WMOD00440

• Title: HIPLEX-1: Statistical Evaluation
• Date: 4/4/1984
• Weather Modification - Doc Type: Report

APRIL 1984

MIELKE ET AL.

521

TABLE 9. Univariate P-values for secondary response variables
FCC5 and MAXZ.

P-value

~

Cloud class Values FCC5 MAXZ
A-I Raw data 0.200 0.486
A-I Ranks 0.200 0.657
B Raw data 0.117 0.203
B Ranks 0.159 0.225
A-I and B Raw data 0.653 0.321
A-I and B Ranks 0.660 0.205

the nonseeded and seeded experimental units for either
of these response variables. In fact, none of the sec-
ondary response variables indicated any obvious dif-
ferences between nonseeded and seeded experimental
units.
The point estimates and exact one-sided P-values
associated with the final test statistics PCP A (based on
TIPA), PCPR (based on TIPR), A VRA (based on
AER), and A VRC (based on RERC) are given in Table
10. Even though some of the seeded to nonseeded
rainfall ratios are either extremely small or extremely
large, these results may not be important because none
of the corresponding P-values (except that for the
A VRA rain ratio associated with Class B clouds) is
very small and/or the sample sizes are also very small.
In addition, the rain ratio for both Class A-I and Class
B clouds is dominated by a single case, and II of the
17 zero values associated with the average rain amounts
determined from RERC are due to default decisions.
The corresponding results for A VRC must be inter-
preted with these points in mind. The AER univariate
P-values (Table 6) and the A VRA P-values (Table 10)
suggest an increase in precipitation due to seeding for
Class B clouds, even though much of the details of the
mechanism in the physical hypothesis and the locations
and times of observations are open to question.

5. Discussion and conclusions

The differences found between nonseeded and
seeded experimental unit responses during the first 5-
min period following the seeding treatment (viz., CIC2,
CIC5, and CCR5) were substantial. Thus, a strong
inference can be drawn that the HIPLEX-l seeding
treatments succeeded in producing ice crystal plumes
more rapidly than they would have appeared naturally.
The P-values for the multivariate responses in Figs.
3 and 4 are roughly monotonically increasing, espe-
cially beyond the initial5-min period. This fact suggests
that the strong results for the initial 5-min period fol-
lowing the treatment are increasingly diluted by the
addition of the subsequent responses specified by the
physical hypothesis. The experiment was therefore not
successful in following the chain of events through the
subsequent development of earlier and increased pre-
cipitation in the seeded clouds. However, there was
some evidence of earlier precipitation in one type of
cloud (Class A-I) and of increased precipitation in
another type (Class B).
It should be emphasized that the sample sizes of
this experiment are extremely small (7 Class A-I clouds,
13 Class B clouds, and no Class A-2 clouds) due to
the premature termination of the experiment. There-
fore, it cannot be determined whether the absence of
significant differences after the 5-min mark is due to
the lack of a seeding effect, inadequacies in the physical
hypothesis, or the small sample sizes. Besides the small
sample sizes, possible reasons for the lack of significant
differences in responses beyond the initial 5-min period
include the lack of conditions conducive to further
growth of the artificial precipitation embryos. Cooper
and Lawson (1984) investigate this possibility. They
find that many of the clouds treated had very short
lifetimes, which prevented the formation of precipi-
tation in them and thus precluded significant statistical
results for the post 5-min variables.

TABLE 10. Point estimates and exact one-sided P-values involving PCPA (based on TIPA), PCPR (based on TIPR),
A VRA (based on AER), and A VRC (based on RERC).
Class A-I and B
Class A-I Clouds Class B Clouds Clouds
Measure S NS S NS S NS
PCPA Proportion with rain 3/4 1/3 4/8 2/5 7/12 3/8
Exact P-value 0.371 0.587 0.325
PCPR Proportion with rain 0/4 0/3 2/8 115 2/12 1/8
Exact P-value 1.00 0.685 0.656
AVRA Average rain (mm km2) 0.077 0.879 1.81 0.067 1.23 0.372
Ratio (SINS) 0.088 26.9 3.32
Exact P-value 0.714 0.078 0.237
AVRC Average rain (mm km2) 0.00 0.00 0.480 0.0024 0.320 0.0015
Ratio (SINS) 200 213
Exact P-value 1.00 0.359 0.347