require complete knowledge of the storm environment down to the meter scale to trace accurately the path of an individual acous tic ray. These requirements can be relieved if we relax somewhat our expectations regarding the accuracy of our ray path. The three fluid properties that cause an acoustic ray to change its direction of propagation are the components of thermal gradient, velocity gradient, and velocity that are perpendicular to the direction of propagation. Beyond the overall thermal structure of the environment, which will be approximately adiabatic, we do not expect that the thermal perturbations due to turbulence will be systematic. In fact, the turbu lent thermal perturbations should be random with a zero average value; hence, an acoustic ray propagating through turbulence should not deviate markedly owing to thermal gradients associated with the turbulence from the path predicted by the overall thermal structure of the environment. Similarly, velocity and velocity gra dients should produce a zero net effect on the acoustic ray propagating through the turbulence.
This argument of compensating effects is not valid for large eddies whose dimensions are equal to or greater than the path length of the ray because the ray path is over a region containing a systematic component of the gradients associated with the large eddy. We can obtain a worst-case estimate of these effects by examining a horizontal ray propagating from a source at the center of an updraft of 30 m/sec through 2 km to the cloud boundary where the vertical velocity is assumed to be zero; we also assume a linear decrease in vertical veloc ity between the center and boundary. The ray will be "advected" by 90 m upward during this transit, which requires approximately 6 sec, while the direction of propagation of the ray will be rotated through 5 down ward (maximum angle = tan-1 A VIC). Owing to this rotation, which is a maximum computation, the "ap parent" source by straight-ray path would be 180 m above the real source. These two effects have been esti mated independently when, in fact, they are coupled and are to some extent compensatory; when we merely add them the result is an overestimate of the apparent source shift, which in this example is 270 m. If this worst case is the total error in propagation to the receiver at 5

km then this error represents 5 percent of the range; over the length in which it occurs, 2 km, it represents 13 per cent error.
Now we turn our attention to the large-scale refrac tion effects that can be incorporated in an atmospheric model that employs horizontal stratification. The two strongest refractive effects of the atmosphere—the ver tical thermal gradient and boundary-layer wind shears— fall into this category along with other winds and wind shears of less importance.
The nearly adiabatic thermal structure of the atmo sphere during thunderstorm conditions has been recog nized for a long time as a strong influence on thunder propagation (Fleagle, 1949). This thermal gradient is effective because it is spatially persistent and unavoid able. Even though the temperature in updrafts and downdrafts— inside and outside the cloud— may differ (sometimes significantly), the thermal gradients in all parts of the system will be near the adiabatic limit (or pseudoadiabatic in some cases) because of the vertical motion. Hence, the acoustic rays propagate in this strong thermal gradient throughout its existence.
We can employ a simplified version of ray theory to illustrate some of the consequences of this thermal struc ture. If we assume no wind, a constant lapse rate (r = - d Tldz), and ATI T0 « 1 (AT is the change in temper ature and Tq is the maximum temperature along the path) , then the ray path may be described as a segment of a parabola

P-45-*.

(4.8)

In Eq. (4.8), T0 also corresponds to the vertex of the parabola where the ray slope passes through zero and starts climbing, h and I are, respectively, the height above the vertex and the horizontal displacement from the vertex. To apply Eq. (4.8) to all rays it is necessary to ignore (mathematically) the presence of the ground be cause the vertices of rays reflecting from the ground are mathematically below ground. In addition, we must in other cases visualize rays extending backward beyond the source to locate their mathematical vertices.
If T0 is set equal to the surface temperature, a special acoustic ray that is tangent to the surface when it reaches the surface is defined; this is depicted in Figure 4. 10. This same ray is applicable to any source, such as S[, S2, or S3, that lies on this ray path. For the conditions assumed in this approximation it is not possible for rays from a point source to cross one another (except those that reflect from the surface) . The other acoustic rays emanating from S2 must pass over the point on the ground where the tangent ray makes contact; this is also true for rays reflecting from the surface inside the tan

56

ARTHUR A. FEW, JR.

FIGURE 4.10 Parabolic acoustic ray from sources S,, S», or S, tangent to the surface at P. This ray was generated utilizing Eq. (4.8) with T„ - 30°C and T = 9.8 K/km. Observ ers on the surface to the right of P cannot de tect sound from sources S|, S^, S3, or Sj; an observer at P can only detect sound originat ing on or above the parabolic ray shown.

40 Si

Distance In km

gent point. The shaded zone in Figure 4. 10 corresponds to a shadow zone that receives no sound from any point source on the tangent ray beyond the tangent point. Point sources below the tangent ray, such as source S4 in Figure 4. 10, have their tangent ray shifted to the left in this representation and similarly cannot be detected in the shadow zone. However, sources above the tangent ray, S5 for example, can be detected in some parts of the shadow zone.
For each observation point on the ground one can de fine a paraboloid of revolution about the vertical gener ated by the tangent ray through the observation point; the observer can only detect sounds originating above this parabolic surface. For this reason we usually hear only the thunder from the higher parts of the lightning channel unless we are close to the point of a ground strike. For evening storms, which can often be seen at long distances, it is common to observe copious light ning activity but hear no thunder at all; thermal refrac tion is the probable cause of this phenomenon. For T0 = 30°C, T = 9.8 K/km, and h = 5 km we find that I = 25 km; as noted by Fleagle (1949) thunder is seldom heard beyond 25 km. (See also the discussion in Ribner and Roy, 1982.)
Winds and wind shears also produce curved-ray paths but are more difficult to describe because they af fect the rays in a vectorial manner, whereas the temper ature was a scalar effect. If you are downwind of a source and the wind has positive vertical shear (duldz > 0), the rays will be curved downward by the shear; on the upwind side, the rays are curved upward. Wind shears are very strong close to the surface and can effec tively bend the acoustic rays that propagate nearly par allel to the surface. The combined effects of tempera ture gradients, winds, and wind shears can best be handled with a ray-tracing program on a computer. With such a program one can accurately trace ray paths through a multilevel atmosphere with many variations in the parameters; it is usually necessary in these pro grams to assume horizontal stratification of the atmo sphere. The accuracy of the ray tracing by these tech

niques can be very high, usually exceeding the accuracy with which temperature and wind profiles can be deter mined.
MEASUREMENTS AND APPLICATIONS
A number of the experimental and theoretical re search papers dealing with thunder generation have been discussed in earlier sections and will not be re peated here. In this section we describe additional results, techniques, and papers that deal with thunder measurements.
Propagation Effects Evaluation
The reader should have, at this point, an appreciation for the difficulty in quantitatively dealing with the propagation effects on both the spectral distribution of thunder and the amplitude of the signal. If, however, we are willing to forfeit the information content in the higher-frequency ( > 100 Hz) portion of the thunder sig nal, which is most strongly affected by propagation, we can recover some of the original acoustic properties from the low-frequency thunder signal.
If the peak in the original power spectrum of thunder is assumed to be below 100 Hz, then the co2-attenuation effects deplete the higher frequencies without shifting the position of the peak. Most spectral peaks of thunder tend to be around or below 50 Hz; therefore, this as sumption appears to be safe even with finite-amplitude stretching effects considered. Further assume that the spectra are not substantially altered by turbulent scat tering and cloud aerosols. To the extent that these as sumptions are valid, the finite-amplitude stretching can be removed from the thunder signal and its peak fre quency at the source can be estimated. This technique enables a rough estimate of the energy per unit length of the stroke to be made; the result is corrected for firstorder propagation effects. Holmes et al. (1971) found that the spectral peak overestimated the channel energy using Few's (1969) method; if corrected for stretching

ACOUSTIC RADIATIONS FROM LIGHTNING

57

these measurements are in closer agreement, with the exception of those events containing other lower-fre quency acoustic sources.
There are a number of experiments that could and should be done to evaluate the propagation effects. Us ing thunder as the acoustic source, several widely sepa rated arrays of microphones could compare signals from the same source at several distances. If carefully exe cuted this experiment could quantify some of the propa gation effects. Another approach would be to employ a combination of active and passive experiments such as point-source explosions inside clouds from either bal loons or rockets. This experiment provides an additional controllable factor that can yield more precise data; it also involves greater cost and hazard.
Acoustic Reconstruction of Lightning Channels
In the section on refraction we mentioned the utility of ray-tracing computer programs that could accurately calculate the curved path of an acoustic ray from its source to a receiver; the accuracy is limited to the preci sion with which we are able to define the atmosphere. An obvious application of thunder measurements is to invert this process; one measures thunder then traces it backward from the point of observation along the ap propriate ray to its position at the time of the flash. Few (1970) showed that by performing this reverse-ray prop agation for many sources in a thunder record it was pos sible to reconstruct in three dimensions the lightning channel producing the thunder signal. The sources in this case were defined by dividing the thunder record into short ( ~ 1/2 sec) intervals and associating the acous tics in a given time interval with a source on the channel.
Within each time interval the direction of propaga tion of the acoustic rays are found by cross correlating the signals recorded by an array of microphones. The position of the peak in the cross-correlation fraction gives the difference in time of arrival of the wave fronts at the microphones; from this and the geometry of the array, one calculates the direction of propagation. At least three noncollinear microphones are required. Close spacing of the microphones produces higher corre lations and shorter intervals thus more sources; how ever, the pointing accuracy of a small array is less than that of a large array. Based on experiences with several array shapes and sizes, 50 m2 has been adopted as the optimum by the Rice University Group (see Few, 1974a).
The reconstruction of lightning channels by ray trac ing was described by Few (1970) and Nakano (1973). A discussion of the accuracy and problems of the tech nique is given in Few and Teer (1974) in which acousti

cally reconstructed channels were found to agree closely with photographs of the channels below the clouds. The point is dramatically made in these comparisons that the visual part of the lightning channel is merely the "tip of the iceberg."
Nakano (1973) reconstructed, with only a few points per channel, 14 events from a single storm. Teer and Few (1974) reconstructed all events during an active pe riod of a thunderstorm cell. MacGorman et al. (1981) similarly performed whole-storm analyses by acoustic channel reconstruction and compared statistics from several different storm systems. Reconstructed lightning channels by ray tracing have been used to support other electric observations of thunderstorms at the Langmuir Laboratory by Weber et al. (1982) and Winn et al. (1978).
A second technique for reconstructing lightning channels has been developed that is called thunder rang ing. This technique was developed to provide a quick coarse view of channels (within minutes after lightning if necessary) as opposed to the ray-tracing technique, which is slow and time consuming. Thunder ranging requires thunder data from at least three noncollinear microphones separated distances on the order of kilome ters. Experience with cross-correlation analysis of thun der signals has shown that the signals become spatially incoherent at separations greater than 100 m owing to differences in perspective and propagation path. How ever, the envelope of the thunder signals and the gross features such as claps remain coherent for distances on the order of kilometers. As discussed earlier these gross features are produced by the large-scale tortuous sec tions of the lightning channel. Thunder ranging works as follows: (1) The investigator identifies features in the signals (such as claps) that are common to three thunder signatures on an oscillograph. (2) The time lags between the flash and the arrival of each thunder feature at each measurement point are determined. (3) The ranges to the lightning channel segments producing each thunder feature are computed. (4) The three ranges from the three separated observation points for each thunder fea ture define three spheres, which should have a unique point in space that is common to all of them. (5) The set of points gives the locations of the channel segments pro ducing the thunder features (see Few, 1974b; Uman et al., 1978).
The basic criticism of the thunder-ranging technique is that the selection of thunder features is the subjective judgment of the researcher; for many features the selec tion is unambiguous; other features, which are close to gether, may appear separated at one location and merged at another. The program developed by Bohannon (1978) included these uncertainties in the estima

58
tion of errors associated with such points. Most of the recent thunder research has used a combination of rang ing and ray tracing.
The whole-storm studies in which an extended series of channels are reconstructed have proven to be the most valuable use of thunder data to date. They define the volume of the cloud actually producing lightning, the evolution of the lightning-producing volume with time, and the relationship of individual channels with other cloud observations such as radar reflectivity and envi ronmental winds (Nakano, 1973; Few, 1974b; Teer and Few, 1974; Few et al., 1977, 1978; MacGorman and Few, 1978; MacGorman etal., 1981).
ELECTROSTATICALLY PRODUCED ACOUSTIC EMISSIONS
The concept of electrostatically produced acoustic waves from thunderclouds goes all the way back to the

ARTHUR A. FEW, JR.
writings of Benjamin Franklin in the eighteenth cen tury; Wilson (1920) provided a rough quantitative esti mate of the magnitude of the electrostatically produced pressure wave. McGehee (1964) and Dessler (1973) de veloped quantitative models for this phenomenon — McGehee for spherical symmetry and Dessler for spheri cal, cylindrical, and disk symmetries. The theory developed by Dessler is of particular importance be cause it made specific predictions regarding the direc tivity and shape of the wave. The predictions were sub sequently verified in part by Bohannon et al. (1977) and Balachandran (1979, 1983).
The charge in a thundercloud resides principally on the cloud drops and droplets. In a region of the cloud where the charge is concentrated producing an electric field E, the charged particles will experience an electric force, which is directed outward with respect to the charge center, in addition to the other forces expressed on them. These particles quickly (on the order of milli-

'i

!

-i_l

- fPS

I— ssec—i
FIGURE 4.11 Low-frequency acoustic pulse thought to have been generated by an electrostatic pressure change inside the cloud during a light ning flash. The higher-frequency signals from thunder have been removed from this record. From Balachandran (1979) with permission of the American Geophysical Union.

ACOUSTIC RADIATIONS FROM LIGHTNING

59

seconds) come into dynamic equilibrium where the hydrodynamic drag force associated with their motion is balanced by the sum of all the externally expressed forces. When the electric field is quickly reduced by a lightning flash the cloud particles readjust to a new dy namic equilibrium. The change in the hydrodynamic drag force requires a change in the pressure distributions surrounding all the charged cloud particles; hence, the pressure in the volume continuing the cloud particles is altered by the sudden reduction of the electric field. Since the electric force from a charge concentration is outward, the pressure inside the charged volume will be slightly lower than the surrounding air. When £ is re duced by the lightning flash the charged volume pro duces a slight implosion; this radiates a negative wave. Few (1982) derived a general expression for the internal pressure gradient produced by the electrostatic force; when integrated the result is

PE = P0 - (n + 1) fo(£°2" £2) .

(4.9)

that the positive component of the wave is not de scribed. Recently, Few (1984) suggested that the diabatic heating of the air in the charged volume by posi tive streamers may be the source of the positive pulse.
Colgate and McKee (1969) described theoretically an electrostatic pressure pulse using this same mechanism but applied to a volume of charged air surrounding a stepped leader. This particular signature has not been experimentally verified because it has the regular thun der signal, which is 300 times more energetic, superim posed on it.
ACKNOWLEDGMENT
The author's research into the acoustic radiations from lightning has been supported under various grants and contracts from the Meteorology Program, Division of Atmospheric Sciences, National Science Foundation, and the Atmospheric Sciences Program, Office of Naval Research; their support is gratefully acknowledged.

In Eq. (4.9) the parameter n takes the value 0 for plane geometry, 1 for cylindrical geometry, and 2 for spherical geometry; P0 and E0 are the values at the edge of the charged volume.
The amplitude of this pressure signal is related to the electric field, the wavelength to the thickness of the charged region, and the directivity of the wave to the geometry to the source (Dessler, 1973). If the theory can be quantitatively verified, the signal can be used to determine remotely internal cloud electric parameters.
The experimental search for electrostatic pressure waves has been difficult. The wave is low frequency ( — 1 Hz), small amplitude ( — 1 Pa), and buried in large background pressure variations produced by wind, tur bulence, and thunder. Prior to Dessler's prediction of beaming, one wondered why the signal was not more frequently seen in thunder measurements. Holmes et al. ( 197 1) measured a low-frequency component in a few of their power spectra of thunder but found these compo nents completely missing in others. Dessler showed that that signal would be beamed for cylindrical and disk geometry; the disk case would require that the detectors be placed directly underneath the charged volume for observation. This relationship has been observed by Bohannon etal. (1977) and by Balachandran (1979, 1983).
The electrostatic pressure wave predicted by the the ory discussed above is a negative pulse. The measured acoustic signature thought to be the verification of the prediction actually exhibits a positive pulse followed by a negative pulse (see Figure 4.11). The negative pulse appears to fit the theory, but the theory is deficient in

REFERENCES
Arnold, R. T. (1982). Storm acoustics, in Instruments and Techniques for Thunderstorm Observation and Analysis, E. Kessler, ed.. U.S. Department of Commerce, Washington, D.C., pp. 99-116.
Balachandran, N. K. (1979). Infrasonic signals from thunder, /. Geophys. Res. 84, 1735-1745.
Balachandran, N. K. (1983). Acoustic and electric signals from light ning,/. Geophys. Res. 88, 3879-3884.
Bass, H. E. (1980). The propagation of thunder through the atmo sphere,/. Acoust. Soc. Am. 67, 1959-1966.
Bohannon, J. L. (1978). Infrasonic pulses from thunderstorms, M.S. thesis, Rice Univ., Houston, Tex.
Bohannon, J. L., A. A. Few, and A. J. Dessler (1977). Detection of infrasonic pulses from thunderclouds, Geophys. Res. Lett. 4, 49-52.
Brode, H. L. (1955). Numerical solutions of spherical blast waves, /. Appl. Phys. 26, 766.
Brode, H. L. (1956). The blast wave in air resulting from a high tem perature, high pressure sphere of air. Rand Corp. Res. Memoran dum RM-1825-AEC.
Colgate, S. A., and C. McKee (1969). Electrostatic sound in clouds and lightning,/. Geophys. Res. 74, 5379-5389.
Dessler, A. J. (1973). Infrasonic thunder, /. Geophys. Res. 78. 18891896.
Few, A. A. (1969). Power spectrum of thunder, /. Geophys. Res. 74, 6926-6934.
Few, A. A. (1970). Lightning channel reconstruction from thunder measurements./. Geophys. Res. "5.7517-7523.
Few, A. A. (1974a). Thunder signatures. EOS 5.5. 508-514. Few, A. A. (1974b). Lightning sources in severe thunderstorms, in
Conference on Cloud Physics (Preprint volume), American Mete orological Society. Boston. Mass.. pp. 387-390. Few. A. A. (1975). Thunder, Sri. Am. 233(1). 80-90. Few, A. A. (1982). Acoustic radiations from lightning, in Handbook of Atmospherics. Vol. 2, H. Volland. ed.. CRC Press, Inc.. Boca Ra ton, Fla., pp. 257-289. Few, A. A. (1984). Lightning-associated infrasonic acoustic sources, in

60
Preprints: VII International Conference of Atmospheric Electric ity. American Meteorological Society, Boston, Mass., pp. 484-486. Few, A. A., and T. L. Teer (1974). The accuracy of acoustic recon structions of lightning channels, /. Ceophys. Res. 79,5007-5011. Few, A. A., H. B. Garrett, M. A. Uman, and L. E. Salanave (1970). Comments on letter by W. W. Troutman, "Numerical calculation of the pressure pulse from a lightning stroke," /. Ceophys. Res. 75, 4192-4195. Few, A. A., T. L. Teer, and D. R. MacGorman (1977). Advances in a decade of thunder research, in Electrical Processes in Atmospheres, II. Dolezelak and R. Reiter, eds., Steinkopff, Darmstadt, pp. 628632. Few, A. A., D. R. MacGorman, and J. L. Bohannon(1978). Thunder cloud charge distributions, inferences from the intracloud structure of lightning channels, in Conference on Cloud Physics and Atmo spheric Electricity, American Meteorological Society, Boston, Mass., pp. 591-596. Fleagle, R. G. (1949). The audibility of thunder, J. Acoust. Soc. Am. 21.411. Georges, T. M. (1982). Infrasound from thunderstorms, in Instrumentsand Techniquesfor Thunderstorm Observation and Analysis, E. Kessler, ed., U.S. Department of Commerce, Washington, D.C., pp. 117-133. Hill, R. D. (1971). Channel heating in return-stroke lightning, /. Ceophys. Res. 76, 637-645. Hill, R. D. (1977). Thunder, in Lightning, R. H. Golde, ed., Aca demic Press, New York, pp. 385-408. Holmes, C. R., M. Brook, P. Krehbiel, and R. A. McCrory (1971). On the power spectrum and mechanism of thunder, /. Ceophys. Res. 76,2106-2115. Kinsler, L. E., and A. R. Frey (1962). Fundamentals of Acoustics, 2nd ed., Wiley, New York, 523 pp. Krider, E. P., and C. Guo (1983). The peak electromagnetic power radiated by lightning return strokes, /. Ceophys. Res. 88, 84718474. Krider, E. P., G. A. Dawson, and M. A. Uman (1968). Peak power and energy dissipation in a single-stroke lightning flash, /. Ceophys. Res. 73, 3335-3339. Landau. L. D., and E. M. Lifshitz (1959). Fluid Mechanics, Pergamon Press, London, 536 pp. MacGorman, D. R., and A. A. Few (1978). Correlations between ra dar reflectivity contours and lightning channels for a Colorado storm on 25 July 1972, Conference on Cloud Physics and Atmo spheric Electricity, American Meteorological Society, Boston, Mass., pp. 597-600. MacGorman, D. R., A. A. Few, and T. L. Teer (1981). Layered light ning activity, 7. Ceophys. Res. 86, 9900-9910.

ARTHUR A. FEW, JR.
McGehee, R. M. (1964). The influence of thunderstorm space charges on pressure,/. Ceophys. Res. 69, 1033-1035.
Nakano, M. (1973). Lightning channel determined by thunder, Proc. Res. Inst. Atmospherics (Nagoya Univ.) 20, 1-7.
Orville, R. E. (1968). A high-speed time-resolved spectroscopic study of the lightning return stroke, /. Atmos. Sci. 25, 827-856.
Otterman, J. (1959). Finite-amplitude propagation effect on shockwave travel times from explosions at high altitudes, /. Acoust. Soc. Am. 3i, 470-747.
Pierce, A. D. (1981). Acoustics an Introduction to Its Physical Princi ples and Applications, McGraw-Hill, New York, 642 pp.
Plooster, M. N. (1968). Shock waves from line sources, NCAR-TN-37. National Center for Atmospheric Research, Boulder, Colo., 83 pp.
Plooster, M. N. (1971a). Numerical simulation of spark discharges in air, Phys. Fluid 14, 2111-2123.
Plooster, M. N. (1971b). Numerical model of the return stroke of the lightning discharge, Phys. Fluids 14, 2124-2133.
Ribner, H. S., and D. Roy (1982). Acoustics of thunder: Aquasilinear model for tortuous lightning, J. Acoust. Soc. Am. 72, 1911-1925.
Sakurai, A. (1954). On the propagation and structure of the blast wave, 2, /. Phys. Soc. Japan 9, 256-266.
Teer, T. L., and A. A. Few (1974). Horizontal lightning, /. Ceophys. Res. 79,3436-3441.
Uman, M. A. (1969). Lightning, McGraw-Hill, New York. Uman, M. A., and R. E. Voshall (1968). Time interval between light
ning strokes and the initiation of dart leaders, J. Geophys. Res. 73, 497-506. Uman, M. A., A. H. Cookson, and J. B. Moreland (1970). Shock wave from a four-meter spark, /. Appl. Phys. 41, 3148-3155. Uman, M. A., W. H. Beasley, J. A. Tiller, Y.-T. Lin, E. P. Krider, C. D. Weidman, P. R. Krehbiel, M. Brook, A. A. Few, Jr., J. L. Bohannon, C. L. Lennon, H. A. Poehler, W. Jafferis, J. R. Gluck, and J . R. Nicholson (1978) . An unusual lightning flash at Kennedy Space Center, Science 201, 9-16. Weber, M. E., H. J. Christian, A. A. Few, and M. F. Stewart (1982). A thundercloud electric field sounding: Charge distribution and lightning,/. Geophys. Res. 87, 7158-7169. Weidman, C. D., and E. P. Krider (1978). The fine structure of light ning return stroke wave forms, /. Geophys. Res. 83, 6239-6247. Wilson, C. T. R. (1920). Investigations on lightning discharges and on the electric field of thunderstorms, Phil. Trans. R. Soc. London. Ser. A 221, 73-115. Winn, W. P., C. B. Moore, C. R. Holmes, and L. G. Byerly (1978). Thunderstorm on July 16, 1975, over Langmuir Laboratory: A case study, /. Geophys. Res. 83, 3079-3092.

Application of Advances in Lightning Research to Lightning Protection
5
MARTIN A. UMAN University of Florida

INTRODUCTION
Significant advances in lightning protection have been made during the last decade. These advances have been a result of progress in two general areas of lightning research: (1) lightning phenomenology, including the technology for determining real-time strike locations, and (2) lightning physics, particularly the characteris tics of return stroke currents and electromagnetic fields (see Krider, Chapter 2, this volume, for a description of the return-stroke phase of a lightning flash, as well as of the other salient events that make up the flash).
(1) By phenomenology, we mean those characteristics of thunderstorms that are associated with numbers of lightning events, as opposed to the physical properties of the individual events. A phenomenological parameter of particular interest is the average lightning flash den sity, that is, the number of lightnings per square kilome ter per year (other units are possible) as a function of location. This parameter represents the starting point for almost all lightning protection designs (for example, the lightning overvoltage protection of utility power lines) because the number of lightning failures per year for which a system is designed is directly proportional to the number of ground flashes per unit area per year. Real-time identification of phenomenological parame ters such as the total number of lightning events per

storm and the lightning flashing rate is now possible with newly developed detection equipment. This equip ment also makes possible real-time decisions on utility system repair and repair preparation, early warning and detection of lightning-caused forest fires, and a va riety of other warning functions in situations that allow protective action to be taken, such as launches at the NASA Kennedy Space Center.
(2) When an object (e.g., aircraft, building, power line, or person) is struck directly by lightning, or is ex posed to the intense electromagnetic fields of a nearby flash, the potentially deleterious currents and voltages that appear in the object are determined by the physical characteristics of the lightning currents and fields and by the electric characteristics of the object that is struck. For example, it is thought that, to a first approximation, the voltages that are induced in electronics within an airborne metal aircraft that is struck by lightning are indirectly initiated by the fastest part of the current rate of rise. This fast change in current induces resonant os cillations on the metallic exterior of the aircraft (like a pestle striking a bell) that are then coupled inside the aircraft via holes or apertures, such as windows, in the conducting metal skin. Lightning protection is cur rently of considerable concern for the latest generation of military and commercial aircraft that operate with low-voltage computer circuits and have lightweight ep

61

62

MARTIN A. UMAN

oxy surfaces (potential apertures) replacing the moreconventional conducting metal. For these types of air craft and other similar advanced systems, the microelectronic components used are often more easily damaged by lightning-induced voltages and the shield ing against the intrusion of those voltages is often less adequate than is the case for more conventional systems.
In the following three sections, we examine in more detail the recent and widespread use of lightning-detec tion techniques for protection; those properties of light ning that cause damage, the mechanisms of lightning damage, and new methods of protection; and some re maining questions that research can answer to facilitate additional improvements in lightning protection.
APPLICATIONS OF NEW LIGHTNINGDETECTION TECHNIQUES TO PROTECTION
In 1983 the Electric Power Research Institute (EPRI), the research arm of United States power utilities, funded a long-term study of lightning flash density in the United States for the purpose of making possible bet ter lightning protection design for power lines. The EPRI research is being carried out using lightning-locat ing technology recently developed through basic re search (Krider et al., 1980). For the initial part of the study a network of automatic lightning direction find ing (DF) stations called the East Coast Network and op erated by the State University of New York at Albany (SUNYA) (Orville et a/., 1983) is being used. Future flash density studies can be expected to involve addi tional portions of the United States and perhaps Can ada. As is clear from Figure 5. 1 in which the East Coast Network is evident, over three quarters of the area of the United States and Canada is covered by DFs, a develop ment that has occurred since 1976. In addition, light ning-locating systems of the DF type developed in the United States have been installed in Australia, Norway, Sweden, Mexico, South Africa, Japan, Hong Kong, and the People's Republic of China during the same time pe riod.
The primary user of lightning-location data in the United States at present is the Bureau of Land Manage ment (BLM) , which is responsible for the majority of the DFs in the western United States and Alaska (Figure 5.1). The BLM and the Forest Services of most Cana dian provinces utilize the time and location of lightning storms to determine when and where to look for forest fires. Early detection of these fires results in consider able savings in natural resources and in the cost of fight ing the fires. BLM data are also disseminated in real time to all National Weather Service Offices in the west ern region via AFOS, to the National Severe Storms

Forecast Center in Kansas City, to Vandenburg Air Force Base, and to Nellis Air Force Base. Data from the SUNYA East Coast Network are currently being dis played in real time at the FAA Washington Air Route Traffic Control Center (ARTCC) in Leesburg, Vir ginia, the National Weather Service Forecast Office in Albany, New York, and Langley Air Force Base in Hampton, Virginia.
In addition to applications-oriented research, opera tional forest fire management, and Weather Service storm warning, the newly developed lightning-locating equipment is used to warn of the approach of storms in a variety of practical applications where protective action can be taken. Examples range from power utility com panies (e.g., Tampa Electric Company, China Power of Hong Kong) to missile launches (e.g., Kennedy Space Center, Vandenburg AFB) to sensitive military installa tion (e.g., Buckley Air National Guard Base, Colorado, Cudjoe Key AF Station, Florida) . In addition, lightning maps from these lightning locating systems are becom ing widely shown on TV weather shows, as they are of ten more meaningful to the typical viewer than the more conventional radar displays. An example of a 1-day lightning map from the Tampa Electric Company (Peckham etal., 1984) is shown in Figure 5.2.
AMELIORATION OF LIGHTNING DAMAGE
Mechanisms of Lightning Damage
The amount and type of lightning damage an object suffers is due to both the characteristics of the lightning discharge and the properties of the object. The physical characteristics of lightning of most interest are the cur rents and electromagnetic fields, particularly those from the return stroke since these are usually the largest: hence protection against the return stroke will usually protect against the currents and fields from other light ning processes.
Four properties of the return stroke current can be considered important in producing damage: (1) the peak current, (2) the maximum rate of change of cur rent, (3) the integral of the current over time (i.e., the charge transferred), and (4) the integral of the current squared over time, the so-called action integral. Let us examine each of these properties and the type of damage that it can produce.
For objects that present a resistive impedance, such as a ground rod driven into the Earth, a long power line, or a tree, the peak voltage on the object will be propor tional to the peak current. For example, a 50,000-A cur rent injected into a 400-fl power line produces a line voltage of 20,000,000 V. Such large voltages lead to

APPLICATION OF RESEARCH TO LIGHTNING PROTECTION
FIGURE 5.1 A map showing the location (dots) of lightning direction finding (DF) stations in place in summer 1984. Connected circles around each DF indicate area of lightning coverage. The area of coverage along the East Coast is the SUNYA East Coast Network.

64

MARTIN A. UMAN

SO

100 150 200 250 300 350

AUGUST 8, 1979 13:00 TO 16:00 (EDST)

FIGURE 5.2 A map of the cloud-to-ground lightning strikes in the Tampa Bay area for August 8, 1979. Individual storms have been cir cled. The two DF locations in the Tampa Electric Company's light ning location system are identified. Map scale is in thousands of feet.

electric discharges from the struck object to the ground through the air or through insulating materials. Such flashovers can, for example, short-circuit a power sys tem or kill people that are standing close to the object that is struck. An example of discharges across the ground caused by the high voltage on a struck golfcourse green marker is shown in Figure 5.3. The mag netic forces produced by the peak lightning currents are large and can crush metal tubes and pull wires from walls.
For objects that have an inductive impedance, such as wires in an electronic system, the peak voltage will be proportional to the maximum rate of change of the lightning current ( V = L dildt). For example, if 1 m of wire has an inductance L of 10 " 7 H and dildt = 1010 A/ sec, 1000 V is generated across the wire. Voltages of this level often cause damage to solid-state electronic de vices.
The heating or burn through of metal sheets such as airplane wings or metal roofs is, to first approximation, proportional to the lightning charge transferred (aver age current times time). Generally, large charge trans fers are due to long-duration (tenths of a second to sec

onds) lightning currents in the 100- to 1000-A range rather than to peak currents that have a relatively short duration. An example of a hole burned in an aircraft skin by lightning is shown in Figure 5.4, and some infor mation on hole size versus charge transferred is given in Figure 5.5. A typical lightning transfers 20 to 30 C and extreme lightnings hundreds of coulombs, but, fortu nately, the lightning does not often stay attached to one place on an aircraft in flight for the duration of that transfer.
The heating of many objects and the explosion of in sulators is, to first approximation, due to the value of the action integral. In the case of wires, the action integral represents the heat that is generated by the resistive im pedance of the wire. Some data on wire temperature rise for typical lightning action integrals is given in Figure 5.6. About 1 percent of negative strokes to ground have action integrals exceeding 106. About 5 percent of posi tive strokes is thought to exceed 107. In the case of a tree, this heat vaporizes the internal moisture of the wood, and the resultant steam pressure causes an explosive fracture. An example of typical tree damage from light ning is shown in Figure 5.7.
Two properties of the electromagnetic fields are suffi cient to describe most of the important damage effects: the peak value of the field and the maximum rate of rise to this peak. For certain types of antennas or metal ex posed to the lightning field, the peak voltage on the metal is proportional to the peak field. These antennas are commonly referred to as capacitively coupled. For other antennas, such as a loop of wire in an electronic circuit or an underground communication cable, the peak voltage is proportional to the maximum rate of change of the field.
New Results on Lightning Characteristics
Chapter 2 (this volume) by Krider describes the re cent findings on lightning current and field characteris tics. Much of this work has application to protection. A major step forward has been made in identifying the maximum rates of change of currents and fields, and it should be noted that these are now thought to be at least 10 times larger than was believed to be the case a decade ago. These recent results have important implications for the design of protection against damage that is caused by fast rates of change of currents and fields.
Chapter 3 (this volume) by Rust discusses positive lightning, recently identified and only partially charac terized. Positive lightning apparently produces very large peak currents, charge transfers, and action inte grals, much larger than the usual negative lightning. The Japanese report that their power systems are dis-

FIGURE 5.4 A lightning hole burned in the wing tip of a Boeing 707 (Uman. 1971).

66

MARTIN A. UMAN

rupted by a large fraction of the positive lightning strike, whereas only a small fraction of negative light nings have this effect (Nakahori etal., 1982) .

0 40 80 120 160 200 240 280 LIGHTNING FLASH CHARGE (COULOMBS)
FIGURE 5.5 Area of holes melted through aluminum and titanium of various thicknesses by lightning charge (Fisher and Plumer, 1977).

Protection Techniques
There are two general types of lightning protection: (1) diversion and shielding and (2) limiting of currents and voltages. On a residential or commercial building, for example, the diversion of lightning currents to ground via a standard system of lightning rods, down leads, and grounds is sufficient to protect the building structure itself and to decrease by imperfect shielding potentially harmful effects to electronic equipment in side. An example of a diversion and shielding system is shown in Figure 5.8.
More complete protection of electronic equipment must include limiting of currents and voltages induced by the direct strike to the structure or by traveling waves into the structure on electric power, communication, or other wires connected to the outside world. The design of the current- and voltage-limiting system is obviously dependent on an understanding of the wave shapes of the deleterious signals that are to be controlled; and this in turn requires a knowledge of the lightning character-

APPLICATION OF RESEARCH TO LIGHTNING PROTECTION

67

FIGURE 5.7 Typical damage to a tree due to a direct lightning strike (Uman, 1971).

istics and how the properties of the system under consid eration change these characteristics. Once such a deter mination is made, three general types of current- and voltage-limiting devices can be used for electronic or power systems: (1) voltage crowbar devices that reduce the voltage difference effectively to zero and short cir cuit the current to ground (the carbon block and gas tube arrestors used by the telephone company are good examples of crowbar devices) ; (2) voltage clamps such as recently developed solid-state metal oxide varistors (MOVs) or Zener diodes, which do not allow the voltage to exceed a given value; and (3) electric filters that re flect or absorb the higher and generally more damaging frequencies in the lightning transient. Frequently, all three of these forms of protection are used together in a

coordinated way. Examples of some of these protective devices are shown in Figure 5.9.
In recent years, a systematic approach has been de veloped that allows an optimal lightning protection sys tem to be designed for most structures. This new tech nique is called "topological shielding," and it uses both diversion and shielding and the limiting of currents and voltages discussed above. The technique consists of nest ing shields and "grounding" each shield to the one en closing it. All incoming wires are connected to the out side of each successive shield by a transient protective device, and therefore, at each successively inner shield, the voltage and power levels to be protected against are reduced. In Figure 5.10, we illustrate the principles of topological shielding. In Figure 5.10a, the equivalent

68

MARTIN A. UMAN

[jSround Terminal

/WW///

(o)

(b)

FIGURE 5.8 A standard lightning protection system for a small structure.

Metal Sheet Transient Protactlon >
Metal ahleld for underground eervlce

FIGURE 5.9 Examples of typical lightning protective devices: a and b are sealed spark gaps (crowbar devices): c is a solid-state metal oxide varistor; and d is a solid-state Zener diode. The diameter of c is about 1 inch.

FIGURE 5.10 A diagram illustrating the principles of topological shielding, a. Equivalent electric circuit for a grounded building served by power lines and a communications tower, b. External view of building after topological shielding, c. Schematic of the topological shielding.

circuit is shown for the grounding of a building associ ated with a communications tower. Figure 5. 10b shows an external view of the building after topological shield ing, and Figure 5. 10c shows a schematic of the topologi cal shielding technique.
FUTURE RESEARCH NEEDED FOR IMPROVEMENTS IN PROTECTION
The detailed physics of how lightning strikes a struc ture, power line, or aircraft is still not well understood. The approaching lightning leader is not influenced by

the object to be struck until it is perhaps a few tens to hundreds of meters away. At that time, an upwardmoving spark leaves the object to be struck eventually and similar sparks may also leave nearby objects. The upward-moving spark connects to the downward-mov ing leader attaching the leader to ground. (See Krider, Chapter 2, this volume, for a discussion of the attach ment process.) When this process is better understood through basic research, we should be able to determine with higher probability what will and what will not be struck and to provide better lightning protection ac cordingly. For example, the positioning of overhead

APPLICATION OF RESEARCH TO LICHTNING PROTECTION

69

ground wires above transmission lines should be able to be optimized.
More information is needed about the character of lightning currents, particularly those in processes other than return strokes. Is there, for example, an upper limit on the maximum rate of change of current? We need more data on positive lightning to be able to character ize all aspects of it in a statistical way. Only then can it be taken account of properly in protection design.
Much work needs to be done on the interaction of lightning currents and fields with objects like aircraft. For example, how are aircraft resonances affected by channel attachment? Computer models are now being developed with which to study these problems even in the presence of nonlinear discharge properties.
CONCLUSIONS
Basic research over the last decade has made possible impressive improvements in lightning protection. Those related to lightning detection and to the specification of current and electromagnetic-field wave shapes have

been discussed in this chapter. As with all research, new discoveries raise new questions. With the present inter est in lightning among scientists, due partly to recent successes and partly to the important unsolved prob lems, we can expect continued progress in lightning pro tection during the next decade.
REFERENCES
Fisher, F. A., and J. A. Plumer (1977). Lightning protection of air craft, NASA Reference Publication 1008.
Krider, E. P., R. C. Noggle, A. E. Pifer, and D. L. Vance (1980). Lightning direction-finding systems for forest fire detection. Bull. Am. Meteorol. Soc. 61, 980-986.
Nakahori, K., T. Ogawa, and H. Mitani (1982). Characteristics of winter lightning currents in Hokuriku District, IEEE Tram. PAS101, 4407-4412.
Orville, R. E., R. W. Henderson, and L. F. Bosart (1983). An east coast lightning detection network, Bull. Am. Meteorol. Soc. 64, 1029-1037.
Peckham, D. W., M. A. Uman, C. E. Wilcox, Jr. (1984). Lightning phenomenology in the Tampa Bay area,/. Geophys. Res. 89, 1178911805.
Uman, M. A. (1971). Understanding Lightning, BEK, Pittsburgh, Pa.

The Role of Lightning in the Chemistry of the Atmosphere

6

WILLIAM L. CHAMEIDES Georgia Institute of Technology

ABSTRACT
The high temperatures in and around the discharge tube of a lightning stroke cause the dissocia tion of the major atmospheric constituents N2, 02, C02, and H2 and the formation of trace species such as NO, CO, N2, OH, N. O, and H. As this cylinder of hot air cools, the levels of these trace species drop. However, if the cooling is sufficiently rapid the concentrations of these trace species can be "frozen-in" at levels significantly above their ambient, thermochemical equilib rium abundances, thereby leading to a net source of these gases to the background atmosphere. It is estimated that about 3 tg of N yr" ' as NO are produced in the present-day atmosphere by lightning through this process. Other gases produced by lightning are CO and N20 in the Earth's atmosphere; HCN in the Earth's prebiological atmosphere: CO, NO, and 02 in the cytherian atmosphere; and CO, N2, and a variety of hydrocarbons in the jovian atmosphere. The major uncertainty in quantifying the role of lightning in the chemistry of the terrestrial atmosphere, as well as that of other planetary atmospheres, arises from the lack of accurate statistics on the energy and frequency of lightning. The role of coronal discharges in the chemistry of clouds also needs to be investigated.

INTRODUCTION
In addition to the spectacular visual and aural effects that accompany a lightning flash, intense chemical re actions occur, which, on a relatively short time scale, can radically alter the chemical composition of the air in and around the discharge tube and, on longer time scales, can ultimately affect the composition of the at mosphere as a whole. The short-term chemical changes associated with lightning have been well documented by spectroscopic studies of lightning strokes (cf., Salanave, 1961; Uman, 1969). For instance, the strong
70

emissions from neutral atomic nitrogen (N 1), singly io nized atomic nitrogen (N 11), neutral atomic oxygen (O 1), and singly ionized atomic oxygen (O n) typically observed from the hot core of discharges, are indicative of the widespread dissociation of atmospheric N2 and 02 and the subsequent ionization of their atomic daugh ters. Other prominent spectroscopic features are the emission lines from CN and H, species arising from the dissociation of CO2 and H2O.
For the most part the large changes in the chemical composition of the air in and around the discharge tube can be related to the rapid variations in temperature in

THE ROLE OF LIGHTNING IN THE CHEMISTRY OF THE ATMOSPHERE

71

this region. The lightning bolt and associated shock wave produce a cylinder of very hot air within which chemical reactions between the atmospheric gases pro ceed rapidly to bring the mixture into thermochemical equilibrium. Immediately after the energy deposition, the temperature in the discharge tube approaches 30.000 K and the gas is a completely ionized plasma. As the gas cools by hydrodynamic expansion and turbulent mixing, the equilibrium composition of the gas changes from a plasma to a mixture of neutral atoms such as N and O and then to a mixture of molecular species and ultimately as the temperatures return to ambient to a mixture of N2, O2, H2O, and CO2 much like the back ground composition of the atmosphere. This variation in the equilibrium composition of air as a function of temperature is illustrated in Figure 6.1; note that as the temperatures fall below 5000 K the equilibrium shifts from N, O, H, and CO to NO, OH, and CO and then to N2,O2, H2O,andCO2.
If the gas around the lightning discharge was always to remain in thermochemical equilibrium, the net effect of lightning on the atmospheric composition would be negligible; once the temperature of the gas returned to its ambient level, its composition would be essentially the same as that of the background atmosphere's, and thus there would be no net production or destruction of atmospheric chemical species by lightning discharges. On the other hand, it is well known that laboratory sparks can have significant effects on the composition of air; most notable is the fixation of atmospheric nitrogen (N2) by sparks to produce nitric oxide (NO) . Given the basic equivalence between laboratory sparks and light ning discharges it would seem reasonable to expect that lightning also affects the composition of air. In fact, the knowledge that NO is produced by laboratory sparks led von Liebig to propose in 1827 that the NO "3 typically observed in rainwater arises from the fixation of atmo spheric N2 by lightning discharges. This nineteenth cen tury hypothesis of von Liebig's has only recently been qualitatively confirmed by direct observations of en hanced levels of NO and NO2 in and around active thun derclouds (Noxon, 1976, 1978; Davis and Chameides, 1984) and in the vicinity of a cloud-to-ground lightning flash (Drapchoefa/., 1983).
The identification of the mechanisms responsible for the net production of trace species such as NO by light ning and the quantification of their source rates on a global scale define the current frontier in the field of the chemistry of atmospheric lightning and will, therefore, be the major subject of this review. The discussion be gins by focusing on the production by lightning of atmo spheric NO, a species of special interest because of its central role in the photochemistry of the atmosphere

>
1000 2000 3000 4000 5000
t[-k]
FIGURE 6.1 Equilibrium volume mixing ratios of selected atmo spheric species as a function of temperatures in heated tropospheric air.
(cf., Crutzen, 1983). Following the discussion of NO production by lightning, a more general presentation will be given of the production of other trace species in both the present and the prebiological, terrestrial atmo sphere, as well as in other planetary atmospheres. A brief discussion is then presented on the possible effect of electrical discharges on the chemistry of cloudwater and the generation of acids in precipitation. Finally, a brief outline of the needs for future work in this area is pre sented.

72

NO PRODUCTION BY LIGHTNING

Similar to the Z'elovich mechanism for the fixation of nitrogen in explosions (Z'elovich and Raizer, 1966), the production of NO in lightning discharges is believed to be driven by high-temperature chemical reactions within a rapidly cooling parcel of air; the rapid cooling causes NO levels above its thermochemical abundance to be "frozen" into the gas. A simple physical analogy to this chemical production mechanism is that of dropping a bead through a column of rapidly cooling water in a gravitational field. Because the bead wants to minimize its potential energy with respect to the gravitational field, the bead will tend to fall to the bottom of the wa ter column. If, however, the water were to cool so rap idly that it froze before the bead reached the bottom of the column, the bead would be frozen in the column at a position of higher potential energy and would be pre vented from reaching its energetically preferred posi tion at the bottom of the column.
In the case of NO production in lightning, the high temperatures in and surrounding the discharge channel result in a series of chemical reactions that both produce and destroy NO. NO production is initiated by the ther mal dissociation of O2.

O + O (Reaction 6.1)

followed by the production of NO via the reaction chain

O + N2

NO + N (Reaction 6.2)

and

N + O2

NO + O. (Reaction 6.3)

In competition with these NO-producing reactions are

NO + N

N2 + O (Reaction 6.4)

and

NO + O

*- N + O2, (Reaction 6.5)

which convert NO back to N2 and O2 as well as

NO -*—*- N + O, (Reaction 6.6)

the thermal dissociation of NO itself, and

NO + NO

*" N2O + O, (Reaction 6.7)

the formation of N2O from NO. The equilibrium NO concentration, /l^0, is the NO

level at which NO-producing and NO-destroying reac tions are in balance. As illustrated in Figure 6.2, /°0 is a strong function of temperature. As the temperature rises above 1000 K the dissociation of N2 and O2 causes an increase in the NO equilibrium level. At about 4000 K,

WILLIAM L. CHAMEIDES f^Q peaks at a value approaching 10 percent. For higher temperature, N and O atoms become increasingly more stable relative to NO (See Figure 6.2) and/'^0 decreases.
Thus if NO were always to maintain thermochemical equilibrium, its concentration would reach a maximum when the temperature in and around the discharge tube was — 4000 K and would then decrease to a negligibly small value as the heated air cooled to ambient tem peratures. However, similar to the equilibrium NO concentration, the time, rNO, required to establish ther mochemical equilibrium for NO also varies with tem perature. As illustrated in Figure 6.2, this time becomes increasingly longer as temperature decreases because the reactions acting to establish equilibrium become slower. (In this figure, rNO was calculated by summing the loss frequencies for NO due to Reactions 6.4-6.7.) Whereas only a few microseconds are required for NO to equilibrate at 4000 K, equilibrium requires millisec onds at 2500 K, a second at 2000 K, and approximately 103 years at 1000 K. Hence, as the air cools, a tempera ture is eventually reached at which the rates of reaction
2000 3000 4000 SOOO 6000 TEMPERATURE, *K
FIGURE 6.2 The NO equilibrium volume mixing ratio/0, repre sented by the solid curve, and the NO chemical lifetime tno, repre sented by the dashed curve, as a function of temperatures in heated tropospheric air. (After Borucki and Chameides, 1984.)

THE ROLE OF LIGHTNING IN THE CHEMISTRY OF THE ATMOSPHERE

73

become too slow to keep NO in equilibrium. Instead of falling to the thermochemical equilibrium concentra tion of the ambient temperature, a higher NO level be comes frozen into the gas. This higher concentration, which corresponds to the NO equilibrium level at the temperature at which the NO concentration departs from equilibrium, is called the "freeze-out" tempera ture.
The freeze-out temperature of NO, TF, is approxi mately determined by the relationship

TT(Tp) = TNO(TF),

(6.1)

where tt is the characteristic cooling time of the heated air. When T > TF, then rNO < tt and the chemical reac tions are sufficiently rapid to keep NO in the thermo chemical equilibrium. However, for T < TF, tNo > rT and chemical reactions are too slow to adjust to the rap idly decreasing T; NO, therefore, freezes out with a mix ing ratio /No(^f) , Although a lower abundance of NO is favored thermodynamically at low T, the kinetics are too slow for readjustment. P, the net yield of NO pro duced by this process, is then approximated by
P(NO) =/?,o(TF)M(rF)£o-1moleculesJ-1, (6.2)
where M is the number of molecules per meter heated to, or above, TF in the region where NO is being pro duced for a discharge energy of E0 (in units of J/m). Thus it is necessary to determine values for TF and M that, when combined with the results of Figure 6.2, will al low an estimate of P(NO) fromEq. (6.2).
Once P(NO) is obtained, the global rate of NO pro duction by lightning, </>(NO), can be estimated from

</>(NO) = P(NO) . R
(14g/mole)(1Q-12tg/g)(3.16 x 107sec/yr) (6.3)
(6.02 x 1023 molecules/mole)
in units of teragrams (tg) (i.e., 10 "12g or 106 metric tons) of N per year, where R is the number of joules dissipated globally by lightning per second. Because of the current interest in developing global budgets for the flow of fixed nitrogen and nitrogen oxides through the atmo sphere, reasonably accurate estimates for <£(NO) are de sirable. A brief discussion of how the parameters needed to solve </>(NO) are calculated is presented below.

Estimate ofP(NO)
Following the approach of Borucki and Chameides (1984), we infer values for TF and M that are needed to

calculate P(NO) in Eq. (6.2) from the laboratory study of linear discharge channels by Picone et al. (1981) . This study indicated that lightning-like discharge channels cooled with a rj of about 2.5 x 1 0 " 3 sec . For this choice of TT, TF and/No(T/) can be estimated from Figure 6.2 to be about 2660 K and 0.029, respectively. Furthermore, using the result of Picone et al. (1981) that in spark dis charges 1 J of energy is required to heat each 1 cm3 of air to a temperature of 3000 K and assuming that the gas cools from 3000 K to the freeze-out temperature of 2660 K by mixing with the ambient atmosphere, it can be inferred that
M(2660K)/£0 = 3.2 x 1019molecules/J. (6.4)
Substituting the above values for/No(Tj.) and M(TF)IE0 into Eq. (6.2),
P(NO) = (0.29)(3.2 x 1018)
= 9.2 x 1016molecules/J. (6.5)
A comparison of the above-estimated NO yield with those of previous investigators is presented in Table 6. 1 and indicates a rather good agreement with a wide vari ety of theoretical calculations, laboratory spark experi ments, and atmospheric measurements. The largest dis crepancy appears to be with the NO yield attributed to Drapcho et al. (1983). The yield of Drapcho et al. was based on their observation of a sudden increase in NO and NO2 levels in the vicinity of a cloud-to-ground dis charge; given the many assumptions necessary to infer a yield from this observation the disparity between the yield of Drapcho et al. and the others in Table 6. 1 is not very surprising.

The Global Dissipation Rate, R
The rate at which energy is dissipated by lightning globally can be expressed as a function of two other pa rameters, i.e.,

R = EFF,

(6.6)

where EF is the average number of joules dissipated per lightning flash and F is the number of lightning flashes occurring globally per second. Borucki and Chameides (1984) recently examined the existing data base on light ning flashes to estimate these parameters. Combining optical and electrical measurements of the energy of a single stroke, observations of the number of strokes per flash, as well as measurements of the distribution of en ergy among the first and subsequent strokes, EF was esti mated to be about 4 x 108 J/flash with a factor of 2.5 uncertainty. From satelliteborne optical detection sys

74

WILLIAM L. CHAMEIDES

TABLE 6. 1 Estimates of NO Yield from Lightning Discharge

P(NO) (molecules/J)

Investigator

A. This work B. Theoretical calculation
C. Laboratory spark experiment
D. Atmospheric measurement

(9 ± 2) x 1016
3 x 1016» (3-7.5) x 1016 (4-6) x 1016» 80 x lO16* 16 x 10'6
(8-17) x 10" (6 ± 1) 1016 (5 ± 2) x 1016 (2 ± 0.5) x 1016 (20-30) x 10""" (25-2500) x 1016»

Based on the calculations of Borucki and Chameides (1984)
Tuck (1976) Chameides eta/. (1977) Griffing(1977) Hill et al. (1980) H ill et al . , as corrected by
Borucki and Chameides (1984) Chameides (1979) Chameides etal. (1977) Levine etal. (1981) Peyrous and Lapeyre (1982) Noxon (1976) Drapcho et al. (1983)

"The NO yields obtained by these investigators were expressed as molecules/flash. These yields were converted to units of molecules/Joule by
assuming Er = 4 x lO^/flash. fcDerived from Hill etal. (1980) by dividing their NO yield (6 x 1025 molecules/flash) by their energy per flash [(1.5 x W]lm)(5 x MPm/flash)].

tems, a value of 100 flashes/sec was assigned to F by Borucki and Chameides (1984) with an uncertainty fac tor of 25 percent. Substituting these parameters into Eq. (6.6), R was thus estimated to be about 4 x 1010Wwith a possible range of (1.3 to 12) x 1010W.
The Global NO Production Rate, <i>(NO)
The above estimates for R and P(NO) can be com bined in Eq. (6.3) to yield a global NO production rate of ~2.5tgof N/yr. However, it should be noted that this number is highly uncertain; Borucki and Chameides (1984) in a similar analysis arrived at a possible range in tfi(NO) from 0.8 to 8 tg of N/yr. By far the largest source of uncertainty arises from the uncertainty in £F, the en ergy dissipated by a lightning flash. A comparison be tween the estimate for the global fixation rate calculated here and those of previous investigators is presented in Table 6.2. For the most part our result is consistent with, although somewhat smaller than, the other esti mates.
Biological processes fix atmospheric N2 at a rate of about 200 tg of N/yr and anthropogenic fixation (pri marily due to the synthesis of fertilizers) occurs at a rate of about 60 tg of N/yr (Burns and Hardy, 1975). Thus it would appear that, at present, lightning is responsible for at most a few percent of the Earth's total nitrogen fixation. On the other hand, lightning appears to repre sent one of, if not the, major natural source of NOx to the atmosphere. The other natural sources of atmospheric NOt include stratospheric oxidation of N2O at a rate of 0.6 tg of N/yr (Levy et al., 1980); oxidation of NH3, which is not well known but could be important (Mc-

TABLE 6.2 Estimates of the Amount of Nitrogen Fixed by Lightning

Investigator

Nitrogen Fixed per Year (tg)°

Tuck (1976) Chameides rt a/. (1977) Chameides (1979) Dawson (1980) Hill etal. (1980) Levine etal. (1981) Kowalczyck and Bauer (1982) Ehhalt and Drummond (1982) Peyrous and Lapeyre (1982) Logan (1983) Drapcho et al. (1983) Present result [based on
calculations of Borucki and Chameides (1984)]

4.2 30 to 40 35 to 90 3 4.4
1.8 5.7 5 9 S 30 Best Estimate: 2.6 Range: 0.8 to 8

"l tg = 1012 g = 106 metric tons.

Connell, 1973); and NO emissions from soils as a result of microbial activity at a rate of (1 to 10) tg of N/yr (Galbally and Roy, 1978; Lipschultz et al., 1981). As noted earlier, qualitative confirmation of the impor tance of lightning as a natural source of atmospheric NOx has been obtained from a variety of NO and NO2 measurements (Noxon, 1976, 1978; Drapcho et al., 1983; Davis and Chameides, 1984), which reveal anom alously high concentrations of NO and NO2 in air within and above clouds in remote regions of the globe. In re gions strongly affected by anthropogenic activities, however, this natural NO source is swamped by NO production from the burning of fossil fuel and biomass

THE ROLE OF LIGHTNING IN THE CHEMISTRY OF THE ATMOSPHERE

75

at a combined rate of about 30 tg of N/yr (Crutzen, 1983) . It is for this reason that NOx levels in urban areas are some 104 to 105 times higher than in remote regions of the marine atmosphere (McFarland et al., 1979) and in conjunction with the anthropogenic release of nonmethane hydrocarbons is the cause of photochemical smog and related air pollution problems. Nevertheless, it is interesting to note that even the extremely low levels of NOx that are characteristic of the remote troposphere are believed to have a significant effect on the chemistry of the atmosphere, catalyzing the photochemical pro duction of tropospheric O3 and enhancing OH levels (Logan et al., 1981; Davis and Chameides, 1984). To the extent that lightning is responsible for the NOx levels in the remote troposphere, it would appear that light ning plays an important role in the photochemistry of the atmosphere.

OTHER TRACE GASES PRODUCED BY LIGHTNING
While NO has received the bulk of the attention with regard to production by lightning because of its impor tance in the photochemistry of the atmosphere, research has revealed that a myriad of other trace gases in addi tion to NO can be generated by lightning in a variety of interesting environments. These gases and their yields in electrical discharges as determined by both theoretical calculations and laboratory experiments are listed in Table 6.3. The comparison between experimentally and theoretically derived yields is quite good over a wide range of gases and atmospheric compositions.
The production of HCN in a reducing, prebiological terrestrial atmosphere is of particular interest because it has been proposed that lightning-produced HCN was an

TABLE 6.3 Calculated and Experimentally Derived Yields of Trace Cases in Various Atmospheres0

Species

Calculated Yield (molecules/J)

Experimental Yield (molecules/J)

Present-Day Terrestrial Atmosphere

NO

9 x lO"

(2-6) x 1016

CO

(0.1-5) x 10"

1 x 10"

N2O

(3-13) x 1012

4 x 1012

Reducing Prebiological Terrestrial Atmosphere (95% Nj, 5% CH4)

HCN

(6-17) x 10ie

-10"

C. Cytherian Atmosphere (95% CO2,5% N2)

CO

(1-1.4) x 10"

3 x 1016

NO (o2 + O)

(5-6) x 1015 (6-9) x 1016

4 x 1015
—

D. Jovian Atmosphere (99.95% H2, 0.05 % CH4)

CO

5 x 10"

—

N2

5 x 10"

—

HCN

9 x 1013

—

C2H2

3 x 1013

—

C2H,

2 x 1012

—

HCHO

8 x 10"

—

CO,

3 x 10"

—

C2H8

4 x 10"

—

Titan Atmosphere (97% N2, 3% CH4)

HCN

1.2 x 10"

C2N2

2.5 x 1014

C2H2

7.5 x 1015

C2H4

5 x 10"

"References: 1. Borucki and Chameides (1984) 2. Chameides et al. (1977) 3. Levine era/. (1981) 4. Peyrous and Lapeyre (1982) 5. Chameides (1979) 6. Levineera/. (1979) 7. Chameides and Walker (1981) 8. Sanchez etal. (1967)

9. Bar Nun and Shaviv (1975) 10. Bar Nun etal. (1980) 11. Chameides etal. (1979) 12. Bar Nun (1980) 13. Levineera/. (1982) 14. Lewis (1980) 15. Bar Nun (1975) 16. Borucki etal. (1984)

Reference
1,2,3,4 5,6 6
7, 8, 9, 10
6,11,12 11,12,13 11
14,15 14,15 14,15 14,15 14,15 14,15 14,15 14,15
16 16 16 16

76

WILLIAM L. CHAMEIDES

organic precursor that ultimately led to the chemical evolution of life on Earth (Miller and Urey, 1959). Both laboratory and theoretical calculations indicate that in a highly reducing atmosphere, rich in hydrocarbons, lightning could have produced HCN in copious quanti ties, possibly large enough to allow HCN levels in ponds and ocean water to build to levels large enough to trig ger the formation of peptide chains and similar precur sors to amino acids. On the other hand, the calculations of Chameides and Walker (1981) indicate that the HCN yield rapidly decreases as the atmosphere becomes less reducing. For an atmosphere where C is primarily in the form of CO, the HCN yield decreases by about 3 orders of magnitude from that of a hydrocarbon atmosphere, and it decreases by an additional 3 orders of magnitude for a CO2 atmosphere. Thus in order to better under stand the role of lightning in the evolution of life, studies are needed to better determine the relative amounts of CH4, CO, and C02 in the primitive atmosphere.

THE POSSIBLE ROLE OF ATMOSPHERIC DISCHARGES IN CLOUD CHEMISTRY

In recent years the growing concern over the possible deleterious effects of acidic precipitation on lakes, forest ecosystems, and crops has led to an increased interest in the chemistry of clouds, a region where acids can be effi ciently generated and incorporated into rainwater. One aspect of cloud chemistry that has yet to be adequately studied is the role of atmospheric electrical phenomena in acid generation in electrified clouds. One possible ef fect of electrical discharges on cloud chemistry is briefly described below.
Suppose, under the appropriate conditions, continu ous, low-level positive point coronal discharges from droplets occurred in a cloud. These discharges would cause 1 electron to be deposited on the droplet and 1 positive ion (most often O 2 ) to be produced in the gas phase for each ion pair produced (Loeb, 1965). The electrons deposited on the droplet would be incorpo rated into the droplet and become hydrated electrons [i.e., (e")a(J. In the presence of dissolved O2, these hy drated electrons would rapidly form O \ via

(e-)aq + (O2),

O,

(Reaction 6.8)

The O 2 species is related to the aqueous-phase HO2 rad ical by the acid-base equilibrium reaction

HO2 -*—»- O2 + H+. (Reaction 6.9)

The O 2 ions produced in the gas phase would lead to the eventual formation of an OH radical and a hydrated oxonium ion, H3O+ (Good etal., 1970). Heterogeneous scavenging of the H3O+ . nH2O ion and its incorpora tion into the droplets to form H+ would maintain the

nominal charge neutrality of the droplets. A sizable fraction of the gas-phase OH radicals produced by the discharge would also be scavenged, either as OH or H02 in the gas phase, and incorporated into the droplet rep resenting an additional radical source to the aqueous phase. Calculations similar to those of Chameides (1984) indicate that about 1.5 aqueous-phase HO2 free radicals would be produced for each ion pair generated.
The HO2 radicals thus produced in the aqueous phase would rapidly react to form dissolved H2O2 via reactions such as

HO, + O;

H2O2 + O2. (Reaction 6. 10)

H

Since aqueous-phase H2O2 is believed to be, in many cases, the most important oxidant of dissolved SO2 in cloud droplets leading to the production of sulfuric acid (Martin, 1983), it is conceivable that this electrical pro cess could play a significant role in the generation of ac ids in clouds.

CONCLUSION
The agreement between theoretical calculations and experimental determinations of chemical yields from discharges for a wide range of gaseous species and a wide range of atmospheres suggests that the basic chemical mechanism by which trace species are produced by lightning is fairly well understood. However, in order to infer global production rates from these chemical yields, accurate values for the rate at which energy is dissipated by lightning is needed. Because these dissipation rates are not well known (uncertainty factors of 10 for the Earth and much larger for other planets are estimated), the role of lightning in the global budgets of species such as NO remains uncertain. To reduce this uncertainty, studies are needed to characterize more accurately the energy and frequency of lightning strokes on a global scale.
Another area where research is needed concerns coro nal discharges and their role as local sources of trace spe cies. Mechanisms exist, for instance, by which positivepoint corona from cloud droplets could lead to enhanced generation of sulfuric acid in cloudwater. To determine if this and similar processes occur at a signifi cant rate, the magnitude of low-level coronal currents in clouds needs to be more accurately established.
REFERENCES
Bar Nun, A. (1975). Thunderstorms on Jupiter, Icarus 24. 86-94. Bar Nun. A. (1980). Production of nitrogen and carbon species by
thunderstorms on Venus, Icarus 42, 338-342. Bar Nun, A., and A. Shaviv (1975). Dynamics of the chemical evolu
tion of Earth's primitive atmosphere, Icarus 24, 197-211.

THE ROLE OF LIGHTNING IN THE CHEMISTRY OF THE ATMOSPHERE

77

Borucki, VV. L., and VV. L. Chameides (1984). Lightning: Estimates of the rates of energy dissipation and nitrogen fixation. Rev. Geophyti. 22,364.
Borucki, W. L., C. P. McKay, and R. C. Whitten (1984). Possible production by lighting of aerosols and trace gases in Titan's atmo sphere, Icarus 60, 260-274.
Burns, R. C, and R. W. Hardy (1975). Nitrogen Fixation in Bacteria and Higher Plants, Springer- Verlag, Berlin.
Chameides, W. L. (1979). The implications of CO production in elec trical discharges, Geophys. Res. Lett. 6, 287-290.
Chameides, W. L. (1984). The photochemistry of a remote marine stratiform cloud, /. Geophys. Res. 8.9, 47-39-4755.
Chameides, W. L., and J. C. G. Walker (1981). Rates of fixation by lightning of carbon and nitrogen in possible primitive atmospheres, Origins of Life 1 1 .
Chameides, W. L., D. H. Stedman, R. R. Dickerson, D. VV. Rusch, and R. J. Cicerone (1977). NO, production in lightning, /. At mas. Sci. 34, 143-149.
Chameides, W. L., J. C. G. Walker, and A. F. Nagy (1979). Possible chemical impact of planetary lightning in the atmospheres of Venus and Mars, Nature 280, 820-822.
Crutzen, P. J. (1983). Atmospheric interactions— Homogeneous gas reactions of C, N, and S containing compounds, in The Major Biogeochemical Cycles and Their Interactions, B. Bolin and R. B. Cook, eds., SCOPE, Parts.
Davis, D. D.,andW. I.. Chameides (1984). The atmospheric chemis try of electrified clouds, presented at VII International Conference on Atmospheric Electricity, Albany, N.Y.
Dawson, G. A. (1980). Nitrogen fixation by lightning, /. Atmas. Sci. 37,174-178.
Drapcho, D. L., D. Sistcrson, and R. Kumar (1983). Nitrogen fixation by lightning activity in a thunderstorm, Atmas. Environ. 17, 729734.
Calbally, I.E., andC. R. Roy (1978). Loss of fixed nitrogen from soils by nitric oxide exhalation, Nature 275, 734-735.
Good, A., A. Durden, and P. Kebarle (1970). Mechanism and rate constants of ion-molecule reactions leading to formation of H " . (H2O)„ in moist oxygen and air, /. Chem. Phys. 52, 222-229.
Criffing, G. W. (1977). Ozone and oxides of nitrogen during thunder storms, /. Geophys. Res. 82, 943-950.
Hill R. D., R. G. Rinker, and H. Die Wilson (1980). Atmospheric nitrogen fixation by lightning, /. Atmos. Sci. 37, 179-192.
Levine, J. S., R. E. Hughes, W. L. Chameides, and VV. E. Howell (1979). N/) and CO production by electric discharge: Atmospheric implications, Geophys. Res. Lett. 6.
Levine, J. S., R. S. Rogowski, G. L. Gregory, W. E. Howell, and J. FLshman (1981). Simultaneous measurements of NO,, NO, and O, production in a laboratory discharge: Atmospheric implications, Geophys. Res. Lett. 8, 357-360.

Levine, J. S., G. L. Gregory, G. A. Hervey, VV. E. Howell. VV. J. Borucki, and R. E. Orville (1982). Production of nitric oxide by lightning on Venus, Geophys. Res. Lett. .9. 893-896.
Levy, H., II, J. D. Mahlman, and W. J. Moxim (1980). Stratospheric NO,,: A major source of reactive nitrogen in the unpolluted tropo sphere, Geophys. Res. Lett. 7, 441-444.
Lewis, J. S. (1980). Lightning synthesis of organic compounds on Jupi ter, Icarus 42, 85-95. ,
Lipschultz, F., O. C. Zafirious, S. C. Wofsy, M. B. McElroy, F. VV. Valois, andS. VV. Watson (1981). Production of NO and N»O by soil nitrifying bacteria: A source of atmospheric nitrogen oxides, Nature 294,641-643.
Loeb, L. B. (1965). Electrical Coronas: Their Basic Physical Mecha nisms, Univ. of California Press, Berkeley, 694 pp.
Logan, J. A. (1983). Nitrogen oxides in the troposphere: Global and regional budgets, /. Geophys. Res. 88, 10785-10807.
Logan, J. A., M. J. Prather, S. C. Wofsy, and M. B. MeElroy (1981). Tropospheric chemistry./. Geophys. Res. 86, 7210-7254.
Martin, L. R. (1983). Kinetic studies of sulfite oxidation in aqueous solution, in Acid Precipitation, J. G. Calvert, ed., Ann Arbor Sci ence, Ann Arbor, Mich.
McConnell, J. C. (1973). Atmospheric ammonia,/. Geophys. Res. 78, 7812-7821.
McFarland, M. C, D. Kley, J. VV. Drummond, H. L. Schmeltekopf, and R. H. Winkler (1979). Nitric oxide measurements in the equato rial pacific region, Geophys. Res. Lett. 6, 605-608.
Miller, S. L., and II. C. Urey (1959). Organic compounds synthesis on the primitive Earth, Science 130, 245-251 .
Noxon, J. A. (1976). Atmospheric nitrogen fixation of lightning, Geophys. Res. Lett. 3, 463-465.
Noxon, J. A. (1978). Tropospheric NO;, /. Geophys. Res. 83, 30513057.
Peyrous, VV., and R. M. Lapeyre (1982). Gaseous products created by electrical discharges in the atmosphere and condensation nuclei re sulting from gaseous phase reactions, Atmos. Environ. 16, 959-968.
Picone, J. M., J. P. Boris, J. R. Greig, M. Rayleigh, and R. F. Fernster (1981). Convective cooling of lightning channels, /. Atmos. Sci. 38, 2056-2062.
Salanave, L. E. (1961). The optical spectrum of lightning. Science 134, 1395-1399.
Sanchez, R. A., J. P. Ferris, and L. E. Orgel (1967). Studies in prebiotic synthesis. II. Synthesis of purine precursors and amino acids from aqueous hydrogen cyanide, /. Mol. Biol. 30, 223-253.
Tuck, A. F. (1976). Production of nitrogen oxides by lightning dis charges, O. /. R. Meteorol. Soc. 102, 749-755.
Uraan, M. A. (1969). Lightning, McGraw-Hill, New York, 264 pp. Z'elovich.Y. B.,andY. P. Raizer (1966). Physics of Shock Waves and
High-Temperature Hydrodynamic Phenomena, Academic Press, New York.

II CLOUD AND THUNDERSTORM ELECTRICITY