WMOD00534 - Laserfiche WebLink

Home Browse Search

MAY 1980 TABLE 1. Statistics for crest without stratification. M and N are the number of seed and no-seed events, respectively. Zeros are the number of zero precipitation events and mean is the average 6 h precipitation [mm(6 h)-']. Ratio is the ratio of mean seed precipitation to mean no-seed precipitation. The test statistic is computed using the Wilcoxon nonparametric two-sample test. P value is the probability by chance alone of having a test statistic value more extreme in either direction than the actually observed test statistic value. The first number is from original work. The second, after the slash, is the new work (i.e., originaUnew) . Projects All Precipitation group Crest Stratifications None M + N = Total number of cases 1248/1003 Seeded Nonseeded M = 499/479 N = 749/524 Zeros = 51150 Zeros = 145/63 Mean = 2.84/2.87 Mean = 2.1212.57 Ratio (seeded/nonseeded) = 1.34/1.12 Test statistic = 4.956/1.835 P value < 0.00110.066 that this problem did not exist in the remainder of the data set. The problem with the Climax experiments came about because neither Climax'l nor Climax II had project rawinsondes but substituted an interpolation scheme to derive a twice daily sounding from the National Weather Service grid. This meant that there was a pseudo-sounding available every 12 h NOTES 623 whether the time period was declared an experimen- tal day or not. Varidman and Moore lifted every sounding available during the winter season with an algorithm designed to determine cloud presence and dimensions. They used every sounding that formed a respectable cloud in their analysis, a consider- able number of which were not within the strict randomization. This problem with the Climax analy- sis introduced bias in the seed/no-seed differences and, when corrected, produced a major change in the overall results as well as the stratified results. Both the corrected data set and changes in results will be discussed. 2. Discussion of the problem Let us refer to all the cases that were outside of the strict randomization as "renegade cases." A few of the renegade cases were categorized as seeded because of suspected contamination, but an over- whelming majority were categorized as non-seeded because they were, in fact, not involved in the ex- perimental operations. This inclusion of the renegade cases resulted in the non-seeded population of Climax being approximately double that of the seeded population. The rationale was that even though the renegade cases were not within the strict randomiza- tion, they were valuable informative data and should be used in the analysis. However, when one reviews how experimental TABLE 2_ Stratification over ranges of LCTTI, the lifted cloud-top temperature with respect to ice: (A) warmer than -lOoC, (B) between -10 and -30oC and (C) colder than -30oC. Projects All All Projects A. Warmer than -lOoC Precipitation group Crest Stratifications LCTTI > - 100C M + N = 93/80 Seeded Nonseeded M = 36/34 N = 57/46 Zeros = 10/11 Zeros = 19/12 Mean = 2.9412.98 Mean = 2.33/2.91 Ratio (seeded/nonseeded) = 1.26/1.02 Test statistic = 1.230/0.143 P value = 0.220/0.888 C. Colder than -30oC Precipitation group Crest Stratifications LCTTI < - 300C M + N = 294/213 Seeded Nonseeded M = 110/89 N = 184/124 Zeros = 6/6 Zeros = 18/4 Mean = 3.20/3.19 Mean = 3.09/3.69 Ratio (seeded/nonseeded) = 1.03/0.87 Test statistic = 0.413/ -1.392 P value = 0.64010.164 Projects All B. Between -10 and -30oC Precipitation group Crest Stratifications -lOoC < LCTTI < - 300C M + N = 8611710 Seeded Nonseeded M = 353/356 N = 508/354 Zeros = 35/33 Zeros = 108/47 Mean = 2.7212.78 Mean = 1.74/2.13 Ratio (seeded/nonseeded) = 1.57/1.31 Test statistic = 5.444/3.108 P value = <0.00110.002 .J