Laserfiche WebLink
<br />- 64 - <br /> <br />If the raingage record is adequate to mathematically describe <br />the natural rainfall distribution, the second task is to make an initial <br />estimate of the required length of an experiment in order to detect'a <br />particular seeding effect. This is done with a type of Monte Carlo1pro- <br />cedure whereby two random samples of equal size are generated--the first <br />corresponding to non-seeded (control) rainfall and the second to seeded <br />rainfall with a specified seeding effect superimposed. Selected te~t <br />statistics are then calculated to determine whether the hypothesis of no- <br />seeding effect is rejected versus the alternative of either a positive or <br />negative effect. This procedure is done a number of times for each paired <br />sample to give the Monte Carlo power (probability of detecting a seeding <br />effect) as a function of seeding effect. The results as a function of <br />sample size provide an estimate of how long an experiment must run in <br />order to detect a particular seeding effect. An example of this type of <br />simulation using radar and raingage data as input is provided by Olsen <br />and Woodley (l975) for the Florida Area Cumulus Experiment (FACE). This <br />type of exercise is essential to modern weather modification programs be- <br />cause it provides valuable information before a decision is made to begin <br />field experimentation. If the required sample and/or the hypothetical <br />seeding effect is unreasonably large, then alternative plans might be <br />advisable. It is better to reach this decision before beginning the, field <br />studies rather than after years of fruitless experimentation. Of course, <br />satisfactory results from the Monte Carlo simulation do not guarant~e <br />positive results in the field endeavor. They are a necessary, but hardly <br />a sufficient criterion for success in a randomized seeding experiment. <br /> <br />~ <br /> <br />3.0 The Estimation of Precipitation <br /> <br />Accurate precipitation measurements in sufficient quantity must <br />be the first priority in the planning for and the execution of weather <br />modification experiments. Because the historical record from climatolog- <br />ical stations is rarely adequate for these purposes, supplemental rain <br />measurements are usually required. Such measurements can be provided by <br />raingages or radar, or a combination of the two. How such measurements <br />are obtained is treated in this section. Most of the studies to be, dis- <br />cussed were made in Florida and Illinois as described by Woodley et al <br />(1975) and Huff (1971), respectively. <br /> <br />3.1 Estimates of Rainfall Using Raingages <br /> <br />Raingages in sufficient density are the accepted standard for <br />rain measurement even though they can be in error by 5 to 15% for P9int <br />measurements (Huff, 1955; Woodley et aI, 1975). To maximize the infor- <br />mation content, recording raingages are essential. Cloud seeding is <br />usually not conducted on all types of storms, and seeding selectivity is <br />likely to increase in the future. Consequently, it is important to de- <br />termine gaging requirements as a function of season, storm type, pre- <br />cipitation type, storm duration and average or total precipitation as was <br />done by Huff (1971) for the Illinois environment. Huff found that: <br /> <br />-~- <br />