Experimental cumulus dynamics

Cumulus dynamics is approached as an experimental science. Its development since 1947 is briefly reviewed and the current convection models described, in particular models evolved from laboratory studies, such as the ‘starting plume’. A preliminary series of experiments on real tropical cumuli has been conducted as a method for testing and improving models and for assessing the range of applicability of the laboratory results. A silver iodide seeding technique was used in a joint U. S. Weather Bureau‐U. S. Navy experimental program called Project Stormfury. Eleven tropical cumulus clouds were probed by instrumented aircraft on four days in August 1963. Six were seeded by dropping pyrotechnic silver iodide generators into their tops at intervals of 100 meters or less; five were studied as ‘controls.’ All control clouds died without further growth. Of the treated clouds, one was seeded by misunderstanding when its top was below 10,000 feet; it collapsed. The remaining five treated clouds grew. Of these, three were seeded at low enough elevations so that natural glaciation was unlikely; these are called the ‘test’ clouds. All three test clouds were observed to ‘explode’ in two phases. The first phase was a vertical growth of 10,000–20,000 ft, occupying 10–12 minutes; the second was a horizontal expansion, with the resulting giant cloud persisting more than 30 minutes. Using laboratory and theoretical results, we construct a dynamic model of the first growth phase and test it against the aircraft measurements and photographic data. The model predicts tower ascent rates, excess temperatures, and water contents, with environment and cloud base conditions as input. It predicts that unmodified clouds could not have attained the observed heights. Furthermore, the postulated effects of seeding (fusion heating, expansion, and altered precipitation fallout) are incorporated into the model and are shown able to account for the excessive growth of the test clouds following treatment. A set of extensive control calculations shows that the main deductions are insensitive to the existing uncertainties in the input data and the modeling assumptions. Improved experiments are suggested to resolve some of the basic questions in cumulus dynamics, as well as seeding effects.