Statistical Issues in designing a weather modification experiment: A case study

The Wyoming Weather Modification Pilot Project is a multi-year, two-site field experiment which attempts to quantify the effects of orographic cloud seeding on target areas within the North Platte River and Green River basins. The statistical design of this experiment has evolved as a result of the analysis of both historical and recent precipitation data, reviews and comments from reviewers, and pragmatic concerns relating to conducting a field experiment in a remote region. This talk will focus on the statistical issues associated with the design of the experiment. The positive and negative attributes of the Wilcoxon test and the ratio test will be discussed. These tests differ in several respects. For example, the Wilcoxon test primarily relies on responses of individual events while the ratio test typically aggregates responses from individual events across the experiment period. An additional important consideration was the overall design of the experiment. In particular, two options considered were (1) to operate the seeding at the two sites independently and pool the results; or (2) to apply a crossover design in which the seeding action at each site is randomly assigned. The crossover design has been selected, and the ratio test will be used in the confirmatory analysis. The impacts of the choices of the statistical test and the design on the sample size required, as well as the power and significance of the test, will be discussed.