The experimental design, analyses, and results of the first Nelspruit randomized cloud seeding experiment are described. The experiment ran for three years, commencing in October 1984, and involved the on-top seeding of new cloud turrets growing on the flanks of isolated multicellular storms using dry ice delivered from a Learjet at around the height of the −10°C isotherm. All storms were tracked by a radar operating in computer-controlled volume scan mode. A total of 169 storms were examined, of which 94 passed the selection criteria. The most important criterion was based upon a microphysical classification scheme obtained from measurements made by the instrumented Learjet. This scheme, based upon a ratio of cloud-base temperature to potential buoyancy at 500 mb, rejected those storms in which the production of precipitation via coalescence was unlikely.
A key element of the experiment was the ability to objectively track the storms using an automatic storm tracking algorithm. Storms were analyzed in terms of their track properties, some of the more important of which were storm volume, area, and rain flux. Analyses of these track properties in 10-min time intervals either side of decision time (the time the seed/no-seed decision was made) proved to be the most revealing in terms of observed changes and rates of changes in convective cloud processes. This analysis showed an almost fourfold percentage increase in radar-measured rain flux and storm area when the seeded and control storms were compared.
A confirmatory experiment was conducted in the third season. Storm track properties that showed an apparent response to seeding in each of the first two seasons were selected prior the commencement of the third season. All but one of these track properties either stayed the same or showed increases in the third season, confirming the hypothesis that there were radar-detected differences between the seeded and control storms.