Laserfiche WebLink
<br />,.."~ <br /> <br />Further discussion of such features is beyond the scope of this report, but operational <br />programs do not incorporate them, and so are much more difficult to statistically evaluate <br />with any degree of confidence. Readers interested in the details of statistical design of <br />experimental programs may refer to such sources as Dennis (1980), Gabriel (1981), and the <br />Weather Modification Advisory Board (1978). <br /> <br />As pointed out by Dennis (1980), some statisticians have argued that the uncertainties in <br />operational cloud seeding programs make all analyses ofthese programs essentially useless. <br />However, others hold the view that such analyses can provide useful information if potential <br />pitfalls are borne in mind and care is taken to avoid insupportable claims. <br /> <br />This author maintains that operational programs can produce useful information, and has <br />supported this point of view by undertaking the analysis to be presented. However, the <br />reader should not expect anything approaching "scientific proof' from statistical analysis of <br />an operational program, and certainly not from the Idaho program at this time because <br />seeding was conducted for less than a single winter. At best, the analysis will produce some <br />suggestions or "hints" of possible seeding effects. An expectation of more specific results is <br />unrealistic, and the reader should interpret any statistical suggestions with care. No <br />justification exists to claim any level of statistical significance for suggested results from a <br />single season of operational, nonrandomized seeding. <br /> <br />3. OVERVIEW OF TARGET-CONTROL EVALUATION <br /> <br />3.1 Choosing Measurement Sites <br /> <br />The most common approach used to evaluate operational seeding is the "target-control" <br />method, which will be used here. In the case of the Idaho program, this method involves <br />comparing snowfall observations from two or more areas. The "target" is that area intended <br />to be affected by seeding with the purpose of increasing its snowfalL Throughout this report, <br />target areas are as defined by NA WC in their reports to the program sponsors. The "control" <br />refers to one or more measurement locations (sites) assumed to be unaffected by seeding. <br />Control areas should be located upwind or crosswind from the target to increase the <br />likelihood that they remain unaffected by seeding. Control sites may be chosen over a <br />widespread region so long as their snowfall measurements have a reasonable relationship <br />with similar measurements made in the target area. <br /> <br />Control areas are usually chosen near target areas to increase the degree of association <br />between target and control snowfall measurements. However, such a choice often involves <br />a "trade-off' between selecting controls close to the target, and increasing the risk that <br />. seeding will contaminate the controL Any such contamination would be unknown in the <br />absence of special measurements (e.g., silver-in-snow concentrations). Contamination would <br /> <br />2 <br />