WMOD00478 - Laserfiche WebLink

Home Browse Search

96 . on which clouds suitable to the seeding hypothesis may have been plentiful, while substantial decreases occurred on the more common southerly wind days when large clouds containing much natural ice prevailed. While post-factum stratifications are not statistically convinc- ing, this suggestion is sufficiently reinforced by later physical understanding and numerical simulation of cloud processes to warrant testing in a second-stage experiment, to be proposed later. In fact, in my opinion, the only "failure" associated with Whitetop was the delay of second-stage experi- mentation, only recently recognized as necessary in virtually all successful weather-modification trials. There were several reasons for this problem. As Braham points out, Whitetop was tacitly regarded as a "confirmatory" experiment with a probable overall rainfall decrease, rather than as an exploratory experiment, as it should have been to enable improved phased experimentation. The distinction between "exploratory" and "confirma- tory" experimentation had not been incorporated into weather modification in the 1960s because it was first formulated by Tukey in 1969 (Tukey 1969, 1970). The problems were compounded by conflicting critical public pronouncements by a group of prominent but meteoro- logically inexperienced statisticians. These stirred up counterproductive controversy and muddied the waters to the extent that full publication and calm, objective discussion of the procedures and results by a wide com- munity of informed scientists were greatly hampered. . The delay, however, may have been compensated for by the advances which have been made in seeding tech- niques, in cloud physics, and particularly in the numeri- cal simulation of clouds and mesoscale processes (lQ-iOO km). Braham makes the point that to establish knowledge about cloud seeding, Whitetop should be repeated. He hesitates to urge this, however, on the ethical grounds that Whitetop I probably decreased rainfall in a water-needy region. The science could benefit and the ethical problem could probably be alleviated by phased experiment de- sign, provided that Whitetop II could eliminate from treatment most of those clouds or cloud situations where the learning. from all previous experimentation suggests a negative effect. Experience shows, however, that eliminating untreatable or negatively responding cloud conditions from treatment in a soundly designed con- firmatory experiment is easier said than done. Braham . discusses briefly two other important experiments, namely, Colorado River Basin and FACE 1, afflicted with this problem. In the former, cloud conditions had to be forecast 24 hours in advance (beyond the state of the art), I with the result that the experimenters' flexibility to leaye I unsuitable clouds untreated was restricted, which ren- 'dered overall results inconclusive. Post-factum stratifica- tion on cloud-top temperature did support earlier results of Climax I and II that supercooled clouds with top temperatures warmer than about - 200C were amenable to treatment. : I '~i~;'~~3;i.~J'~'~:'~;'b"';~":~'" .gz;;;-"{~-.l, ." _~ ~,_ ;._~:i.ik'~_~~c~b~~'.--:..~__;~_~ . Journal of the America" Statistical Association, March 1979 In F'ACE 1, two levels of subjectivity have been neces- sary to cope with unsuitable cloud situations. The first is to terminate a "GO" day after 60 flare drops if clouds become either too developed or too stunted (or even absent) for the hypothesis to apply. The second is to select within the total target "floating targets" ,i.e., groups of clouds with the characteristics suitable to the seedinj!1; hypothesis. Since the experiment design is double blind, with identical flights and dummy flare drops on random control days, inferences would be ac- ceptable if it were not possible to guess the seeding decision on many occasions from watching cloud be- havior. The hypothesis, in fact, would be in serious doubt were the cloud behavior not detectably different! Al- though it is rarely possible to guess the decision after only 60 flares out of a usual 200-300 and although terminated days did not show seed/control differences, the experi- mental results of the exploratory phase (1970-1976) are still challengeable on grounds of subjectivity and po- tential bias (Woodley et al. 1978; Tukey, Brillinger, and Jones 1978). Worse yet, intense interaction between mete- orologilsts and statisticians has not yet been able to hammer out objective criteria for the confirmatory phase which started in the summer of 1978. The expense of additional seasons of experimentation is a major obstacle here siince workable, objective criteria could be formu- lated if a considerably longer experimentation period and sample size were foreseeable. Returning to a possible Whitetop II, we should keep in mind. that major advances in targeting and tracing seedinJ?; materials and improved measurements and model simulations have intervened since Whitetop I. Relatively uncomplicated research would show whether persistent south-wind days with mainly unsuitable clouds could be eliminated from the sample by an objective early decision. Two different continental seeding experiments have been proposed, one in the Midwest and one already in the early field stages in the high plains region, which should answer some of the questions raised by Whitetop. Bfaham, however, did not have space to discuss the only weather-modification experiment which has success- fully been confirmed in the opinion. of virtually all meteorologists and statisticians, namely, that in Israel (Gagin and Neumann 1974; Tukey et al. 1978). This experiment tested the "optimum nucleus concentration" strate~~y in the winter storm systems which bring nearly all of ][srael's rain. F'ortunately, rain is so well correlated between areas within these storms that the much more efficient design of randomized crossover was possible; this design would not work in the summer showery con- ditions of FACE 1 or Whitetop. After 12 years and 600 cases of an exploratory followed by a confirmatory phase, the Israel experiment has dem- onstrated, without subjectivity or the multiplicity (Tukey 1977) .introduced by subsamples, that a useful 15-25 percent precipitation increase was accomplished. This concept has now been implemented operationally in Israel. It needs to be replicated in another area with