|
<br />\~\J %
<br />_ .v:#
<br />-~~"-T~ . .
<br />..,."f
<br />..~"i
<br />~.i~' ,
<br />
<br />-'.'-""
<br />
<br />i
<br />\1_. ..
<br />
<br />4- I
<br />~
<br />
<br />. ,i'.
<br />
<br />Field Ex:perimentatior11 in Weather Modification
<br />ROSCOE R. BRAHAM, JR..
<br />
<br />Weather modification provides a fertile field for interaction andcol-
<br />laboration between meteorologists and statisticians. Cloud seeding
<br />experiments, including Project Whitetop, provide a background for
<br />illustrating some of the statistica.l issues in meteorologica.l field
<br />experimentation and for exposing Bome of the difficulties that can
<br />arise when meteorologists and statisticians look at the same experi-
<br />ments from different points of view. Severa.l specific points relating
<br />to statistics and statisticians, raised by meteorologists involved in
<br />weather modification, are discussed.
<br />
<br />KEY WORDS: Experiment design; Cloud seeding; Weather
<br />modification; Project Whitetop.
<br />
<br />1. INTRODUCTION
<br />
<br />Although the association of, and cross-fertilization be-
<br />tween, statistics and meteorology goes back to the early
<br />part of this century, the past three decades of weather
<br />modification have served to bring these two sciences into
<br />intimate contact. This contact has been fruitful for
<br />meteorology and, I. believe, also for statistics; it also
<br />has produced its share of problems, some of which are
<br />. discussed in this article. But before getting down to
<br />specifics, several generalizations are in order.
<br />Statistics has many important roles in meteorology.
<br />At the University of Chicago I have been fortunate in
<br />having &- close association with. Professor William
<br />Kruskal, and in earlier years with Professor K. Alexander
<br />Brownlee, and wi~h their associates and graduate'stu-
<br />dents. I believe that statisticians should be involved in
<br />weather modification experiments from the predesign ex-
<br />ploratory studies through the design, operations, eva.lua-
<br />tion, and reporting P9ases. Unfortunately, this ideal is
<br />not often achieved. With a few notable exceptions, sta-
<br />tisticians have been involved only in ~he design and
<br />evaluation phases. Ai? a result, they miss out on the excite-
<br />ment and the sobering realities of the great "in-between,"
<br />namely, the operational phases of this research. With this
<br />constraint, it seems hard for them to get a feel for mete-
<br />orological systems. The unfortunate result has been a
<br />tendency to promote an attitude among meteorologists
<br />that statisticians prefer to sit as judge and jury, but
<br />never to be called to the witness box.
<br />Many'meteorologists have backgrounds in physics,
<br />chemistry, or engineering. As physical scientists, we
<br />
<br />· Roscoe R. Braham, Jr., is Professor of Meteorology, Department
<br />of Goophysica.l Sciences, The University of Chicago, Chicago, IL
<br />60637. Research was supported in part by Nationa.l Science Founda-
<br />tion Grants G8214 and G22419. This article is the edited text of a
<br />talk given at the Invited Societics Session, American Statistical
<br />Association Annual Meeting in Chicago, on August 15, 1977. I wish
<br />to thank the many individuals associated with Project Whitetop for
<br />their long and devoted assistance. I also thank the National Science
<br />F~undation for its sponsorship over many years under a number of
<br />different grants.
<br />
<br />".l<~"~..,l~"N.k""'.._M " ,,-,,.~~~,,~ _,
<br />
<br />-.,'~':..--.,,,,;.ri~->..,,,::;.,,
<br />
<br />;-'i:'~~~:"
<br />
<br />natw~ally think in terms of physical models. We are in-
<br />clined to accept evidence which seems compatible with
<br />established meteorological physics, even though it may
<br />have modest statistical support, over evidence that runs
<br />counter to accepted physics, even though the latter may
<br />have stronger "statistical" support.
<br />Within meteorology and statistics alike, weather
<br />modification and weather modifiers are often viewed with
<br />suspicion and disdain. Unfortunately, this has tended to
<br />depriv.e the field of some badly needed intellectual
<br />resources.
<br />Wl~ather modification is an exciting field of enormous
<br />scientific challenge and offers the prospect of immense
<br />societal benefits. But the path to progress often has been
<br />obscured by differing perceptions as to the state of the
<br />scien,ce, differing types and levels of evidence upon which
<br />these perceptions are based, and differing views as to
<br />when and how an emerging technology should be used.
<br />Even among serious researchers, both statisticians and
<br />meteorologists, there are honest differences in opinion.
<br />Sometimes these differences are overplayed in the popular
<br />press. Sometimes one or the other opinion is used un-
<br />critically to support a particular point of view. At other
<br />times, subtleties of our differences are not appreciated,
<br />'Or are intentionally ignored, even by supposedly serious
<br />writers.
<br />There are many reasons for this confusion. The sub-
<br />jects of our experiments,.viz., clouds and cloud systems,
<br />are complex, highly variable and interactive, and poorly
<br />understood. Theories of cloud response to seeding are, in
<br />general, simplistic and inadequate. In deliberate attempts
<br />to change some weather parameter, experimenters have
<br />control .over treatment methods but must rely on
<br />weather forecasts or other prescreening devices to provide
<br />a measure of homogeneity to the experimental subjects.
<br />Measurement deficiencies limit the amount we know
<br />about the subjects, even after the fact. It is never possible
<br />to replicate exactly. We have neither the laboratory
<br />control and repeatability of physics, nor the identical
<br />twine: of biology. Experiments can only be conducted in
<br />the public domain, with consequent sociological, jurisdic-
<br />. tional, and ethical issues. In addition, experiments at the
<br />interface of statistics and meteorology are liable to differ-
<br />ent interpretations when seen from our different points
<br />'Of view.
<br />With these difficulties, why does anyone want to seed
<br />
<br />57
<br />
<br />R.~,r1nt.d from: iC> Journal of the American Statistical Association
<br />
<br />March 1979, Volumca 74, Number 365
<br />
<br />Invited Paper
<br />P.gq 57-104
<br />
<br />'":i~",j',;,>.,,,
<br />
<br />",'.;..:..:..._,_-.:'.;.:~-,....2~~;:";;;~~:i.i~,.~~..... -~~
<br />
<br />'~.-,;~,........o'.i
<br />
<br />!fi~"-
<br />
<br />""',~-w_
<br />
<br />...
<br />
|