PHYSICAL VIEW OF CLOUD SEEDING

• Document Type: CREST
• Collection: NGA Records (Formerly NIMA)
• Document Number (FOIA) /ESDN (CREST): CIA-RDP80T01137A000200030004-2
• Release Decision: RIPPUB
• Original Classification: K
• Document Page Count: 13
• Document Creation Date: December 28, 2016
• Document Release Date: March 25, 2010
• Sequence Number: 4
• Publication Date: April 10, 1970
• Content Type: OPEN SOURCE

Sanitized Copy Approved for Release 2010/03/25 :CIA-RDP80T01137A000200030004-2
FROM DATE.
~~ L , ~ % `f
TO INITIALS DA E REMARKS
DIRECTOR
~
~ ~
~
~
DEP DIRECTOR
A,
'^a
~
~~
EXEC/DIRECTOR
SPECIAL ASST
%
ASST TO DIR
HISTORIAN
CH PPBS
DEP CH PPBS
EXO PPBS
--
__-
-
1,U L
3
~_
. x~,
~ ~.
CH;SS
~~
r ;,,.
DEP CH-SS
SC$P
RECORDS MGT
PERSONNEL
LOGISTICS
TRAINING
SECURITY
FINANCE
CH IEG
DEP CH IEG
EXO IEG
CH PSG
DEP CH PSG
EXO PSG
CH;TSG
DEP CH TSG
EXO;'TSG
DIR; IAS/DDI
CHr'DIAXX-4
--- -
--
-
CH; DIAAP-9
CHiSPAD
Sanitized Copy Approved for Release 2010/03/25 :CIA-RDP80T01137A000200030004-2
 Sanitized Copy Approved for Release 2010/03/25 :CIA-RDP80T01137A000200030004-2
EXECUTIVE OFFICE OF THE PRESIDENT
OFFICE OF SCIENCE AND TECHNOLOGY
WASHINGTON, D.C. 20506
August 6, 1970
ARGO Steering Committee Members
In view of the Steering Committee's interest in the possibility of
modifying storms such as the one which hit Lubbock, Texas, the
attached paper by Myron Tribus is furnished for your information.
STAT
Sanitized Copy Approved for Release 2010/03/25 :CIA-RDP80T01137A000200030004-2
 Sanitized Copy Approved for Release 2010/03/25 :CIA-RDP80T01137A000200030004-2
Preprinted from the 10 April 1970 issue
SCIENCE
Physical View of Cloud Seeding
A review of experimental data indicates that we are
considerably further ahead than is generally realized.
Myron Tribus
From the very beginning of cloud
seeding in 1946, the subject has been
contl?oversial. There are many reasons
for the controversies, not the least of
which have been the strong personali-
ties, beliefs, and convictions of the
people involved. Throughout it all there
has been an underlying uncertainty
about weather phenomena in general.
The question "Would it have rained,
anyway?" has been unanswered and,
until recently, unanswerable.
In this article I wish to present some
views based on 25 years of intermittent
association with this field. For the last
decade I have been intensely involved
in studies on the foundations of in-
ductive logic and scientific inference.
These two fielfls are basic to attempts
to establish scientific hypotheses about
weather modification. In the title of
this paper I have emphasized the word
physical because I wish to distinguish
it from the statistical view. The distinc-
tion is extremely important to the plan-
ning and the prosecution of research.
Statistical methods now routinely em-
ployed do not, in general, take into ac-
count known mechanisms or the physics
of a process. I have written elsewhere
on this weakness (1) and have used the
following example: Suppose I claimed
clairvoyant capabilities and, in support
of this claim, I correctly predicted the
makeup of the front page of the New
York Trmes one week in advance,
down to the smallest detail, including
a few typographical errors. Would not
your common sense tell you my claim
was valid? For problems that use as
evidence the results of a single ob-
servation, there are no classical statis-
tical procedures. Most of us rely on
common sense when evaluating this
type of problem. It is possible (1) to
develop procedures that can handle this
class of problem, but the new proce-
dures have not yet been generally ac-
cepted as valid by the majority of
statisticians. (See the appendix for a
brief mathematical analysis of the prob-
lems of clairvoyance and weather modi-
fication.) It is my belief, however, that
the methods will be accepted in the
rot too distant future. The reason is
simple: today we are dealing with many
problems that require these new tech-
niques, and necessity is still the mother
of invention.
As I have argued (1), the main weak-
ness of most existing statistical ap-
proaches is that they do not permit
us to use all that we really know. It
is not that the available methods are
"wrong"-it is that they are inadequate.
Cloud seeding is a complex process.
During the process there are many
intermediate stages of development,
each with its own measurable charac-
teristics. Thus, when a cloud is seeded,
we observe such variables as the cloud
base height, the updraft velocity distri-
bution, the radar echoes from the core,
the rate of growth of the top of the
cloud, the pattern of the convective
activity, the temperature distribution
and the time at which water starts to
fall from the cloud, where the water
falls, and how much evaporates. When
we compare these results with the pre-
diction of a digital computer, a single
observation in which we obtain very
good agreement in all details obviously
weighs more heavily in our minds than
does a single statistical measure which
merely considers the ratio of rainfall
measured in gauges for seeded and
nonseeded clouds. Of course, we must
observe a few nonseeded clouds, in
the same fine detail, to see if their
behavior agrees with our computer pre-
diction. But our common sense tells us
that experiments that reveal this fine
detail are more significant than experi-
ments that do not.
The issue is not a trivial exercise in
logic-a mere comparison of prefer-
ences as regards the proof of scientific
claims. The perspective adopted de-
termines how an experiment is planned
and therefore how much money is spent
in coming to a particular state of
knowledge. Usually the biggest single
item of expense involves the operation
of aircraft. In addition to the capital
costs of aircraft and sophisticated in-
strumentation, such factors as the cost
of flight crews, fuel, and aircraft over-
haul make each hour of flight
time very expensive. Today it is not
uncommon for experimenters to fly
about half the missions without seeding,
just to produce some "randomized re-
sults." My quarrel is not with the va-
lidity of the statistical approach-it is
whether the classical statistical ap-
proach to design of experiments ought
to be followed without regard to the
expense. A more useful tool is decision
analysis, which adds statistical consid-
erations to cost parameters. Some ex-
perimenters, incidentally, recognize this
fact intuitively and bias the randomiza-
tion by making the odds two to one
in favor of seeding on a given flight.
Even if decision analysis is used, how-
Dr. Tribus is Assistant Secretary for Science
and Technology of the U.S. Department of Com-
merce and is chairman of the Interdepartmental
Committee for Atmospheric Sciences of the Fed-
eral Council for Science and Technology. Thls
paper was presented at the Second National Con-
ference on Weather Modification, held in Santa
Barbara, California, from 6 to 9 Aprll 1970.
Sanitized Copy Approved for Release 2010/03/25 :CIA-RDP80T01137A000200030004-2
 Sanitized Copy Approved for Release 2010/03/25 :CIA-RDP80T01137A000200030004-2
ever, it should not be applied without
taking into account the physics of the
process. Decisions are based on de-
scriptions, and it is important that, in
the process of developing a statistical
description of our state of knowledge,
we not ignore some data just because
our mathematical tools cannot encom-
pass them.
The key to what we call scientific
understanding is the elucidation of
mechanisms-that is, chains of events-
each of which we can understand and
link with others to describe an out-
come. If all these individual events per-
mit different outcomes, the final out-
come may indeed be so variable as to
make it very difficult to prove our
knowledge by study of only the final
outcome. If purely statistical methods
are employed, some intermediate
measures (that is, parameters) may be
included in a statistical analysis via
"stratification," but stratification re-
quires that there be more, not fewer,
data points to reach a given "confidence
level." In systems as complex as meteor-
ological ones, I argue that we will move
ahead faster if we spend money to get
a greater variety of carefully planned
observations instead of a greater
number.
I wish to consider first the general
scientific view of cloud seeding and
then particularize this view to specific
applications. Then I shall return to
the use and misuse of statistics in this
field.
As Joanne Simpson (2) has ob-
served, analyses of cloud seeding may
be considered to be either static or
dynamic. A static analysis is one in
which the natural flow field is not
markedly affected by seeding. Static
analysis seems to be appropriate to
quiescent cloud chamber experiments,
stratus clouds, and orographic lifting
such as occurs either at a mountain
range or in the frictional boundary
layer where the wind over water first
meets the land. In orographic lifting
the buoyancy forces generated by seed-
ing are usually too weak to cause a
relative motion between air masses.
The resultant motions of air are nearly
the same with and without seeding.
The practical effect is that, in describing
the fluid flow field, the energy and mo-
mentum equations are decoupled and
may be solved separately.
In a dynamic analysis, on the other
hand, intervention markedly affects the
flow field (for example, by artificial
buoyancy), and the gross features of
the cloud may be greatly modified by
the presence or absence of phase
changes in the cloud water. Complete
diagnostic analyses are more difficult
to perform because they inherently in-
volve solving the equations of motion
of a three-phase system (air, water, ice)
in a nonisothermal field with the energy
and momentum equations inextricably
linked together. Although complete
solutions starting from the primitive
equation of energy, momentum, ther-
modynamic state, and rate processes
would be desirable, they have not been
obtained for many, even simpler sys-
tems in meteorology. It is not a weak-
ness peculiar to cloud seeding that we
must be satisfied, as of now, with
approximate solutions. The approxi-
mate solutions for mixing, diffusion,
and cloud droplet growth can be in-
dividually checked and put together
to form a system of equations. Even
the approximations are quite complex,
and a digital computer is essential if
any analysis at all is to be carried out.
Computer analysis is therefore an
essential adjunct of useful experimen-
tation, for experiment without analysis
cannot provide the basis for prediction.
Certain features are common to both
static and dynamic analyses. In both,
it is postulated that humid air is cooled
by expansion and that initially small
droplets of water are formed, too tiny
to fall relative to the air. In the cloud
chamber (3) the expansion rate may
be controlled by evacuation pumps.
In orographic lifting the rate of ex-
pansion of cloud mass is often con-
trolled by the larger circulation pat-
terns, which impose the motion but
are relatively unaffected by cloud
seeding.
In cumulus clouds, on the other
hand, the rate of expansion is deter-
mined by the buoyancy forces, which
in turn are determined by the conden-
sation rate and by the vertical variation
of the basic horizontal velocity field
(which is why I said the equations of
energy and momentum were inextrica-
bly linked).
When the cloud temperature goes
below 0?C, most of the droplets do
not turn to ice but remain supercooled.
The size of the droplets at any stage
depends upon the number of conden-
sation nuclei present in the original
air mass, the rate of cooling, the origi-
nal humidity, and the number of vari-
ous kinds of freezing nuclei present-
Lhat is, substances that catalyze the
transition from water to ice. We still
do not have adequate information on
the number of ice and condensation
nuclei necessary to initiate and sustain
the precipitation process. Techniques
now in use have exhibited inconsisten-
cies up to 9 orders of magnitude when
employed by various investigators. De-
velopment of more accurate methods
for measuring these particles together
with increased studies of the basic
mechanisms of the nucleation process
are urgently needed; if successful, they
will result in more efficient cloud seed-
ing techniques.
Langmuir (4) has described the fu-
rious competition between small and
large drops which results in the growth
of large ones at the expense of the
small. This competition is greatly af-
fected by the rate of expansion (which
may be so large as to cause all drops
to grow), the temperature (low tem-
peratures reduce vapor pressures and
hence growth rates), and the presence
of freezing nuclei. Since the vapor
pressure of ice is much lower than
that of supercooled water at the same
temperature, all liquid water drops tend
to vaporize into the freezing nuclei,
thereby liberating additional heat of
freezing and artificial buoyancy. Also,
the lowered saturation vapor pressure
causes more condensation from the
cloud air mass, which further enhances
buoyancy. The "triggering effect" pan
be spectacular; it can create in seeded
clouds strengthened updrafts leading to
a vertical growth 4 to 5 kilometers
higher than the tops of unseeded clouds.
This growth leads to further conden-
sation and frequently to increased "nat-
ural" precipitation.
Of cuurse the effect of the introduc-
tion of "artificial" freezing nuclei into
a cloud depends in part upon how
many were already there from "natural"
causes. I put quotation marks around
the words "artificial" and "natural"
because at many locations man's activi-
ties now so pollute the atmosphere
that we cannot always distinguish "nat-
ural" from "artificial."
The behavior of the particles in a
cloud is strongly dependent upon the
interaction among nuclei, velocity field,
and temperature. For a given amount
of condensation, the more nuclei there
are, the smaller the droplets will be.
Small droplets will fall relative to the
air more slowly than large ones and
may, therefore, be carried up within
the cloud. The temperature determines
not only the amount of water vapor
Sanitized Copy Approved for Release 2010/03/25 :CIA-RDP80T01137A000200030004-2
 Sanitized Copy Approved for Release 2010/03/25 :CIA-RDP80T01137A000200030004-2
in saturated air; it also has an im-
portant effect upon diffusion and growth
processes. In a few systems, especially
systems in which the air motion is not
affected by the thermal processes as-
sociated with condensation, it is pos-
sible to find simple solutions to the
cloud growth equations without re-
course to a digital computer. Lang-
muir's "time of rise" treatment of a
cloud growth on Mount Washington
is an example (S). In this work Lang-
muir successfully predicted the sizes
of drops that would be measured at
the summit, though he worked only
with information about the cloud base
height, the velocity of the wind (which
determines the time for the droplets
to pass from cloud base to summit),
and the temperature. But usually the
equations are so complex that they can
be handled only on a digital computer.
When the energy released during
condensation is large, the cloud density
changes. The velocity field is then de-
termined by the difference between
the density that occurs inside the cloud
and the density outside the cloud. If
this density difference between the in-
side and the outside of the cloud is
large, the result can be an extremely
strong updraft in the cloud. This differ-
ence depends strongly on the ambient
humidity and temperature lapse rate. In
tropical clouds, which develop in a
moist environment, it is not uncommon
to see a cloud grow to 15,000 meters.
Since the buoyancy effects are inte-
grated over the vertical extent of a
cloud, under some conditions seeding
from the top can kill the cloud. This
effect occurs when a dry stable layer
surrounds the cloud's midsection with
a less stable layer above. Then seeding
can cause a too rapid growth of the
cloud top compared with the conden-
sation rate beneath it. Cold air comes
in from the side and causes the top of
the cloud to break away from the
base. The total buoyancy force avail-
able within the cloud is thus diminished,
and the base collapses.
There are more considerations. When
cloud drops grow large enough to fall,
they descend through smaller drops,
collecting and coalescing as they go.
The dynamics of this collection process
have been extensively studied, both ex-
perimentally (4) and analytically.
Cloud seeding therefore involves
many considerations of fluid mechanics,
heat transfer, diffusion, thermodynam-
ics, two-phase flow, particle dynamics,
and surface chemistry. Instrumentation
includes radars, reconnaissance aircraft,
precipitation gauges, radiosondes, drop-
sondes, kites, cloud photography, and
the human eye. Predicting detailed be-
havior in such a complex system re-
sembles the problem of predicting the
front page makeup of the New York
Times.
The interplay of these effects can
produce bewildering results for those
who take the pragmatic view that all
that counts is the results and who pay
no attention to the elucidation of mech-
anisms. Thus, at very low temperatures
there are many kinds of particles that
can act as ice-nucleating agents. At
higher temperatures, there are fewer
active nuclei. Seeding at low tempera-
tures, therefore, may produce anover-
seeded condition and reduce the rain-
fall. Seeding at higher temperatures can
increase the rainfall. Even the level in
the cloud at which the seeding occurred
may make a difference. Neyman ana-
lyzed the Whitetop experiment statisti-
cally by lumping all seedings together
and counting only rainfall. He thereby
showed there had been a net decrease in
rainfall due to cloud seeding (6). But
Fleuck (7) analyzed the same data, this
time stratified as to maximum radar echo
top heights, and showed that low cloud
tops (less than 6100 meters above
mean sea level) resulted in slight but
not statistically significant decreases in
precipitation, that intermediate cloud
tops were associated with substantial
increases, and that sufficiently high lops
(more than 12,200 meters above mean
sea level) favored significant decreases.
The Neyman and Flueck analyses do
not conflict. It is true, as Neyman says,
that indiscriminate seeding without
complete knowledge of the physics in-
volved can lead to an unintended re-
sult. But Flueck's calculations also show
that knowledge of the basic mechanisms
involved permits the selection of opti-
mum techniques for desired control.
Neyman's methods of analysis do not
admit the use of whatever insights
were available from other experiments
about the physics of the process. On
the postulate that all future seeding
activities will be conducted in equal
ignorance, Neyman's warnings certainly
follow from his analysis. But, as I have
written elsewhere, we do know more
than what is learned by reading rain
gauges (8). It was on this basis that
I asserted that we are closer to the
capability to do operational cloud seed-
ing than Neyman's article indicated.
Let us turn to a review of the experi-
ences upon which that assertion is
based.
Orographic lifting is one of the
simpler cases to analyze because, even
with seeding, the resultant buoyancy
forces are usually too weak to have
much effect on the overall motion. In
a fixed fluid flow field, the instrumen-
tation can be used to measure many
details of the system with or without
seeding.
The broad-scale synoptic situation
induces a flow up the mountainside,
which in turn produces a cloud that
appears to hang motionless over the
top of the mountain. Actually the cloud
is extremely active, with moisture en-
tering the cloud base and droplets
growing as the air ascends. Vaporiza-
tion takes place on the lee side as the
air is compressed.
At the South Dakota School of
Mines and Technology, under the spon-
sorship of the Bureau of Reclamation
and the Department of Defense, Or-
ville (9) has used detailed descriptions
summarized by Kessler (10) (see Fig.
1) to study the effect of the various
precipitation processes in a numerical
simulation of cloud development over'
a mountain barrier. The mathematical
model is two-dimensional and incor-
porates a vertical wind shear and an
initially stable, incompressible atmo-
sphere. The rain shower model is pro-
grammed so that each cloud affects its
own development during its life cycle.
As the multiple clouds form and grow
into the mature stage, cloud shadow
effects combine with the downdraft in-
fluences that result from evaporative
processes. The behavior of the clouds
is most realistic, and it will be of great
interest to compare Orville's results
with field measurements.
Under National Science Foundation
sponsorship, Lewis Grant and his col-
leagues at Colorado State University
have produced what may be regarded
as a classic experiment in atmospheric
sciences. Figure 2 shows the system
that Grant chose for study (3). By
using scale models, they have investi-
gated the flow field over the mountain
and have compared it with field mea-
surements obtained by radiosondes,
constant-level balloons, and tethered
kites and parafoils. In a dynamic cloud
chamber they have observed the rates
of droplet growth and nucleation under
conditions approximating conditions on
the mountain. With tracer techniques
they have tracked the release of par-
ticles from ground-based generators to
learn how to put nuclei where they
Sanitized Copy Approved for Release 2010/03/25 :CIA-RDP80T01137A000200030004-2
 Sanitized Copy Approved for Release 2010/03/25 :CIA-RDP80T01137A000200030004-2
Cloud water
(liquid)
Autoconversion
accretion
Lc =0 T 1,
p(H~DX) -~ 1 and, as x -~ 0, p(H~DX)
-> 0.] (ii) If x is not exceptionally close
to 1, find an experiment in which
z Q xy.
If xy ~ 0 (that is, HD not impos-
sible) , Eq. 6 may be divided by xy to
give
P(HIDX) = 1/[1 + (1 -x)/x(z/y)]
If x is neither close to 0 or 1 (that is,
the hypothesis is seriously in question),
the value of p(H~DX) depends mainly
on z/y; that is, it depends on the likeli-
hood of finding D true on H or h. We
can make this clear by simple ex-
amples.
Suppose I were to claim as my hy-
pothesis H that I could dissolve a
stratus cloud by putting COZ pellets into
it, and to support my claim I predict-
ably produced, as data D, the General
Electric monogram burned into a cloud
over Schenectady at a specified place
and time. Clearly, for such evidence z/y
is extremely small, and, since we have
said x ~ 0, p(H~DX) ~ 1. In other
words, my ability to dissolve stratus
clouds can be proven by very few ex-
periments or even by only one of this
type. On the other hand, if I only burn
a small hole in the cloud, even though
z/y is a large number, p(H~DX) is
larger than x but is not raised up to
unity. Here I will need a large number
of randomized experiments to prove my
hypothesis. This result coincides with
the commonsense observation that wa
often see holes in clouds (z is not zero).
It is the design of the experiment that
determines the number of tests.
The words "predictably produce" are
very important. There are patterns to be
found in every random sequence, if we
just look hard enough and creatively
enough. Langmuir and Schaefer re-
ported (D') a 7-day national cycle
in rainfall when they were seeding
in New Mexico on a 7-day period;
this fact would have had a very pro-
nounced effect on the subsequent state
of science if they had predicted it in
advance. The fact they did not predict
it in advance does not make p(H~D'X )
smaller than p(H~X), but it does mean
that the assignment of a value to
p(H~D'X) depends more on X (other
knowledge) than on D'.
If any one observation, say D", has
about the same probability given H true
as it does given h true, it will take a
large amount of data to produce a value
of z/y which is very persuasive. Such
cases correspond to observing a D"
which is of itself not very surprising. It
takes a large collection of different D"
to be persuasive. In such cases, random-
ized experiments are indeed essential.
But, we may also make progress in
physics and its many derivative
branches of science by the performance
of "critical, experiments," which are ex-
periments which enable us to verify in
fine detail a complex theory. When the
predictive process and the instrumenta-
tion have been properly designed, and
if the prediction concerns a very com-
plex process, there is no need for ran-
domization. If these conditions cannot
be met, randomized "blind" runs are
essential. As Bayes' equation demon-
strates, the cases in which randomized
experiments are necessary and the cases
in which randomized experiments are
rot necessary merge smoothly into one
another. It is only necessary to make an
analysis in each case to decide which
conditions can be made to pertain.
1. M. Tribus, Rational Descriptions, Decisions,
and Designs (Pergamon, New York, 1%9).
2. J. Simpson, Med. Opin. Rev. S, 39 (1969).
3. N. Fukuta, J, Meteorol. 15, 17 (1938).
4. I. Langmuir, ibid. S, 175, (1946).
5. in The Collected Works of Irving
Langmuir (Pergamon, New York, 1961), vol.
l o.
6. I. Neyman, E. Scott, J. A. Smith, Selence 163,
1445 (1%9).
7. J. Flueck, Final Report of Protect Whltefop
(Univ. of Chicago Press, Chicago, in press),
part S.
8. M. Tribus, Science 164, 1341 (1%9).
9. J, Y. Liu and H. D. Orville, J. Atmos. Sci.
26, 1283 (1969).
10. E. Kessler, Meteorol. Monogr, 10, No. 32
(1969),
11. L. O. Grant and P. W. Mielke, Jr., in Pro-
ceedings of the Fi/th Berkeley Symposium on
Mathematical Statistics and Probability (Univ.
of California Press, Berkeley, 1%S), p. 113;
P. W. Mielke, Jr., L. O. Grant, C. F. Chap-
pell, private communications; 1. E. Cermak,
L. O, Grant, M. M. Orgill, private communi-
cation.
12. R. D. Elliott, papers presented at the Project
Skywater Conference on Modeling, Denver,
Colorado, 10 February 1970.
13. R. L. Lavoie, thesis, Pennsylvania State Uni-
versity (1%8).
Sanitized Copy Approved for Release 2010/03/25 :CIA-RDP80T01137A000200030004-2
 Sanitized Copy Approved for Release 2010/03/25 :CIA-RDP80T01137A000200030004-2
14. Warm fogs at airports are also dissipated by
use of sprays of various hygroscopic ma-
terials, but this matter is not of concern here.
Suffice it to say that several companies offer
this service at various airports in the United
States.
13. J. Simpson, Rev. Geophys. 3, 387 (1%S); I.
Simpson. G. W. Brier, R. H. Simpson, J.
Armos. Scl. 24, 508 (1%7); J. Simpson and
V. Wiggert, Mon. Weather Rev. 97, 471
(1%9); J. Simpson et al., J. Appl. Meteorol.,
in press.
16. G. K. Sulakvelidze, Trudy (High-Altitude
Geophys. Inst. Nalchik, U.S.S.R.) No, 7
(1966).
17. L. J. Batten, Bufl. Amer. Meteorol. Soc. 51,
925 (1%9).
I8. R. A. Schleusener, J. Appl. Mcterol. 7,
1004 (1968).
19. F. A. Ludlam, Nubile 1, % (1958).
20. R. H. Douglas, Meteorol. Monogr. S, No. 27,
157 (1963).
21. D. J. Musil, Rep, 69-11 (Institute of Atmo-
spheric Sciences, South Dakota School of
Mines and Technology, Rapid City, 1%9).
22. A. S. Dennis, paper presented before the
Weather Modification Association Meeting
(September 1%9). [Available as Pap, 69-73
(Institute of Atmospheric Sciences, South
Dakota School of Mines and Technology,
Rapid City, 1969)].
23. T. J. Henderson, in Proceedingr of the Fist
National Conference on Weather Mod/fication
(American Meteorological Society, Boston,
1%8), p. 474.
24. R. C. Gentry, Buli. Amer. Meteorol. Soc, St,
404 (t%9).
23. R. H. Simpson and I. S. Malkus, Sci. Amer,
211, 27 (1%S).
26. D. M. Fuquay and R. G. Baughman, "Project
Skyfirc Lightning Research" (final report to
National Science Foundation under grant
GP-2617, U.S. Forest Service, Missoula,
Montana, 1%9).
27. By the Phrase "proceed to operational weather
modification," I mean the development of a
specific timetable of operations to proceed
to pilot and then to regular operations spe-
cifically aimed at filling asocial need. The
bcation of the pibt operation should be choun
on the basis of economic and social, not sci-
entific, objectives. The observations should
be aimed at improving the operation, not
merely the state of scientific knowledge.
28. M. Tribus, statement before the Subcommit-
tee on Science, Research, and Development
of the House Committee on Science and
Astronautics, 2 December 1%9.
29. speech presented at the Sixth Tech-
nical Conference on Hurricanes, 2-4 Decem-
ber 1%9, Miami Beach, Florida.
30. I thank Robert M. White, loame Simp-
son, Joachim Kuettner, Vincent I. Schaefer,
Richard A. Schleusener, and Lewis O. Grant
for critically reviewing this paper and pro-
viding many helpful suggestions.
Sanitized Copy Approved for Release 2010/03/25 :CIA-RDP80T01137A000200030004-2