WMOD00404

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

2.10 Projected Duration of the Demonstration Project

The criterion for success of the demonstration project will be plausible, physically based, and
statistically significant indications of increased rainfall from the convective cells that received
AgI treatment and from the experimental units that contained the AgI-treated cells. Based
on the results of the exploratory experiment, the criterion for success is expected to be
satisfied first for the convective cells and then for the experimental units that contain the
cells. The reasons for this expectation are twofold. First, the effect of treatment will be
larger for the cells than for the experimental units because the cells receive the AgI. Second,
the cell sample size will be larger than the sample of experimental units because more th:m
one treated (i.e., AgI or simulated Agl) cell is obtained within each experimental unit.

These expectations are supported and quantified by calculating the number of random cases
(i.e., experimental units) and cells that will be required in Thailand to detect a given seeding
effect at the 5-percent significance level at a given power. The results are based on 2,000
permutations or rerandomizations of the ratios of S to NS rainfalls. For comparison
purposes, the same calculations were run for data obtained in Texas. The computations were
made by Dr. Ronit Nirel of the Department of Statistics at the Hebrew University of
Jerusalem.

The required sample size, N, is calculated according to the following relationship:

N = (Z u + Z P )2 X ( cJ2 / 02)

where Zu is the (1 - a) cumulative fraction of a standard normal distribution; Zp is the (l -
13) percentile of a standard normal distribution; a is the significance level of a one-sided test
that excludes a priori a negative seeding effect (a = 0.05); 13 is the probability of concluding
that no effect of seeding exists when an effect actually exists, and 1-13 is the power of the
experiment, which is defined as the chance or probability of detecting an effect of seeding
when an effect of seeding actually exists; cJ2 is the variance of the SR (single ratio) ofS (seed)
to NS (no seed); 0 is the seeding effect (SR-1); and SE(SR) is the standard error of the
estimate of SR.

Note that a large variance of SR and a small seeding effect means a large N. The
calculations of N for cells and experimental units in Thailand and in Texas are provided in
tables 2.1 and 2.2, respectively.

A power of 0.80 was used in these calculations because of the small sample size from which
to calculate SR and its resulting high variance. The variance will likely decrease as the
sample is increased during the demonstration project, and project duration will be re-
estimated using a power of 0.90. This method should provide a better estimate of the
required project duration, one that should be oflesser duration than that projected currently.

Care should be exercised in interpreting these calculations. However, reaching a definitive
conclusion for individual cells in both Texas and Thailand should be relatively easy because
the observed DELTA is > 1.0 in Texas and also in Thailand for the warm cloud base (i.e., >
16 OC) partition. Thus, one might reasonably expect to reach statistical significance for the
Thai cell experiments in only one or two more seasons of experimentation, mainly for cells
with base temperatures> 16 oC.

14