|
<br />--
<br />
<br />\
<br />
<br />J'''' "'.ooOi~ c........
<br />
<br />..
<br />Braham: field Experimentation in Weather Modification
<br />
<br />can arise when meteorologists draw conclusions which are
<br />fundamentally statistical in nature, or when a statistician
<br />draws conclusions about meteorology. Such conflicts are
<br />minimized when both are members of,the same project
<br />team and are able to enter into discussions an,d debate in
<br />the presence of all statistical and meteorological evidence.
<br />But when statisticians and meteorologists are on differ-
<br />ent teams, or when articles are published in specialized
<br />journals without adequate review, conflicts are more
<br />likely, and resolution of them more difficult. The inter-
<br />pretation of Whitetop is a case in point.
<br />4. Virtually every major cloud seeding research project
<br />from the early 1950s to the present has employed ran-
<br />domization under the advice and counsel of a statistician.
<br />These projects have been useful in helping to develop a
<br />more secure scientific base. for seeding. But as our
<br />capacity to observe and model clouds and our knowledge
<br />about clouds have improved, we have witnessed within
<br />part of the weat.her modification community a growing
<br />disenchant.ment with early-stage randomization and long
<br />randomized experiments. These scientists argue that con-
<br />firmatory, hypot.hesis.testing experiments ought not to
<br />be. made until our understanding of the underlying
<br />'meteorology, and the statistical structure of meteoro-
<br />logical data, are sufficiently advanced to assure reasonable
<br />validity of the assumptions underlying a statistical design.
<br />For example, in the San Juan experiment we have a ran-
<br />domized experiment that seems to have missed its mark
<br />because the natural variations in cloud suitability and
<br />nuclei 'residence times did not mesh with the 24-hour time
<br />unit of the experiment design.
<br />5. Modern cloud seeding experiments are multidi-
<br />m.ensional and have a wealth of associated physical meas-
<br />urements. They provid~ an excellent data base for case
<br />studies and partition studies. Both of these heavily in-
<br />corporate the subjective insight and intuition of the re-
<br />searcher. But meteorologists sense uncertainties within
<br />the statistical community about the rationale for using
<br />the results of these studies to Hestablish" the results of
<br />seeding. How does one take into account the likelihood
<br />that, in addition to the studies and partitions committed
<br />to paper, there may have been many more considered and
<br />rejected for any of many reasons?
<br />Is there a common view within statistics as to the
<br />proper use of exploratory data analysis verus probabilistic
<br />inference? What are the proper roles for each?
<br />To what extent do apparently statistically significant
<br />Hresults" have to be judged against physical plausibility
<br />and rejected if found wanting iI). this respect?
<br />Under what circumstances can data be combined and
<br />treated as a single sample? For example, if the seeding
<br />procedure is changed from one year to the next, or one
<br />changes suppliers of seeding materials, or the time of
<br />treatment is changed by an hour, or the person opening
<br />the seeding Hvalve" is changed, etc., can the data be com-
<br />bined to increase sample sizes? Ultimately, can this
<br />decision be other than subjective?
<br />6. In a field like weather modification, where potential
<br />
<br />
<br />
<br />67
<br />
<br />benefits are large, it is only natural that techniques will
<br />be applied operationally before they are thoroughly
<br />undlerstood and tested. In the U.S. today, probably about
<br />half of the total expenditure for clouQ seeding is for
<br />privately sponsored, nonrandomized, operational-type
<br />projects. Typically these projects last two or three
<br />seasons. One or two of them have operated continuously
<br />for over 20 years. Some statisticians have stated that the
<br />lack of randomization makes it impossible to evaluate
<br />these projects. Is this a view accepted by most statis-
<br />ticil!tns? If a way were found to randomize some of these
<br />efforts, could we learn more from them? How can we
<br />maximize the rate of learning from commercial-type
<br />seeding projects and other nonrandomized experiments?
<br />How can we ensure the most prudent and timely transfer
<br />of new indications from the research community to the
<br />operational sector?
<br />
<br />REFERENCES
<br />
<br />Batt.an, Louis J. (1969), "Whitetop Experiment" (Technical Com-
<br />ment), Science, 165, 618.
<br />Boulrquard, A. Don (1963), "Ice Nucleus Concentrations at the
<br />pmund," Journal of the Atmospheric Sciences, 20, 386-391.
<br />Braham, Roscoe R, Jr. (1964), "What is the Role of Ice in Summer
<br />R.l!Lin-ahowers?," JqurnaZ of the Atmospheric ScUmces, 21, 640-645.
<br />-- (1965), "Project Whitetop, a Five Year Randomized Cloud
<br />Seeding Study," Report to the National Science Foundation on
<br />Glrant G22419, and paper presented at the American Meteoro-
<br />101~ical Society Conference on Cloud Physics and Severe Storms, .
<br />Reno, Nevada, October 25, 1965:
<br />-- (1966), "Final Report of Project Whitetop: Part I-Design
<br />of the Experiment; Part II-Summary of Operations," Depart-
<br />ment of Geophysical Sciences, University of Chicago (available
<br />from NTIS, PB-17tHi22).
<br />-- (1968), "Meteorological Bases for Precipitation Develop-
<br />mlent," BuUetin of the American Meteorological Society, 49, 343-353.
<br />-- (1974), "Cloud Physics of Urban Weather Modification-A
<br />Preliminary Report," BuUetin of the A merican Meteorological
<br />S~~y, 55, 100-106.
<br />--, McCarthy, John, and Flueck, John A. (1971), "Project
<br />Whitetop-Results and Interpretation," Proceedings of the Inter-
<br />national Conference on Wealher Modification, Ca.nberra, Australia,
<br />September 6-11, 1971, 127-129.
<br />Byel's, Horace R (1974), "History of Weather Modification," in
<br />Weather and Climate Modification, ed. W.N. Hess, New York:
<br />John Wiley & Sons, 3-44.
<br />Changnon, Stanley A., Jr., Huff, Floyd A., and Semonin, Richard G.
<br />(1971), "METROMEX: An Investigation of Inadvertent Weather
<br />Modification," 'Bulletin of the American Meteorological Society, 52,
<br />958-967. '
<br />Decker, Wayne L., Chang, Long-nan, and Krause, Gary F. (1971),
<br />"An Evaluation of the Whitetop Cloud Seeding Experiment
<br />through a Covariance Analysis," Jqurnal of Applied Meteorology,
<br />10, 1193-1197.
<br />--, and Schickedanz, Paul T. (1966), "Final Report of Project
<br />Whitetop: Part IV-A Summary of the Re.infall Analyses,"
<br />College of Agriculture, University of Missouri, Columbia, Missouri.
<br />Elliott, Robert D. (1974), "Experience of the Private Sector," in
<br />w.eather and Climate Modification, ed. W.N. Hess, New York:
<br />John Wiley & Sons, 45-89.
<br />--, Shaffer, Russell W., Court, Arnold, and Hannaford, Jack F.
<br />(li~76), "Comprehensive Evaluation Report, Colorado River Basin
<br />Pillot Project," Report ARI-76-1 to Bureau of Reclamation,
<br />AErometric Research Inc., Goleta, California.
<br />Flue<~k, John A. (1968), "A Statistical Analysis of Project White-
<br />top's Precipitation Data," Proceedings of the First Conference on
<br />W,ealher Modification, Albany, New York, April 28-May 1, 1968,
<br />Beeton: American Meteorological Society, 26-35.
<br />-- (1971), "Final Report of Project Whitetop: Part V-Statisti-
<br />call Analyses of the Ground Level Precipitation Data," Department
<br />
<br />_",!;.j;,<100....... -"4-::-"~'_'"
<br />
|