It is easy to produce ''encouraging findings'' by cutting corners in scientific rigor. Methodological and statistical issues in research have parallels in many different fields, and the solutions proposed in one are often instructive for workers in others. Meteorologists may find it interesting to compare their concerns with those of medical scientists and vice versa. ...
The requirement of informed consent is accepted as the patients' right but there are no parallel rights for the populations potentially affected by weather modification, though there have been a few cases of litigation. Application of the results of experiments, on the other hand, has close parallels between medical treatment and weather modification. It is just as much an ethical imperative to apply knowledge gained from a weather experiment in a policy to alleviate a drought, as it is to apply knowledge from clinical trials to a practice that reduces patients' suffering. ...
What effect does introduction of a glaciogenic or hygroscopic agent have on any feature of the clouds? Many clinical trials, on the other hand, are focused on finding out whether a drug, or intervention, will have a specific effect, usually on survival. The findings of clinical trials often translate directly into decisions on treatment of future patients. The findings of cloud seeding experiments, on the other hand, do not translate simply into weather modification policies. It is a long way from an experiment's showing that cloud seeding augments radar reflectivity in a cloud to gauging the economic benefit of a seeding operation, or even to knowing whether it would produce more precipitation to the ground. ...
In weather modification, adaptive experiments with instruments and burners and observation of individual cloud's apparent reactions precede experiments that are designed to confirm the efficacy of a seeding technique in a given time and place. Adaptive experimentation is a learning experience that usually cannot be evaluated objectively. Designed experiments, by contrast, are tightly controlled ''black boxes'' with a protocol that is planned ahead of time and continued unchanged to completion, when the results are analyzed by already planned statistical methods, including significance tests. ...
Limiting analyses to units actually treated, however intuitively appealing, is a source of positive bias and must be avoided. ...
The assessment of the effect of cloud seeding during nonrandomized operations is based on comparisons with historical data and/or data from other areas. How can one ever know whether the difference between the operational precipitation and that in other places and other times is due to seeding? An obvious source of bias is that seeding operations usually occur after a drought and address the most affected area. It is only to be expected that the seeding period has relatively high precipitation in comparison with the preceding period. Cases of similar ''regression to the mean'' are also well known in medical studies ... Data from cloud seeding operations must be used very circumspectly, and can never have the same evidential weight as do data from randomized experiments. This fact has been well known and documented since the early days of weather modification in the 1960s ... The dilemma for operators is that their funding often depends on seeding on every available occasion, so they can never leave an occasion unseeded; as a result, they cannot produce reliable evidence on whether their seeding is effective. ...
Are the results of studies of glaciogenic seeding by AgI relevant to the present possibilities of hygroscopic seeding? The intended physical effect is very different, so why should the observed effects on the ground be the same? ...
Serious doubts persist regarding the effectiveness of cloud seeding for augmenting precipitation, especially with glaciogenic material, and yet some seeding operations may be economically justified. Why not risk a relatively small outlay on seeding clouds if it is considered at all plausible that it will result in increased precipitation or reduced hail, both of which are of con-siderable economic value. After all, other business decisions are also made without ''proof'' that they will turn out to be profitable; economic choices generally are decisions under uncertainty.
The accepted relation between proof and application is very different in medicine. Administration of a treatment is contingent upon rigorous demonstration of its effectiveness. Drugs cannot be marketed unless they have been found to be significantly effective in clinical trials. Patients are permitted to choose a treatment not if they are convinced of its efficacy but only if that efficacy has been confirmed by rigorous experimentation. The standards for scientific demonstration are very distinct from those for taking action. In meteorology it is accepted that the latter are determined by market forces: Seeding operations can be conducted absent scientific proof. In medicine, by contrast, it is accepted that drugs and treatments be made available only if there has been scientific confirmation of their effectiveness. The difference between the public attitudes toward science and policy in these two areas shows that this is not a matter of scientific method or principle but a reflection of what aspect of life one wants protected by governmental and judicial regulation. Evidently, people feel less need to be protected from possibly futile interference with the weather than from possibly unavailing medical treatment.