WMOD00478

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

..
64

-

Journal of the American Statistical Association, March 1979

2. Evaluation of Differences in Average Precipitation in "he Ten-hour Period before the Scheduled Start
of Seeding for Seeded and Not-Seeded E-days8

I Inches of rain per period
% Periods with rain Inches of min per wet period (wet or dry) .
% % %
S NS change pb S Ne- change P S NS change P
"
Near-downwind 29 32 -10 0.88 0.136 0.1:m +13 0.81 0.039 0.038 +1 0.98
Far-downwind 39 47 -16 0.57 0.074 0.0!~1 -18 0.60 0.029 0.043 +32 0.40
Near-upwind 21 32 -33 0.33 0.251 0.H8 +41 0.48 0.054 0.057 -5 0.93
Far-upwind 27 49 -45 0.04 0.079 0.1:~0 -39 0.27 0.021 0.063 -67 0.03
Near-left 29 38 -25 0.40 0.101 0.1'.5 -30 0.39 0.029 0.055 -48 0.20 "-
Far-left 52 45 +16 0.60 0.041 0.2:!1 -82 0.00 0.021 0.099 -79 0.00
Near-right 20 34 -42 0.15 0.199 0.0~r5 +166 0.06 0.039 0.025 +54 0.50 ,
Far-right 25 43 -41 0.10 0.068 0.1()3 -34 0.40 0.017 0.044 -61 0.09
. Reproduced from Lovasich et al. (1971), p. 2646. Relates to Project Whitetop.
· Calculeted P-values (lwo-slded).

.

selected because it contained most of the SINS rainfall
differences occurring during the fourteen (fifteen) hours
following scheduled start of seeding. The well-known
phenomenon of persistence in meteorological variables
leads us to expect this partition also to contain most of
the rain differences for the ten hours immediately prior
to seeding. Moreover, relatively few of the location cells
show "significant" P values; these are scattered through
the three nonindependent rain statistics and presumably
could not have been specified in advance-thus are sub-
ject to the -problems of multiplicity (Tukey 1977).
Finally, to accept these conclusions makes it impossible
to accept the logical and coherent physical picture of
Whitetop results set forth in an earlier paragraph.
This is an example of a case in which some statisticians
and some meteorologists, starting from the same point,
have reached different conclusions because of their
differing points of view. It is unfortunate that this issue
has for so long remained unresolved. It illustrates the
crucial role played by the judgments of specialists. But
when a field is so small that it attracts only a handful of
specialists, their views sometimes can take on an aura
of unchallengeable authority that can be disadvantageous
to the scienc~. We have witnessed this more than once in
weather modification.
Whitetop was undertaken as a definitive experiment.
AB it turned out, it should be regarded more as an ex-
ploratory experiment-one raising more questions than
it answered. From'the standpoint of establishing knowl-
edge about cloud seeding, it should be repeated.. But this
seems improbable because of the ethical issues raised in
repeating an experiment that gave evidence for rainfall
decreases in a region where summer rain usually is barely
sufficient. This situation is reminiscent of the recent
article by Tukey (1977) discussing parallel concerns
about clinical trials in medicine.

4. OTHER PROJECTS

4.1 Climax I and Climax II

We now turn our attention to a few other projects
selected to illustrate some of the seeding studies that have

been carried out. The first of these are three projects
inte:nded' to increase the amount of snow released from
winter storms crossing the mountains of Colorado. The
first two, Climax I and Climax II, were carried out in the
vicinity of Climax, Colorado, by Colorado State Uni-
vernity Professor Lewis Grant (Grant and Mielke 1967;
Grant, ChaplX'll, and Mielke 1968; Mielke, Grant, and
Chll~ppellI971). These experiments used a target/control
arefL design, Experimental units were 24-hour periods
which were independently forecast to produce at least
0.01 inch of precipitation at Leadville, Colorado, 15 miles
from the target. Seeding was carried out from ground-
based generators located upwind of the target. Both
projjects employed restricted randomization to ensure
that large blocks (20-40) would have equal numbers of
seeded and not-seeded experimental units.
Climax I, conducted from 1960 through 1965, gave an
overall increase in snow but with weak statistical support.
Post-factum partitioning of the data suggested that snow
increases had come mainly on days with upslope winds
and 500 millibar temperatures warmer than about
-2:30C. Snowfall decreases from seeding were suggested
whe:n the 500 mb temperatures were colder than about
- 2'70C. Since the 500 mb level was identified with the
toP!! of the clouds, the - 270C temperature was considered
to be the temperature of adequate natural nucleation,
beyond which seeding resulted in overseededconditions.
Seeding results were interpret~d against a simple pre-
cipitation growth model.
R.ecognizing the risks of conclusions drawn from retro-
spe(:tive partitions, Professor Grant repeated the experi-
melllt for another five years. With minor exceptions, the
second experiment (Climax II) confirmed the results of
the first, but with stronger statistical significance.

4.2 San Juan Project

On the basis of the Climax experiments, and others not
summarized here, the Department of Interior, Bureau of
Reclamation designed a project to demonstrate the
feasibility of winter snow enhancement on western moun-
tains through seeding during periods of suitable upslope