WMOD00478

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

Iraham: Field Experimentation in Weather Modification

65

mented. Neither of these facts could have been learned
hadthe project not employed randomization.

winds and cloud-top temperatures. A substantial effort
went into the experiment design. Advice was sought from
many statisticians and meteorologists, and nonscientific
. sectors also had much to say about the final design. In due
course, a project was carried out over a portion of the
SanJuan Mountains of Colorado during five winter
seasons between 1970 and 1976. Although the final report
of the project has not yet been published, many details
are available from a very comprehensive contractor re-
port by Elliott et al. (1976).
The design called for unrestricted. randomized seeding
during 24-hour periods forecast to have suitable winds and
cloud-top temperatures equal to or exceeding -250C.
There were 147 experimental days, 71 seeded and 76 not-
seeded. Silver iodide nuclei were released from ground
generators upwind of the target, the number and location
of generators depending on wind direction and speed. An
attempt was made to tailor the generator output ac-
cording to anticipated cloud-top temperatures, with
higher output.s on days with warmer clouds. Apparently,
cloud-top temperatures were not often measured directly
but were computed from data obtained from project
radiosondes released upwind of the target.
There were eighteen precipitation stations in the tar-
get area for which data were available for all five years..
The ratio of seeded- to not-seeded-day precipitation at
individual stations ranged from 0.7-1.01, with 16 of 18
being less than 1.00. Application of a rank-sum test led
to the conclusion that there were no statistically signifi-
cant differences in precipitation attributable to seeding.
According to the Elliott et al. report, apparently no effort
was made to obtain an overall seeding result for the entire
target. .
A number of post-factum studies were carried out. One
of these indicated that up to 22 of the control days might
have been seeded as a result of carry-over of seeding
material from the previous (seeded) day or from another
project upwind of the target. When these 22 days were
removed, the treatment effect ratios at individual stations
ranged from 0.94-1.51, with 16 of 18 greater than 1.00.
The rank-sum test applied to individual stations again
indicated no statistically significant difference. Another
ex post facto study showed that cloud conditions actually
changed substantially within individual experimental
periods and that seeding had not been limited to clouds
meeting the suitability criteria. Therefore, the seeded
and not-seeded periods were divided into 6-hour blocks
and stratified according to cloud-top temperature and
air-mass stability. This analysis suggested substantial
snowfall increases from seeding during stable thermo-
dynamic conditions and cloud-top temperatures between
-90C and - 290C. Colder cloud tops and most conditions
of deep convection seemed to be associated with snowfall
decreases.
Scientifically the San Juan Project appears to have
been a suc. cess in that it verified the scientific hypotheses
I
from Climax I and Climax II. Operationally, it was a
failhre because the design was not effectively imple-
!

~'4:

~,;:,.~____.i';':<,~L,:,_.".~ .cJ--.-.lo;._"i...'

4.3 I~roject FACE

Project FACE (Florida Area Cumulus Experiment) of
the National Oceanic and Atmospheric Administration is
an experiment to determine whether rainfall from large
cumulus clouds can be increased through application of
the dynamic seeding strategy described in Section 2.
Phase I, the exploratory phase, was carried out during
five summers between 1970 and 1976 (Simpson and
Woodley 1971, 1975; Woodley and Sax 1976; Woodley
et al. 1977).
This project considered both a geographically fixed
target and wind-borne (floating) targets within it. Seeding
was randomized by day. Experimental days were selected
on the basis of certain suitability criteria involving radar
observations and calculations of a numerical model. The
model was used to predict the amount of cloud-top
growth expected from the glaciation of an individual
cloud turret. The design called for seeding individual,
growing cloud turrets near their tops by dropping into
them pyrotechnic flares containing silver iodide. l3Y
glaciating the updraft region of many adjacent cloud
turrets, the experimenters hoped to assist those clouds
to organize their convective motions and to merge into a
larger cloud system. Growth by mergers of many smaller
clouds has long been recognized as a pattern of develop-
ment of convective rainstorms. As yet, there are only
very crude theories to account for this behavior of natural
clouds, much less to anticipate changes in such behavior
as a result of seeding.
The primary response variate was surface rainfall cal-
culated from radar measurements, with and without an
adjustment factor obtained from clusters of surface rain
gages. As a result of local cloud climatology, most of the
experiments were conducted during mid to late
afternoons.
The seeding decision was determined through a random
process after suitable clouds were observed to be present
within the target area. The seeding decision was not
communicated to the scientists responsible for judging
cloud. suitability. Final acceptance of an experimental
day depended on the development of six or more suitable
cloud.s, or of clouds suitable for the expenditure of 60 or
more seeding flares, or both. Application of these selection
criteria led to 39 seeded days and 36 not-seeded days
spre:ll.d over five summers.
AJlthough there were several changes in the experiment
as it progressed, only two have thus far been identified
as a,ffecting the. results: (1) a change in the type of
seeding flares used (1975), and (2) a change in the
number and types of airplanes used for seeding (1976).
However, the analysis underwent continual evolution
both in overall philosophy (Bayesian, significance testing,
exploratory data analysis) and in the partitions and co-
variates considered. A comprehensive reporting of FACE
was given at a 1977 scientific meeting, .from which the