WMOD00473 - Laserfiche WebLink

Home Browse Search

1998 JOURNAL OF CLIMATE AND APPLIED METEOROLOGY VOLUME 22 nonseeded cases. Toward this end, estimates were cal- culated using the "mean double ratio" defined by (Ts/ Tns)/Cs/Cns) where T and C refer to target and control mean precipitation amounts' and sand ns refer to seeded and nonse,eded experimental units. However, as pointed out by Mielke et al. (198Ia), double ratios can be dominated by a few large precipitation events. This depends upon the distribution of precipitation amounts yielded by a particular partition. But, in gen- eral, the lower the association between the target and control and the smaller the subpopulation being ex- amined, the more unstable the double ratios tend to become. Hence, they should be interpreted with care, especially for small populations and for Zones 2 and 3 which have a lesser association with the control gages than Zone I, as will be shown. In an attempt to reduce the impact of a few large storms on the mean double ratio, a second double ratio was calculated, as suggested by a reviewer, by substituting median target and control precipitation amounts for mean amounts. This "median double ra- tio" has a shortcoming in that it can become unstable with subpopulations containing a relatively high fre- quency of zero precipitation amounts. It becomes in- determinate if more than half the cases have zero pre- cipitation. This problem is again greatest iIi Zones 2 and 3, particularly in the latter. These zones are in the lee of the Bridger Range and sometimes have zero precipitation during days with significant snowfall . amounts in Zone I. Both the mean and median double ratios will be presented with the caveats noted above. It is not known which is most appropriate in any particular partition. This would presume knowledge about whether seeding effects (if any) are proportionally greater for large or small storms, as well as whether seeding causes pre- cipitation during naturally non-precipitating periods. The double ratios do provide an indication of the mag- nitude of any seeding effect-a matter of obvious in- terest. 7. Results of partitioning experimental units Except where noted, precipitation data were grouped into three zones for all statistical testing. This was done by the simple expedient of averaging data from all gages less than 20 km, from 20 to 30 km, and beyond 30 km from the midpoint between the two generator sites (see Fig. 1). This resulted in 12 gages on or near the intended Bangtail Ridge target area designated as Zone I. Nine gages were in Zone 2, the lee slope of the Bangtail Ridge and seven gages were in Zone 3, more than 30 km downwind of the generators. a. Partitioning by Main Ridge temperature Since the AgI generator calibration illustrated in Fig. 2 shows a marked temperature dependence in effective ice nuclei, it is reasonable to partition ex- perimental days by plume temperature. The AgI plume measurements of Super (1974), together with the gen- erator calibration of Fig. '2, indicate that the artificial ice nuclei concentrations would not exceed more than a few per liter until plume temperatures decreased be- low about -80e. . The evidence from airborne tracing of AgI plumes indicated that the seeding material was usually in the lowest 400-500 m above the Main Ridge. Thus, the temperature measured by thermograph on the Main Ridge crest is a convenient estimate of plume tem- perature. For typical in-cloud lapse rates, the plume top should have been no more than 30e below the Main Ridge temperature (hereafter called Ridge tem- perature), assuming the plume ascended less than 500 m above the Main Ridge. The average of the 24 hourly Ridge temperature measurements was used to partition each experimental unit. In order to search for possible seeding effects in a consistent manner, the following scheme was applied to the population of Ridge temperatures and to all other parameters used to partition the experimental days into subpopulations. This approach avoided some of the multiplicity involved in continued searching for those particular ranges that yield the lowest probabil- ities. The entire available population of experimental days was subdivided, as closely as possible, into halves, thirds and quarters. These fractions of the whole pop- ulation, as well as the entire population, were then subjected to the statistical tests discussed in Section 6. The resulting one-tailed probabilities for Ridge tem- peratures are shown in Table 2. It can be seen that values for both statistical tests suggest that the colder portion of the total population had significant differ- ences between seeded and nonseeded subpopulations. No significant differences are evident for the warmer half of the total population. The colder portion of the population was examined further, by testing the subpopulations indicated in Ta- ble 3 using 20e intervals. Mean and median double ratios are also shown. It is seen that eJ!,cept for the coldest and warmest subpopulations shown, proba- bilities are 0.01-0.02 by both tests for Zone I, the intended target. The highest mean and median double ratios are 1.90 and 2.35, respectively, both for tem- peratures of -13 oe and colder, which suggests an ap- proximate doubling of the precipitation in that par- tition. Low probability values are also seen in the other two zones for temperatures lower than about -II oe. The highest mean double ratios in Zones 2 and 3 exceed 2.0 and are for those cases ~ 130e. The median double ratios are much larger than the mean double ratios in Zone .3, especially for the colder cases. This was due to very low median values for nonseeded days in Zone 3. For example, for the 27 nonseeded days with Ridge temperatures ~ -130, the median precipitation