WMOD00404

• Title: Thailand Applied Atmospheric Research Program - Final Report - Volume II
• Date: 3/1/1994
• Weather Modification - Doc Type: Report

The situation for the experimental units is more complicated because of the small Texas and
Thailand samples. In Thailand, the current best estimate of the seeding effect on the
experimental units is 0.66 (i.e., SR = 1.66). Examining the appropriate portion of table 2.1,
one can see that detecting this seeding effect will take about 125 randomly-selected
experimental units at a one-sided significance level of 5 percent at a power of 0.80.
Obtaining such a sample during the five years that have been allotted to Phase 2 should be
possible. This possibility assumes, of course, that the current best estimate of the seeding
effect for the unit is representative of the true seeding effect.

In Texas, the apparent seeding effect OIlL the experimental units by 150 minutes after initial
seeding is on the order of 0.27 (i.e., SR = 1.27). Thus, the calculations presented in table 2
suggest that as many as 400 random cases could be needed to detect this effect at a 5-percent
significance level at a power of 0.80. If this set of computations is representative of the real
situation in Texas, obtaining such a sample size using a raw "brute force" approach might
never be practical without the benefit of covariates.

Fortunately, the use of covariates can reduce the number of cases needed to establish an
effect of seeding. It is well-known, for example, that a covariate that is correlated with the
rain volume in the experimental unit by coefficient R will reduce the required number of
random cases N to N(R) according to:

N(B) = (1~R2) N(SR)

To illustrate how this fonnula works, assume that an identified variable is correlated with
rain volume in the experimental. unit by 0.70 (i.e., R = 0.70). Then, the 125 cases that were
cited above as necessary to resolve a seE~ding effect in Thailand of 0.66 decreases to about 64
cases. Therefore, identifying covariates in cloud seeding experiments has great benefits. This
technique should be a major area of research during the Thai demonstration experiment.

2.11 Evaluation

2.11.1 Evaluation parameters

The evaluation of the results of the demonstration project will be similar to that described
in appendix B of this report. The focus will be on the individual cells and on the
experimental unit that contains the cells. The specific evaluation parameters for the cell
analyses will include S versus NS ratios of mean cell heights, reflectivities, areas, durations,
rain-volume fluxes, lifetime rain volumes, and the number of cell mergers. For evaluation
ofthe experimental units, the parameters will include ratios of mean "focused-area" and unit
echo areas, durations, and rainfalls. The focused-area analysis is explained later in this
section. All analyses will proceed with and without partitioning by cloud-base temperature.

2.11.2 Approach

The results cannot depend on the approach to the analyses, as was the case for the analysis
of the exploratory experiment. Further, seeding effects should be indicated in most of the
analyses; they should be consistent with the conceptual model, and some ofthe results should
be statistically significant after accounting for some of the natural variability. In essence,
therefore, the analysis process must not be a mathematical exercise. All results must be
plausible, reasonable, and physically consistent if they are to be believed.

15