Possibility of a Type I Statistical Error in Analysis of a Randomized Cloud Seeding Project in South Dakota

Abstract

Previous tests of rainfall data from a randomized cloud seeding project in northwestern South Dakota led to a tentative rejection of the null hypothesis that seeding had no effect and to a suspicion that seeding had caused a suppression of rainfall, not only in the target area but in nearby upwind controls. The present study shows that the shortage of rainfall on seed days extended over a three-state area and, furthermore, was present in the early morning hours before seeding commenced. It is concluded that the rainfall data collected on the project do not necessarily reflect changes in rainfall due to seeding and that rejection of the original hypothesis (that seeding had no effect) constituted a Type I error.