Abstract
This paper summarizes results of a statistical analysis of the Israeli cloud-seeding experiment (1961–67) based upon precipitation data for the period of the experiment and upon 30 concomitant meteorological variables, both surface and upper air. The experiment had a crossover design in which days are allocated at random to seeding in one of two (North or Center) targets, separated by a Buffer region. Two statistical tests are used, one based upon normal theory and the other a Wilcoxon two-sample (nonparametric) test. On the basis of these tests it is reasonable to accept the hypothesis that cloud seeding affected precipitation; results are significant for two-sided tests at the 2.5% level in the case of the normal test and at the 10% level in the case of the Wilcoxon test. However, when these same test procedures were applied to the Buffer region instead of the Center target, even more significant results were obtained. Analysis of plume directions at the times of seeding do not indicate that this contamination was due to physical transport of seeding material. This result with regard to the Buffer strongly suggests the possibility of contamination of one target by seeding of the other. The meteorological assumptions underlying a crossover experiment are that contamination effects are negligible in comparison to the effects of seeding in the target area. When such assumptions are not valid, interpretation of the results of a crossover experiment may be grossly misleading; in fact, they may be in the wrong direction.
On the assumption that this is not the case for the Israeli experiment, i.e., that crossover effects are negligible, an attempt is made to determine the differential effects of seeding under different meteorological conditions. It is estimated that in an operational year there would be about 35 days of actual seeding with increases of 1.9 mm day−1 in the North and 2.2 mm in the Center. No evidence of large negative effects in target areas was found, but in about 10–20% of the days there may have been moderate or slight negative effects. Meteorological conditions on such days are associated with warm surface air moving from the south. Positive effects are associated with meteorological conditions that are conducive to greater than average rainfall. Tables with details about partitionings of the data by means of meteorological variables are provided.
