WMOD00412

• Title: Final Report on Utah Cloud Seeding Experimentation Using Propane During the 2003/04 Winter
• Date: 3/1/2005
• State: UT
• Weather Modification - Doc Type: Report

I

I

precipitation gauge SWE observations for the population ofEUs. All four used the crosswind control
gauge to reduce natural variance, an approach specified before the experiment was conducted.

I

Three statistical testing methods were specified prior to data collection. These were selected to explore
different means of objectively measuring potential seeding effects. These methods are referred to as pairs
testing, linear ranking of precipitation amounts (Wilcoxon test), and mean double ratio testing. Least
squares linear regression analysis was applied only to the entire EU population to demonstrate that even
this simple and widely used test could provide a strong statistical suggestion of seeding effectiveness at
the primary target gauge.

I

I

The three more sophisticated statistical tests were also applied to the total nonpartitioned population of
98 EUs, or 47 pairs for the pairs test. The traditional but somewhat arbitrary a.-value used to define
"significance" in weather modification evaluation is 0.05. Because it is assumed that seeding cannot
decrease precipitation, the tests were I-tailed. Statistical tests giving probability values (P-values) ~ a. are
termed significant. A P-value equal to a. = 0.05 means there are 5 changes out of 100 for the apparent
seeding increase to be due to chance rather than a real effect. The lower the P-value the more significant
the result, indicating greater likelihood of having detected an actual seeding effect. When applied to all
98 EUs, the pairs and ratio tests resulted in significant P-values between 0.03 to 0.06 for the three target
gauges GSC, located midway from the HAS to the TAR GTR near the TAR, and GSO 1 kIn due south
(crosswind) ofGTR. These three are referred to as the "core" gauges, located within 3 kIn or less of one
another. The less sensitive Wilcoxon test indicated a P-value of 0.05 at GSC and values of 0.1 0 at both
GTR and GSO. The downwind gauge GDN had less encouraging results, with two tests somewhat
suggestive of seeding effectiveness, but not meeting the 0.05 criterion. These tests, applied to the entire
EU population, in total provide a strong statistical suggestion that propane seeding increased the
"seasonal" 3.5 month experimental period snowfall on the windward slope and upwind portion of the
plateau top.

I

I

I

I

The most obvious and physically reasonable partitioning of the available EUs was by seeding site wind
direction. No real seeding effect can exist if the wind did not transport the seeding plume over the target
gauges. Considerable prior plume tracing with tracer gas and AgI had been conducted with the same
seeding site and primary intended target, TAR and associated gauge, GTR. These earlier data sets have
been analyzed in detail for this experiment. They showed that plumes released at the HAS when winds
there were from about 170 to 270 deg routinely reached the GTRlT AR area. Plumes released when HAS
winds had any northerly component did not target that area but were transported further south. The
obvious wind partition was for HAS median wind directions of 270 deg true or less, as the lowest EU
value was 169 deg.

I

I

I

This wind direction partition provided a total of32 pairs, and a total of35 seeded plus 34 nonseeded or
69 EUs. The pairs and ratio tests provided strongly suggestive P-values between 0.02 and 0.05 for the
core gauges with the exception of one GSC P-value of 0.09. Wilcoxon P-values were somewhat higher,
which proved to be the usual case, ranging from 0.11 to 0.15 for the core gauges. In this case the
downwind gauge GDN had somewhat suggestive P-values of 0.06 and 0.09 by the ratio and Wilcoxon
testing, but was inconclusive by the pairs test. The core gauges had mean residuals equivalent to 0.012 to
0.015 inch h-I, averaging 1.5 times as much mean SWE per EU as suggested by testing the entire
population. These results were shown to indicate that the entire seeding effect at the core gauges was
contained within the wind direction partition. This is a physically reasonable result which would be
expected if targeting was limited to southwest quadrant winds at the HAS, as demonstrated by prior
plume tracing studies. If the seeding effect was not concentrated in this wind direction partition, doubts
should be raised about the reality of the statistical suggestions. Conversely, finding the seeding effect
concentrated within this partition lends considerable confidence to the statistical results.

I

I

I

I

It was hypothetically estimated that if all hours containing the 69 EUs had been operationally seeded,
the overall precipitation increase would have been 8% for the 3.5 month experimental period. This result

I

III

I
I