WMOD00478

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

JL

'2

.

As was clearly pointed out by Flueck, the partitioning
.was arrived at ex post facto, subjecting the results to
criticisms of multiplicity (Tukey 1977). Remember that
we do not base our conclusions about Whitetop on a few
selected tests of significance, but rather on the overall
structure of the data when viewed against a framework of
the physical processes involved. Interested readers are
urged to consult Flueck (1971) for full details of these data
and conclusions drawn from them.
Based on differences between target and control areas
on seed and not-seed days, the overall effect of seeding
in Whitetop was a decrease in both rain and radar echo
cover. The largest decreases appeared in the five hours
I
following seeding. Statistical support for decreases in
rain' was very weak; for decreases in echo cover, the
support was moderate to strong, especially in the post-
seeding hours.
The apparent seeding effect was found to vary with
the maximum heights of clouds on any given day. On
days with maximum echo tops below 20,000 feet, about
-lOoC, seeding effects were negative but with little or
no statistical support. On days when maximum echoes
were between 20,000 and 40,000 feet in height, the target-
control differences indicate positive treatment effects of
+38 to +57 percent in echo with weak support, and
+68 to + 100 percent in rain with strong support. On 85
days when echo tops exceeded 40,000 feet (about -400C),
the indicated seeding effects were strongly negative with
very strong statistical support. On these "tall echo" days,
the largest apparent negative seeding effects and the
strongest statistical support occurred near the seeding
line during seeding hours and 70-110 miles downwind
of the seeding line during post-seeding hours. Since days
with these very tall echoes dominate the rainfall pattern
in southern MissoUri in summer, they overbalanced the
positive effects indicated for days with medium-height
echoes.
Two other physically meaningful partitions are total
number of burner hours and mean wind speed, both of
which give indications of the magnitude of treatmeD:t.
Days with 20 or fewer burner hours (indicative of weak
treatment) gave mixed treatment effects without sta-
tistical support. But on 148 days on which we came close
to the programmed level of 36 burner hours, the seeding
effect was strongly negative, with strong support in echo
and weak support in rain. Note the steady progression
from positive to negative effects in echo with increasing
nmriber of burner hours. Flueck (1971) showed that this
inverse relation between precipitation and burner hours
was restricted to the target area. I
D~ys with winds greater than 24 knots (indicative of
strong dilution) gave positive seeding effects, while on
days with slower winds (less dilution), the effects were
mainly negative. During post-seeding hours, when verti-
cal mixing of the seeding materials should have reached
its maximum, the seeding effects were directly propor-
tional to wind speed: for rain, -92 percent for WS S 12
knots, -60 percent for 12 < WS S 24 knots, and +38

~ ...

Journal of the American Statistical Association, March 1979

percent for W S > 24 knots; for echo cover, corresponding
valu.es were -50 percent, -145 percent, and +6 percent.
Statistical support for these values is weak but generally
consistent. Flueck also showed that large rainfall de-
creases were associated with days with southerly winds,
while rain increases came on days with west winds. He
reported little evidence for changes in the frequencies of
precipitation except on west-wind days, where it in-
creased (about +20 percent) and on south-wind days,
where it decreased (about -40 percent).
One aspect of Whitetop which has received considerable
attention is the fact that both target and control areas
received substantially less rain on days we seeded than
on days we did not seed. This happened in four of the five
years. For the experiment as a whole and for individual
years, except for 1962, the statistical support for this
being due to anything but chance is very weak. Some of
the post-factum partitions showed SINS differences in
both target and control areas with considerable statistical
strength.
These statistical findings are to be viewed in the con-
text of several results from the physical studies: (1) the
unseeded clouds in Whitetop were found to have sur-
prisingly high concentrations of natural ice crystals, fre-
quently entirely adequate for effective precipitation
formation, and occurring at temperatures warmer than .
the threshold of AgI nucleation (Koenig 1963; Braham
1964); (2) our AgI nuclei, which we released high in the
subcloud layer, mixed vertically much more slowly than
we had anticipated, and were barely detectable at the
surff~ce at distances of 30-40 miles from the seeding line
(Bourquard 1963); (3) cumulus cloud studies by other
groups have shown that most of the air passing up through
the base of a cumulus cloud comes from fairly close to the
surfB~ce, while relatively little comes from high in the
subcloud layer (Marwitz 1973); (4) application of the
Simpson-Wiggert (1969) model of cumulus cloud dy-
namics to the 6 A.M. upper air temperature and moisture
data predicted maximum heights of unseeded clouds on
seeded days to be about 3,000 feet lower than on not-
seeded days. Actual radar measurements showed maxi-
mum echo heights on seeded days to be about 3,000 feet
higher than on not-seeded days (Braham, McCarthy,
and Flueck 1971).
The results of these physical studies, plus the fact that
we seeded with a heavy dosage rate in the upper part of
thesubcloud region, can be combined with Flueck's
analyses to give the following picture of Whitetop results:

1. Large clouds, because of their ability to stir the
boundary layer by cloud-induced local circulations, most
likely were overseeded, except during periods of high
winds and/or low burner hours, which would act to
reduce the AgI concentrations in the plume. (Recall our
finding that many clouds of this region contained un-
expectedly high concentrations of natural crystals.) The
very large negative seeding effects on days with echo
tops over 40,000 feet are consistent with findings of mQre

",-,'141:.......:.<:.