WMOD00478

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

----

- ,,"f-

-"~~

~;~..~....".-~

Simpson: Field Experimentation in Weather Modification

JOANNE SIMPSON*

9S

Comment

If perfect predictions specifying the natural behavior
of clouds, cloud systems, and storms were feasible and
if all important variables were identified and precisely
measured in treated and untreated weather systems,
then statistical controls and inference would playa less
crucial role in atmospheric experimentation than they do.
Braham's article provides a much needed retrospective
summary of'Project Whitetop (1960-1964) that is par-
ticularly valuable in the context in which he places it,
with subsequent experimentation and progress in under-
standing the physical processes.
Braham himself was probably the original pioneer in
bringing statistical controls to atmospheric experiments;
he did this by means of collaboration with members of
the University of Chicago Statistics Department, par-
ticularly Brownlee and Kruskal. Prior to Whitetop,
Braham had concluded a randomized single-cloud experi-
ment in the tropics, with a treatment consisting of a
droplet spray to enhance coalescence (Braham, Battan,
and Byers 1957). Results showed that radar echoes (in-
dicating rain-sized droplets) did indeed occur earlier in
the treated clouds. Unfortunately, this fascinating experi-
ment was not carried forward to consider rain reaching
the ground or to area-wide cloud systems.
Whitetop, directed by Braham, was among the earliest
statistical-meteorological collaborations in experiment
design applied to area-wide systems. Partly because of
the infancy of the field, the project became unnecessarily
controversial and has not previously been fully reported
outside the immediate meteorological community. White-
top provided an important learning experience, not only
in cloud and precipitation physics, but also in both pro-
ductive and counterproductive interactions between sta-
tisticians and meteorologists.
The two other experimental areas briefly discussed by
Braham, namely, winter clouds over the mountains in
the West (Climax and Colorado River Basin) and sum-
mer cumulus in Florida (FACE 1), illustrate the fruits
of evolving experimental procedures which have resulted
partially from the closer interweaving of meteorological

· Joanne Simpson is William W. Corcoran Professor, Department
of Environmental Sciences, University of Virginia, Charlottesville,
VA 22903. Research wa.s supported by Grant ATM-lOO87 from the
National Science Foundation. I am grateful to the former RANN
Division of the National Science Foundation (under Grant No.
GI-43764) for providing statistical colleagues, in particular Dr.
Ronald Biondini, who helped with this note. I also apprecia.te the
opportunity so provided to help train a few statistical-meteorology
graduate students and to conduct. weather-modification research in
an environment where the interface between meteorologists and
statisticians ha.s become increa.singly permeable.

.L1;.::<::i..."

and statistical expertise. Braham's description of these
randomized experiments and nonrandomized seeding
operations also highlights the difficult and complex chal-
len~~es facing further progress in weather modification,
which will require joint advances in statistical tools
interacting with physical knowledge, perhaps combined
in the same individual or at least in closely interacting
teams.
Like medicine, weather modification first grew mainly
in s.n operational mode from dire needs of man, afflicted
by drought, hail, and severe storm damage. Cloud seeding
was, tried as a cheap remedy for these ills, based on
simplistic hypotheses or even, on too many occasions, on
trial and error. The concept of "experimental mete-
orology" has evolved only slowly over the three decades
sinc:e the discovery of glaciogenic seeding. This approach
is now more well known and accepted, aided by closer
involvement of the statistical community, particularly
tho13e with experience in the related fields of agricultural
and medical experimentation. Its implementation is
cles.rly expensive in sustained application of funds, labor,
and facilities.
Project Whitetop was innovative in bringing statis-
tically controlled design together with physical measure-
ments. It was a full-scale meteorological experiment with
a substantial target area on land. It pioneered in atmo-
spheric hypothesis testing by assigning treatments (seed
or no seed) randomly to objectively defined experimental
units and by ensuring that any subjective decisions
made during data analysis were "blind" to the treatment
decision. The experimenters recognized that the enormous
natural variability of summer showers required protection'
against bias together with statistical tests of a postulated
association between seeding and altered rainfall. The sta-
tisticians were integral components of the project from
earliest conception through the many evaluations and
reevaluations of results, although none actually went out
in the field.
Some practical-minded meteorologists have alleged
that Whitetop "failed" in that it did not show that the
taq~et rain on seeded days exceeded that on unseeded
days. Actually the evidence summarized by Braham is
persuasive that seeding probably decreased target
(plume) rainfall when the entire experimental sample is
lumped together. Post-factum subdivision of the sample
by ,cloud population structure (governed to a significant
extent by air-mass source identifiable by wind direction)
sugl~ests that seeding may have increased target rainfall
on IL small population of westerly or northerly wind days