Cloud Motion and Rainfall Statistics

Abstract

This paper presents some of the results obtained from a new analysis of the data collected during 1968 and 1970 by EML in the Florida cumulus modification experiment (the ?single cloud? experiment). The most important new element in this analysis is the stratification of the data into categories based on cloud motion. Category 1 days are those displaying significant, relatively uniform cloud motion throughout the day. Category 2 days are those displaying no such motion. First the rainfall data are analyzed without regard to motion categories, the log-normal model is introduced, and evidence for multiplicative seeding factor is adduced and its value estimated. The data are also analyzed on a pairwise basis. Next the data are stratified by motion categories and the analysis repeated with categories. It is shown that although a log-normal model can still be applied within categories, there is no basis for assuming a multiplicative seeding factor within categories. It is shown that category 1 clouds respond to seeding in a significantly different manner than do category 2 clouds. Evidence indicating that seeding tends to promote the merger of clouds is presented. The lifetimes of those clouds which did not merge are analyzed. Seeding apparently increased the lifetime of these clouds by about 40%. This holds both across and within motion categories. With the exception of a single outlier, the lifetime data can be taken as log-normal. It is shown that the effects of seeding on rainfall, both within and across motion categories, of non-merging clouds are mainly effects on intensity (average rainfall per unit time) rather than on lifetimes. The implications of these results to cloud modelers, cloud modification operations, and further statistical analyses are briefly discussed. An effort is made to identify a cloud-by-cloud seeding effect. A quantity called the ?rank indicated seeding effect? is introduced and compared to a quantity called ?pair wise seeding effect.? It seems that these quantities behave much more regularly for category 1 clouds than for category 2 clouds. In an appendix, some properties of the log-normal distribution are presented along with some discussion as to the relevance of these properties to these analyses and those anticipated in the future.