WMOD00473 - Laserfiche WebLink

Home Browse Search

1990 JOURNAL OF CLIMATE AND APPLIED METEOROLOGY VOLUME 22 1000 ft Contey,~ Ihown . A Go;. Sit. <<> bow Cour.. Sit. o Anal above 7000 tt Seal. (kllo",ot.n) o 10 FIG. 1. Map of Bridger Range experimental area showing locations of precipitation gages and snow courses. budget for fiscal year 1973. This resulted in fewer ex- perimental days than anticipated and decreased re- sources for analysis. Nevertheless~ considerable analyses were completed and reported in a two-part final reportl (Super et al., 1972, 1974, hereafter referred to as Part I and Part 11). These reports also contain detailed dis- cussions of experimental design, field facilities, oper- ations, all supporting investigations and a complete listing of the precipitation, Main Ridge thermograph and upper-air data used in the analysis reported herein. These reports were provided. to many interested in- 'dividuals and organizations at the time of their pub- lication. Previous statistical analyses of the BRE, noted below, gave strong indications oftype I statistical errors-i.e., rejection of a true null hypothesis. This means that some meteorological partitions yielded distributions of seeded and nonseeded days for which the statistical testing suggested significant changes in precipitation due to seeding when, in fact, no logical cause and effect relationship was evident. Changes were generally in the sense that seeding apparently decreased precipi- tation although some were also in the opposite sense. 1 Copies of the BRE Final Reports and magnetic tape copies of data used herein are available at cost from the Division of Atmospheric ResourcesResearch, Bureau of Reclamation, DFC, P.O. Box 25007, Denver, CO .80225. The suggested type I errors occurred even with rather large populations. The first indication of a type I error was reported in early 1971 by .super and Mitchell (1971) i.n a pre- liminary analysis of data from the 1969-70 winter and first half of the 1970-71 winter. After completion of the BRE, type I errors in both positive and negative sense were reported by Super (1975). These showed apparent increases in snowfall at several target gages associated with a limited number (n = 33) of exper- imental days with thin clouds. Similar apparent in- creases were found at each seeding site and at the control gage well to the south. More strikingly, 162 days with clouds thicker than 1000 m and with westerly winds indicated significantly less snowfall on the seeded days, not only at several target gages, but also at the control gage and at the Billings Airport 185 km to the east. Both these changes are believed to be type I errors. Super and Heimbach (1974) summarized the BRE statistical analysis which was fully presented in Part II. They gave further evidence of a serious type I error using National Weather Service (NWS) precipitation probability forecasts for each experimental day. How- ever, they also presented some partitions which sug- gested that real precipitation increases were associated with seeding. In these cases, snowfall increases were not apparent at the control gage, at gages near the seeding sites, or often even at gages at the crosswind edges of the intended target.