WMOD00464 - Laserfiche WebLink

Home Browse Search

B-3 8. Dangers of SubjE~ctive Inftuences Weather modilication experiments involve many individuals, exceedingly variable condi- tions, great difficulty in data collection and not a lot of agreement on proven facts. Th~: work- ers involved includl~ some whose self-interests are served by demonstrating seeding effects, some who believe it does not work, some who are seeking to learn and others who do not care. Consciously and unc:onsciously, each of these persons can have an undesirable and unacceptable impact on a study. As time has passed, means of rc~ducing subjective elements have been learned and brought into use; however, .there remains a properly skeptical audience who should be much more firmly assured. In this section we comment on some general aspects of bias and subjectivity and in particular on such aspects of certain major weather modification experiments. Difficulties may also arise, if research personnel are aware of whether or not, on Ii given occasion, the experimental intervention has been employed and if then, at a moment in time following their awareness, these personnel decide features of the design or execution of the study. Despite efforts to keep such decisions unaffected by knowledge of the treatment condi- tion, there remains ~ possibility of insidious influence of that knowledge on the decisions that are made. To mention a subtle possibility, were it known that seeding was not to be done on a given day, fewer personnel might be on duty to notice that recording gauges were malfunction- ing for a period that day. As another possibility, an analyst who believes in seeding and who screens the data prior to its reduction might be more likely to check into a "low" rainfall value at a gauge if he knew that seeding had been done that day. Any systematic differcmce in behavior with respect to seeded and non-seeded days might lead to an appearance of significant difference between them. In general, subjective influences seem most likely to be of importance as they affect judg- ment making, though their influence upon data collection and data processing also deserve seri- ous concern. (Of the 9 recent studies which we have reviewed, only a few -- Santa Barbara II, perhaps FACE during some seasons, and, conceivably, Israeli II -- have involved data collection and assembly in ignorance of which days were seeded and which were not.) * FACE * FACE was intc~nded to be a "double blind" experiment, with neither scientists nor clouds "aware or' whether or not clouds were seeded. However, during the first two years of the experiment, the scientists were told by the end of each experimental day whether the day had been (randomly) selected to be seeded. (The decision was made when an airborne "ra.ndomi- zation officer" opened a sealed envelope,) During 1970-1973, since no flares were fired on unseeded days, the scientists might have been able to detect whether or not flares were fired (although the firing was monitored by the randomization officer in the rear of a large aircraft, while the scientist occupied the cockpit). During 1976 AgI flares were fired on seeded days, placebo flares were fired on unseeded days. During these later years, randomized blocks were employed, with the complete schedule having been determined in advance of the season; efforts were made to maintain the schedule's security from the scientists until after the end of the season. Reported data show that, over the course of the study and over the different air- borne scientists employed, scientists correctly guessed whether the day was seeded or unseeded on l\bout 80 percent of the days they were willing to guess -- they were willing to guess on about half the total number of days. (These guesses seem to have been based on changes in cloud appearance, and NOT on knowledge of the randomization,)