WMOD00478

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

92.

.

~e met with a difficulty. It appeared that the number of
experimental days with winds outside the "must" range
'was relatively small. Therefore, in order to have a
reasonable chance of detecting even a substantial differ-
ence in the pattern of the apparent effects of seeding, some
kind of compromise stratification appeared necessary.
Several such compromise stratifications were made. All
of them, including a cross-stratification, are described in
Lovasich et 801. (1969; 1971a,b). However, a pictorial
. illustration of the evaluation is given for only one of them,
labeled East days vs. West days (Lovasich et a1. 1971a).
The striking results are that the East days (but not the
West days) were marked by unbelievably large and un-
believably widespread deficiencies of seed-day rain as-
cribable to seeding. -

As mentioned, all the above results were obtained in the
course of studies undertaken in response to questions
posed by what we found in the "Final Report" of Profes-
sor Braham. No explanatory idea came to our minds, and
our comment (Lovasich et a1. 1971a, p. 151) was: "The
reported results reflect facts thai. happened j plausible
explanations may be expected from specialists in cloud
physics and meteorology."
An intelligible tentative explanatory hypothesis was
suggested to us by Mr. James Hughes, an atmospheric
physicist at the Office of Naval Research. This hypothesis,
termed the cloudiness-temperature (or C-T) hypothesis,
visualizes the possibility that beginning the Whitetop
seeding in the late mornings generated widespread cloudi-
ness and the subsequent widespread decrease in tempera-
tures. This, in turn, could diminish the thunderstorm ac-
tivity and thus the rainfall. In order to verify the C-T
hypothesis, it was necessary for us to examine the area-
wide' diurnal variation in cloudiness, temperature, and
rainfall. In other words, we were faced with the necessity
of timing the apparent effects of seeding on East days
(and separately on West days), this over the whole big
area, up to 180 miles from the Whitetop target center.
Specifically, the question was whether at the beginnirig
of seeding on East days, or soon thereafter, there occurred
a marked change in the area-wide pattern of the three
weather parameters, the cloudiness, the ground tempera-
ture, and the rainfall. To answer this question, we had to
use the available data for a substantial period of time
before and also after the scheduled beginning of seeding.
Also, it was judged important to take into. account the
directions of winds and the distance from the center of
the research area.
The diurnal change in cloudiness and temperature
could be studied on data from only eight weather stations
in the area at which the relevant observations were made
and published. On the other hand, the diurnal variation
in rain could be studied using the data of 104 recording
gages. Here, then, the distinction was made between
"near" areas, up to 90 miles from the seeding center, and
"far" areas, from 90 to 180 miles. For each experimental
day, we covered the period fmm the preceding midnight
and for 36 hours .thereafter.

~~:.~.~~~~~~~,._..L- .

Journal of the American Statistical Ass,ociation, March 1979

The results of 0UI' study were published in Lovasich
et a1. (1971a). As anticipated, the findings for West days
did not, exhibit marked seed/no-seed differences. On the
other hand, the findings for East days were striking. They
flatly contradicted the C-T hypothesis. The seeded East
days were less cloudy and hotter t.han controls right from
the' preceding midnight. Seed day rain deficiencies also
began in the early mornings. Very unexpectedly, the most
pronounced deficiency of seeded precipitation occurred in
"far" a.reas, from 90 to 180 miles from the center of seed-
ing, aUld was particularly regular in the "far" upwind
areas. It is this circumstance that caused us to examine
the significance of these deficiencies of rain that occurred
during the ten hours preceding the scheduled commence-
ment of seeding. OUf results !Lre reproduced by Braham
in his Table 2. Our interpretation of all of the above
findings was, and continues to be, that for some reason or
other, the category of East days with seeding had less
than its fair share of days with bad weather.
Our tests indicated that strict randomization could
produce the observed deficiencies of seed day precipita-
tion, but very rarely. This suggests doubts about ran-
domizB~tion. Our doubts have a qualitative, indirect
backing. The general pattern of winds in the area sug-
gests that, frequently, the faraway upwind area covered
the Ozark Plateau, occasionally described as the "breed-
ing ground" of big, 'widespread storms. For the benefit of
light planes, forecasts of such storms are regularly given
over the radio, and the pilots employed by the Whitetop
project are likely to have been aware of big approaching
storms. Furthermore, as reported by Braham (1966, p.
13), the envelopes with dated decisions "go" or "not go"
were not in Braham's custody, but were left with the
personnel in the field. I need not continue.
However, it is not impossible that Braham's randomiza-
tion was faultless. But, as he rightly states, a given day's
weather exhibits a degree of continuity. Thus, if pure
chance produced a brreat deficiency of rain in an upwind
area during early morning, it should not be surprising to
see' similar deficiencies in the afternoon and later. Such
considerations motivated our warnings that any con-
clusions about the effects of Whitetop cloud seeding must
be made with extreme caution.
At this point it gives me great pleasure to compliment
Professor Braham on his methodological invention, which
I value very highly. T_his invention is published in
Braham (1965), but is not mentioned in the article now
discussed. The invention has to do with the problem of
farawl~y effects of local cloud seeding. To my knowledge,
Braham used it only with radar echo data. Being pre-
domin.antly interested in precipitation that reaches the
ground, we had to adjust Braham's idea to rain gage
data, and our description of the resulting methodology
is "moving grid" or "mogrid" methodology. Its descrip-
tion, crediting Braham, is given in Lovasich et a1. (1971a).
Certain paragraphs of this discussion exhibit my
skepticism about statistical tests used by Braham, oc-
casionally with great confidence in their validity. The