Laserfiche WebLink
<br />conducted during a generally dry period which <br />received only 69% of the long-tenn average SWE for <br />December :!OO3 through \-larch 2004 as monitored by <br />a nearby Sl'\OTEL snow pillow. A total of98 EUs <br /> <br /> <br />were obtained. of which 94 were paired in 2 hour <br />blocks. The remaining 4 had their paired EU rejected <br />because of very light winds, resulting in unreliable <br />estimation ofplun}(' transit times. <br /> <br />- <br /> <br />I <br />~ <br />r <br />r;'':~,~~ <br />. I -- --f~" <br />1. \ <br /> <br />I <br /> <br /> <br />~' <br /> <br /> <br />Figure 4 The emlnll propane dispenser and tower allhe High Altitude Se\.-ding (HAS) location with the vie\\ toward the south- <br />southl;'ast v. here the Wasatch Plateau lop is in doud. The propane controllcr and data logger are althe base oilhe tuv.er. <br /> <br />Statistical Tt'stin~ <br /> <br />Three statistical ~thods. all specified prior to data <br />collection. were applied to the precipitation gauge <br />SWE observations for the population of EUs. All <br />three used the cross....ind control gauge to reduce <br />natural variance, an approach specified before the <br />experiment \\,as conducted. The mcthods were <br />selected to explore different means of objectively <br />measuring potential sel-ding effecls. They are <br />referred to as pairs testing. linear ranking of residual <br />preCIpitation amounts (Wikoxon test), and mean <br />double ratio (\-!DR) testing. All three tests provided <br />impressive and often similar results as discussed by <br />Super and lIeimbach (:!OO5a. :!OO5b) with the <br />Wilcoxon being somewhat less ~nsiti\e as expected. <br /> <br />lIigh sensitivity and indusion of confidence <br />mternls gives the MDR approach advantages. In the <br />interests of brevity and simplicity of explanation <br />most results presented herein will be from MDR <br />testing_ Ilowever. reference will be made to "mean <br />residuals" calculated by the pairs method because <br />they provide estimates of the mean difference in <br /> <br />" <br />, <br /> <br />SWE between gi\'Cn groupings of seeded and <br />nonseeJed EUs. <br /> <br />The traditional but somewhat arbitrary {J-\.alue used <br />to define "significance" in weather modification <br />evaluation is 0.05. Because il is assumed that <br />seeding cannot decrease precipilation. the tests were <br />I-tailed. Statistical tests gi\"ing probability values (P- <br />\-alues) ~ {J are teoned signilicam. A P-\"alue equal <br />to n = 0.05 means there are 5 changes OUl of 100 for <br />the apparent seeding increase to be due to chance <br />rather than a real effect. The lower the P-\'alue the <br />more significant the result. indicating greater <br />likelihood of having detected an actual seeding <br />effect. <br /> <br />When applied to all 98 EUs. \!IDR testing pro\'ided <br />the results shown in Table I where targel gauges are <br />ordered by increasing distance from the IIA$ seeding <br />site. Each target gauge is well correlated with the <br />G1'\O control gauge observations so much of the <br />natural variance was eliminated by using G~O as a <br />co\-ariate. Significant P.\'alues between 0.03 10 0.06 <br />are seen for the three targel gauges GSC. located <br /> <br />6 <br />