WMOD00478

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

.74

Journal of the American Statistical Association, March 1979

The suggestion made by several meteorologists is that
the seeding at Whitetop was too early and/or too heavy,
producing an early concentration of droplets that dis-
sipated the energy of the cloud or created early afternoon
shading of the ground so that the warming whicli is part
of an afternoon thunderstorm did not materialize. There
is an indication of a second buildup of precipitation in the
early morning on the day following a not-seeded experi-
mental day, rising and falling before the start of any
possible seeding on the following day. Why seeding should
inhibit this second buildup seems even more difficult to
explain.
We do not have physical measurements that would
allow investigation of the hypothesis that too early and
too many fine droplets dissipate the energy of the cloud.
We did study the suggestion of Hughes (see Lovasich
et al. 1971b) that these water droplets form clouds that
shade the ground, thus reducing its temperature. If this
hypothesis were correct, we would expect to find more
cloudiness in the early afternoon on seeded E-days than
on not-seeded E-days. Also, the afternoon temperature
on seeded E-days would be lower than on not-seeded
E-days. Eight U.S. Weather Bureau stations within the
area provide observations of cloud cover and of tempera-
ture every three hours. These observations contradict the
suggested cloudiness hypothesis: the cloudiness is less on
seeded E-days than on not-seeded days, and the tempera-
ture is higher.
Braham describes his moving-target method of analysis
in an earlier article (1966). This idea can be extended to
a moving grid, set up in various ways. We have used the
upper winds at the level and time of seeding to define the
moving grid, shortened to mogrid, and have also em-
ployed grids corresponding to other wind levels and times.
Two mogrids are shown in Figure D, corresponding to
frontal days and to air-mass days. In this case, there are
450 sectors ~nd the rings are each 60 miles wide. The
entries are the percentage increase (negative entries are
thus decreases) observed in that wind direction and
distance averaged over all experimental days in that
stratum. The numbers in the squares are corresponding
entries for rainy days. The actual gages in a given sub-
area tend to vary from day to day; each rain gage is used
each day of the stratum, once and only once. The second
entry in each subarea of the mogrids in Figure D is' the
two-tail significance probability. We note that on air-
mass days, the typical entry is negative and appreciable,
accompanied by a small significance probability.
Figure D illustrates that, for the air-mass category of
experimental days, every percentage increase is negative
and many regions show a significant decrease, especially
the regions on the right and those downwind. For the
. frontal days (shown in the upper part of Figure D), there
are no significant effects. There is a tendency for in-
creasingly positive results toward the right downwind.
Evidence of increase right downwind, when the over-
whelming effect is negative, is also shown in the corre-
sponding analysis of the Arizona randomized experiment

[.t. Mogrld Comparison 'of Apparent Eflects of Cloud
SeEldlng on Frontal and on Air-Mass Days (Entries:
Percentage Apparent Increase In 24-hour Precipita-
tion and Two-Tall P)

DAY'S WIND DIRECTION

Bot.:TClI: Lovuich et al. (11171.)

,
,

of Hattan (see Neyman, Scott, and Wells 1973). Many
out(:omes of the Whitetop experiment, such as extended-
area. and extended-time effects, apparent decapitation of
the afternoon thunderstorm buildups, and the overall
negfLtive results, are verified in the Arizona experiment
(see Neyman et al. 1972). The search for explanations
requires the cooperation of meteorologists and
statiisticians.

4. SEARCH FOR TIME OF EFFECTS

In our studies of the cloudiness-temperature hypothesis
as an attempt to explain the lack of afternoon sho'Y-er
buildup, the data used were published at intervals of
three hours which led us to perform analyses of the effect
of Reeding every three hours, rather than hourly or daily,
as before. This was the first time we analyzed precipita-
tion in a period before the start of seeding. One of us
noti,ced that there_ was an apparent effect of seeding in
certu.in strata before seeding started. We then looked into
this question in more detail. In every case, we used all of