WMOD00265 - Laserfiche WebLink

Home Browse Search

~ and comparison of loss-cost data by township with elevation. It was discovered that melting is sufficient to explain much of the decrease in average loss-cost ratios as one goes from western to eastern Kansas. A paper on the results was presented at the Tenth Conference on Weather Modification at Arlington, Virginia in May of 1986 (Rasmussen [10]). Unfortunately, the ability to adjust hail statistics for elevation did not lead to any breakthrough in the evaluation studies to be described below. Statistics derived with the adjusted data looked about the same as those derived with unadjusted data. We f1nally decided that the analyses using unadjusted data already had elevation factored into them, because they consisted essentially of comparisons of events during seed and no-seed years within each grid cell, and thus of hailstonns occurring at the same elevation. All of the ,results presented below are based on the unadjusted data. 4.3 Comparison of Losses in the Target Area during Seed and No-Seed Years The fIrSt step in the analysis was to compare the loss-cost ratios for each grid cell for the seed and no-seed years. Figure 4.2 shows the ratio obtained for each cell by dividing the average loss-cost ratio for the seed years by the average loss-cost ratio for the years without seeding. The calculations were extended outside of the target area by assuming that in each case the years 1948 to 1974 inclusive were no-seed years and the years 1975-85 inclusive were seed years. The results show a region over the eastern part of the target area and extending to the northeast where hail damage was as much as SO percent less during the seed years than during the 27 preceding years. However, this presentation does not provide any indication of the probability that such changes represent anything more than random variations. In order to obtain some idea of the probability of a real seeding effect, the loss-cost data for the years with and without seeding were compared using the Wilcoxon two-sample test. The results are shown in f1gure 4.3. This color plot shows the probability that the loss-cost ratios during the seed years (1975 to 1985 inclusive) exceeded those for the no-seed years (1948 to 1974 inclusive). It should be noted that the analysis again was extended across the entire grid and that a smoothing routine was used to suppress local variations and thereby bring out the overall pattern. Subtracting the plotted probability values from 1.00 yields the probability that the loss-cost ratios for the seed years were less than those for the no-seed years. Even in the absence of systematic differences, random year-to-year fluctuations cause the calculated probabilities to vary. As statisticians sometimes accept probability values of 0.10 or less as evidence of a significant difference between two sets of numbers, we can say that the areas shaded dark red (P>O.90) had significantly more hail during the seed years than during the no-seed years and that the areas shaded dark blue (P<O.lO) had significantly less hail damage during the seed years. .; Some grid cells on the southwest and southeast corners of the target area had significantly more hail during the seed years.The comparison between seed and no-seed years generally becomes more favorable as one proceeds eastward from the western border of the target area. Six grid cells in the eastern part of the target area had significantly less hail during the seed years. 1bey constitute part of the favorable area already identified in the map of seed/no-seed ratios (figure 4.2) as covering the northeast part of the target area and extending to the northeast of it. The general trend agrees well with the results of Hsu and Chen [1] for the years 1975 to 1979 inclusive. ~ Because the target area varied from year to year, we reran the analysis for each grid cell that was 28