WMOD00478

• Title: Field Experimentation in Weather Modification
• Date: 3/1/1979
• Weather Modification - Doc Type: Report

~~:~tWI.-:=~4- --

..

'. Braham: . Field Experimentation in Weather Modification

of obtaining an understanding of the meteorological
pr6Clesses for a [proper interpretation of any seeding
effeCts iWl:!'mightuncover. I feel strongly that this was,
and still is, the best approach for establishing the facts
about weather modification. As a corollary, I am some-
what less than pleased when I see conclusions drawn from
the Whitetop rainfall or radar data in isolation from the
other parts of this experiment.
In the design phase, arrangements were worked out
whereby Professor Wayne Decker, University of Mis-
souri, would reduce the rain gage data and carry out the
initial analyses of it totally independently of the Uni-
versity of Chicago. Under our arrangement with NSF
and Professor Decker, he was not informed of which days
were seeded until the rain gage data reduction had been
completed and plume and nonplume averages computed
and tabulated. We furnished Professor Decker with pibal
data and assisted with the plume charts which he used in
his data reduction. He reported data for both the percent
of gage-hours with rain and the area average rain, for
both plume and nonplume, beginning with the first year's
operations. Professor Decker also carried out a limited
analysis of these data. At Chicago the initial efforts were
on analyzing the cloud measurements. Near the end of the
field operations, we began serious analysis efforts on the
rain and radar data" culminating in the monograph by
Flueck (1971).
Another activity that went on during the project
design phase resulted in the so-called selection criteria.
The most desirable subjects for our study were afternoon
and early evening convective clouds, which have a
climatological maximum frequency between about 11
A.M. and 6 P.M. and which account for a substantial
traction of the area rain. Because these clouds are rela-
tively numerous and dynamically coupled to the surface
and lower levels of the atmosphere, we reasoned that our
chances of inducj.Ilg and detecting seeding effects in them
~ere much better than in the .much larger nocturnal
squall-line storms. The latter are associated with large-
scale, dynamical structures that often begin in the high
plains during' late afternoon and sweep across the plains
at night to arrive in Missouri after midnight. Accordingly,
we concentrated our efforts on the afternoon and evening
clouds. But we needed some device to help select days
for the experiment, because only about half of all summer
days could be expected to have, somewhere in our re-
search circle, supercooled con\tectiVe clouds capable of
responding to our seeding materials. Limited resources
and a desire to reduce the number of "zeros" in the data
precluded extending the experiment to all summer days.
After testing several simple "prediction" schemes on data
for three prior summers from Springfield, Missouri, we
selected one which led to our selection criteria (Braham
1960, 1966). Our plan was to use these criteria for select-
ing operational (experimental) days. Once we were in the
field, we encountered situations requiring occasional de-
viations from the criteria. These deviations have already
been reported in the form of criteria values for all days

~_i,~d>J;:,;;'",_,:::~_ ~.,...i ~- ..

~.,.

99

and an indication of which days were actually used in
the experiment. In every case, the r:~ndomization decision
was made after all other decisions regarding selection of
days, selection of the location of possible seeding, etc.,
had been made.
Thus the purpose of the criteria was to help in selecting
days which would have clouds capable of responding to
sibrer iodide. It is a mistake to aS80ciate any particular
seeding hypotheBis with days which met, or did not meet,
these criteria.
I am very plen.sed that so many of the discussants ad-
dress one or more of the issues with which I closed my
article. These discussions may prove to be as important
as the article itself in that they reveal an increasing
amount and diversity of interest among statisticians in
meteorological issues. I think this E:ignals an opportunity
for progress in weather modification.
lOne of these issues relates to nonrandomized experi-
ments, and in particular to those carried out by com-
mercial cloud seeders. In general, it is clear that there are
types of inquiries where randomized experiments are
difficult or impossible, and where remarkable progress
has been made without them. (Consider the areas of
geology and astronomy.) But in cloud seeding, we agree
thl~t randomization is highly desirable. The reason given
by commercial operators for nonr,andomized operations
is that their customers object to giving up many (perhaps
half) of the opportunities for receiving the benefits they
believe to come from cloud seeding. I thank Professors'
Flueck and Mosteller for setting forth several of the
cautions and conditions to be observed in working with
nonrandomized data.
In the time that has elapsed between preparation of my
paper for the Chicago meeting of ASA and preparation
of this response, the two-volume final report of the
Weather Modification Advisory Board has been issued.
(I had the honor of serving on that Board with 16 other
persons, mostly nonmeteorologists.) This Board was ap-
pointed by Dr. Juanita Kreps, Secretary of Commerce,
in response to Public Law PL94-490, in which the
Congress asked the Secretary to undertake a thorough
re'new of weather modification, assess its current status,
and recommend public policy for it. The first volume of
the report discusses the subject in much more detail than
I was able to do in my article, including societal, environ-
mental, legal, and institutional and governmental factors.
The second volume was prepared for the Board by three
sULtisticians: David R. Brillinger, Lyle V. Jones, and
John W. Tukey; it deals exclusively with statistical issues
relating to wea.ther modification experimentation and
gives a statistical critique of nine recent seeding projects.
I recommend both volumes for further information and
dil>cussion of many of the issues raised in my article and
in these discussions.

Responses to Individual Discussants:

Cook-Holschuh: I agree with the view that meteorol-
ogists must develop an ability to Plredict additive subsets
,