WMOD00478 - Laserfiche WebLink

Home Browse Search

.' 100 ~ ~~~~ , before weather modification experiments ean progress much beyond the exploratory stage. At the same time, it 'is no surprise that these discussants found some meteo- rologists who felt that they already had such an ability. This is implicit in all commercial cloud seeding activities. It would be interesting to go back to their respondents and ask about the evidence ~pon which such feelings were based, so that the authors could form their own. opinions on this matter. Clearly, if a group of meteo- rologists really do have the ability to predict, a priori, additive units in cloud seeding for rain enhancement or hail suppression, then such ability should be incorporated immediately into the design of confirmatory experiments. I would like to amplify a point raised by these discus- sants. In cloud seeding for rain enhancement, commercial operators normally intentionally choose naturally rainy days. Many research projects try to eliminate both exces- sively dry and wet days. The reason is that present seed- ing techniques for rain "enhancement" simply cannot be effective when natural conditions are not favorable for rain, and we believe that seeding effects are relatively small and, harder to detect during very heavy natural rains. >ill' [ij I I: Dawkim-Scott: AB a meteorologist I am unable to comment adequately about the use of the C(O!) teat and the statement that it is more powerful and more valid than the several tests employed in the analysis of White- top data. I note, however, that at least one other discus- sant is of the view that simple robust techniques seem to be indicated at this stage in our understanding. Additional discussion on the pros and cons of this test will be found in the 1973 National Academy of Sciences report (pp. 216-222). In discussing the 24-hour rainfall analyses, these discus- sants found that their analysis was insensitive to the gage density, and they suggest that this was contrary to a statement in my article. It should be remembered that rainfall measurement errors associated with gage spacing depend upon the type of storms sampled, plus the time duration and area over which. the integration is per- formed. In our Whitetop analysis, we used one-hour time intervals for determining plume and nonplume, and our maximum area of interest was the 6O-mile-radius re- search circle. In this case, the requirements on gage spacing are quite different from those for 24-hour integra- tions over much larger areas. Several studies have been carried out to determine how rainfall estimates are affected by'size of area sampled, gage density, and storm type. Interested readers should refer to Huff (1971). In response to the discussion about experimental days that did or did not meet the selection criteria, recall that the criteria were used to help us select days thought to be generally favorable for the experiment and should not be construed as a hypothesis for any particular seeding effect. As shown by Braham (1966), these criteria per- formed quite well, since precipitation occurred somewhere in the research, circle on over 90 percent of the experi- Journal of the American Statistical Associcltion, March 1979 mental days, but on less than half of the nonexperimental days. To stratify Whitetop data according to 'whether the winds at a number of levels, and oyer a considerable portion of the day, remained within criteria limits' is an interesting exercise; but as a meteorologist, I ask why it has any particular merit over countles:s other stratifica- tions in looking for, lI.nd in judging, seeding effects. I also would comment tha,t labeling some of these stratifica- tions "Braham days'" or "non-Braham days" was a little misleading, since they have little to do with the criteria, and the logic behind the criteria,' which I introduced in the prescreening step. The stratification into air-mass and frontal weather has meteorological foundation, though its value in in- terpreting Whitetop is diminished because it was applied to 24-hour rain total over a circle 180 miles in radius. Over such a large space-time dimension, for the season and region in question, one may often find only air-mass convection. Seldom would frontal convection dominate. In additioD, one sho'~ld note that the upwind-downwind specification in Figures C and D is based on the direction of our middle seeding site from our radar. This C01?"e- sponds to our early-morning forecast for the midday lower- level winds inside ou.r 6O-mile-radius research circle. The analyses might be more interpretable if Dawkins and Scott had made use 'of actually measured winds through- out the 24-hour period, and at the levels of importance in the meteorological model being examined. TheBe discussants, raise a question about 30 days, during the five summers, which were not included in the experiment even though they apparently met the selec- tion criteria. (V alue8 for all three criteria for every day of our experiment are found in Braham. 1966.) They also question another 22 days which were included in spite of the faet that they fLppeared to fall outside the .criteria limits. Both of these conditions were discussed in our reporu;, but exact dll.tea and reasons were not listed. To complete the record, such a list is given here. Conl~erning the 30 "missing days": Two days, 17 July 1961 llmd 3 July 1964, actually were outside criteria limits, thus not missing at all, but were reported as being on the limits of acceptability because 'of rounding. Five days in 1964 were Sundays, declared to be "days-off" prior to the start of the season (14 and 21 June, 5 July, 2 and ao August). On the afternoon of 29 July 1961, after we hadl worked a string of 21 consecutive days, I declared that the next day (30 July 1961) would be nonexperi-' mental (I deemed this necessary for crew morale). Five days in 1960 (24, 25, and 27 July, 7 and 8 August) and two in 1961 (12 and 13 June) were declared nonexperi- mental because we knew before the day started that our radar l~cope cameras (main and spare) were inoperative (in fac:t, we were involved in flying them to a Chicago manuf:!l.cturer for emergency repair service). 1 declared. 17 June 1962 to be Donoperational because the radar was known to be inoperative before the envelope was opened (cases where the radar went down during a day were left in the experiment).