WMOD00478 - Laserfiche WebLink

Home Browse Search

~_~- --::::.rr'~.....-i-"- Dawkins and Scott: Field Experimentation in Weather Modificcltio'n the hourly precipitation data in the strata under study. The mogrid study reproduced as Braham's Table 2 in- dicates effects of seeding before actual seeding starts in certain of its cells j so do other tables. This problem also occurs in the Grossversuch III experiment and requires careful study to avoid difficulties in the interpretation of the experimental results. Several explanations suggest themselves: one possi- bility is a partial breakdown in the objective protocol set up in the experimental design j another possibility is carry-over effects of the seeding done on the previous day; or there may be some artifact in the statistical analysis. We have been able to examine some of the suggested explanations. There were 198 projec~ days j that is, days declared experimental days by Braham. Of these 198 days, 102 were seeded days and 96 not seeded j these project days have been used in the analyses. Three objective criteria were set up by Braham (1966) at the start of the experi- ment; 22 of the project days did not satisfy all three criteria and thus strictly speaking should not have been used. Of these 22 days, 10 are in 1964, the last year of the experiment. On the other hand, 30 nonproject days did satisfy all three criteria and should have been included but were not, based on a strict application of the original protocol. However, 5 of these 30 days were Sundays in 1964, which w,ere declared to be nonoperational before the start of the 1964 season. Five other days which satisfied all criteria were declared nonoperational before the seeding decision envelope was opened because the seeder planes could not fly in bad weather (Braham 1977). Since the decision envelope for nonoperational days was returned unopened, we cannot assess the consequences of eliminat- ing these 30 (or 25 or 20) days which satisfied the criteria from the roster of experimental days. The precipitation was heavy on all five days eliminated for bad flying weather, ranging from 0.22-2.34 inches per gage in, the target (the largest by far for any days in the experimental seasons). For the 20 days where no reason has been sug- gested for their omission, 4 had precipitation greater than 7S 0.0:5 for the average target gage, of which 3 had more than 0.1!!}, but 10 had less than 0.01. Except for the five days "off" due to bad flying weather, there does not appear to be :anything unusual about the distribution of precipita- tion on these omitted days. Neither does there seem to be anything unusual about the values of the criteria on these days j they do not appear to be borderline cases. Returning to the 22 project days which should not have been included, most of these are far from borderline C&Sles, and few of them appear to change into a Braham- day. Only 6 are W-days, while 16 are E-days j 10 are frontal days, while 12 are air-mass . days. When the ana.lysis of the effect of seeding in the target area is repeated with the 22 noncriteria days omitted (as they should be), there is no striking change in the conclusions although there are, of course, differences. Even the results of l~nalyses of the precipitation in the ten-hour period before seeding starts show no striking change (see the table). We do not know why these extra 22 days were included in the project, but they do not seem to make any spe,cial difference in the conclusions. Inspection of the hourly rainfall plots for the 36-hour period starting at midnight (10 or 11 hours before the start of seeding) .and ending at noon of the next day (see Figure C) suggests that there may be a carry-over effect into the morning following seeding. Analyses were made using the seeding decision of the day before as well as that of the day itself to categorize each day during the experimental seasons (except for those beginning and ending days where we do not have enough information about the day before or the day after, as the case may be). The table shows the results of the analyses of precipita- tion in the target for the lO-hour period from midnight to llO A.M. of the today of each pairwise category. This is the period when one might expect confusion between the caflry-over effects from seeding on the previous day (ye.!terday) and the before-seeding effects of todaY'8 operations. Here S denotes a seeded project day, N denotes a not-seeded project day, and 0 a nonproject day. The first line of the table compares the pairwise category SS, Search for Evidence of Aftereffects of Cloud SeedinfJ'. Comparison of Precipitation in Target Area during the Ten-hour Period before Seeding St61rtS. Pairwise Stratification by Decision of Yesterday, Decision of Today. Results in Parentheses are Results of Computations when Noncritel'ion Days are Omitted Compare Against Block Percentage Yesterday Today Yesterday Today No. of pairs Increase ProbabJ/lty I 8 8 N N 37 (35) 22 (19) +297 (+205) 0.055 (0.12) II .8 0 N 0 34 (34) 40 (40) -52 (-52) 0.15 (0.15) 8 8 N 8 37 (35) . 32 (31) +8 (-5) 0.81 (0.99) I S N N N 30 (27) 22 (19) +153 (+134) 0.29 (0.43) l 8 N 102 (97) 96 (92) -23 (-28) 0.42 (0.31) I III 0 8 0 N 33 (28) 44 (36) -61 (-65) 0.18 (0.16) ,. " 8 8 8 N 37 (35) 30 (27) +57 (+30) 0.31 (0.46) J N 8 N N 32 (31) 22 (19) +269 (+222) 0.091 (0.17) 8 N 102 (94) 96 (82) +59 (+39) 0.15 (0.32)