Laserfiche WebLink
<br />prove to be unacceptable to a majority of the populace. ranging from the man on the street to <br />those in academia to those in the media. Equally imponanl. the DWB requires evidence of how <br />well the seeding progr.un \\ork\.-d in order to make rational decisions about any future seeding. <br />Ag.ain. sihcr-in-snow analysis cannot provide the total answer regarding project success. but it <br />can provide imponant evidence concerning the critical process of seeding agent T&D over the <br />target area from the particular Agl generator nern'ork that was employed. <br /> <br />1.2 Statistiul r.,,'aluation Limitations <br /> <br />This section briefly deals with the nC\..~ for at least some level of physical evaluation of <br />cloud seeding projects. Most operational projects have been evaluated only by statistical <br />approaches which will ~ seen to have serious poh:mial naws when used alone. Evcn some well <br />kno....n randomized cloud seeding experiments. ....-hich u.'ied sophisticated statistical methods. <br />have been largcl}' discredited in the scientific literature upon closer examination of their physical <br />plausibility. The most recent American Meteorological Society (1998) Policy Statement on <br />Wealher Modification recommends that. "\\'"herea'i a statistical evaluation is required to establish <br />that a significant change resulted from a given S\.,,'tiing activity. it must be accompanied by a <br />physical evaluation to confirm that the statistically observed change was due to thc seeding." <br />The trace silvcr analySl..'S herein is the most a.:onomical mcans of beginning to meet this <br />requirement set forth by the American Meteorological Society. <br /> <br />Statistical evaluations of cloud seeding programs prescnt a special challenge. especially <br />for operational projects where no experimental units arc randomly SC't aside for comparison <br />purposes, Operational programs are usually evaluated by a statistical approach known as the <br />"historical regression method" or "Largct..control" analysis, This statistical approach develops <br />relationships bet....ecn "target area" and upwind or cross....ind nonsceded. "control" observations of <br />St.-asonal snowpack accumulation for pa..t nonseeded winters, ThI.'SC rclationships arc used to <br />predict the exptXled target accumulation during sceded \\inters if sc:eding wa.. not cmploy\.-d. It <br />is assumed that departures from these predictions arc caused by the St."Cding. 111c DWB <br />contracted with a privatc cloud st-eding company, Nonh American \\'eather Consultants <br />(N"A We). to perform such analyses. Thc procedures and precipitation (bolh snow and rain) <br />observations intended for the NA we analyscs "ere published in a repon dated December 31. <br />2002 (Solak et al.. 2002). well before the April I. 2003 observations to be used in the analyses <br />were availahle, The NA WC final repon concerning application of their anal~sis approaches had <br />not been St:en when this rcpon .....as written. This is as was intended in order to avoid suspicions <br />of biased analyses and intcrpretations, <br /> <br />It has long bt.--en recognized that the historical regression method ha.. serious potential <br />shortcomings related to (1) possible sources of bias and (2) uncontrolled variance. Dennis (1980) <br />discussed these concerns in detail in his book on weather modification, A possible source of bias <br />is selection of target and control stations after completion of seeding. which raises suspicion that <br />panicular stations may have been chosen to enhance the results. The D\\'B wisely minimized <br />this type of bias by requiring station selection and analysis procedures to be completed and <br />documented well before the sea<;onal snowpack data were available tor the 2002103 wimer. as <br />was done by Solak et al. (2002). <br /> <br />5 <br />