WMOD00293

• Title: Summary of the NOAA/Utah Atmospheric Modification Program: 1990-1996
• Date: 9/1/1998
• State: UT
• Weather Modification - Doc Type: Report

1.2 Objectives and Research Reported

The research reported herein is based on the last several years of the NOAAlUtah AMP when field work
'was conducted on the Wasatch Plateau (hereafter Plateau) of central Utah, approximately between the
towns of Fairview and Price. Earlier work was conducted on the Tushar Mountains of southern Utah.
That work, which has its own body of publications (e.g., Huggins and Sassen 1990), is not discussed in
this report. The single exception is the paper by Sassen and Zhao (1993), which is quite relevant to the
Plateau findings and was published during the Plateau phase. A review of the earlier years of NOAA
AMP work, with an extensive list of references, is given by Reinking (1992).

The NOAAlUtah AMP had two main objectives. First, the program was designed to physically evaluate
the effectiveness oftheUtah operational seeding program which has been partially funded by the State of
Utah. Second, the program was to recommend to the Utah Division of Water Resources any changes
which might improve future effectiveness of the operational seeding program., The operational
program's goal was to increase the high mountain snowpack, which should lead to springand summer
-streamflow augmentation (Super and McPartland 1993). Numerous findings and recommendations
which could improve the operational program are to be found in the articles and papers summarized
herein. Any decisions to implement such recommendations are the responsibility of the Utah Division of
Water Resources and cooperating local water management agencies which jointly fund the operational
program.

1.3 Problems with Statistical Evaluations

Three statistical evaluations of the operational program have been reported by Thompson and Griffith
(1981), Griffith et al. (1991), and recently by Griffith et al. (1997). The reader is referred to these same
articles for details of the operational program which used valley-basedAgI generators, typically spaced
on the order of 16km apart (Griffith 1996). A small minority of all generators were sited in or ~ear
canyon mouths which could help transport and dispersion over mountainous terrain.

While all three of the statistical evaluations have suggested seasonal snowfall increases in the
10-20 percent range, one should be very cautious about accepting such indications. Griffith et al. (1997)
correctly point out that their (statistical) techniques, "are not as rigorous or scientifically acceptable as is
the randomization technique used in research." A sizeable body of literature exists which discusses the
many difficulties of after-the-fact statistical analysis of operational seedirig programs and why such
attempts should be viewed with caution. Dennis (1980) and Gabriel (1981) discusses some of the
problems that can result from improperly applied statistical approaches. Most of these arguments will
not be repeated here. However, a major problem is the lack of any randomization with operational
seeding programs, considered by many statisticians and meteorologists as essential for valid statistical
testing. Target-control analysis of the type applied to'Utah's operational seeding program must assume
that precipitation relationships are stable over decades. This assumption presents a major difficulty
(Dennis 1980). In the Utah analyses, these relationships are assumed stable over long distances, from
central and southern Utah target gages well into Nevada and Arizonawhere control gages were selected.
But it is well known that precipitation relationships can change over time and space for a variety of
reasons ranging from large scale climatic changes (e.g., EI Nino) to local changes in the environment of a
precipitation gage (e.g., growth of vegetation affecting gage catch).

The three analyses cited above were done by the company hired to conduct the operational seeding.
While it is common practice for cloud seeding operators to analyze their programs, statisticians have
pointed out that this approach can leadto bias, favoring positive results. Dennis (1980) discusses a

2