WMOD00213 - Laserfiche WebLink

Home Browse Search

Despite the difficulties invol....ed. some techniques are available for estimation of the effects of operational sccding programs. These techniques are not as rigorous or scientifically dcsirablc as is the randomization technique used in research, \\-'here roughly half the sample of storm e....cnts is randomly left unseeded. The less rigorous techniqucs do. hO\\'ever. ofTcr an indication of the long-ternl effects of sceding on opcrational programs. A commonly cmployed technique is the "target" and "control" comparison. This technique is one described by Dr. Arnett Dennis in his book entitled "Weather Modification by Cloud Seeding (1980)". This tcchnique is bascd on the selection ofa variable that would be affected by seeding (e.g.. liquid precipitation. snowpack or streamflow). Records orthe variable to bc tested arc acquired for an historical (not seeded) period of many years duration (20 years or more ifpossiblc). Thcse rccords are partitioned into those located within the designated "target" area of the project and those in a ncarby "control" arca. Ideally the control sites should be selected in an arca metcorologically similar to the target. but onc that would bc unaffected by the seeding (or seeding from other adjaccnt projects). The historical data (e.g.. precipitation) in both thc target and control areas arc taken from past years that have not bccn subjcct to cloud seeding activities in eithcr area. These data are evaluated for the same seasonal period as that of the proposed or previolls seeding. The targct and control sets of data for the llnseeded seasons are lIsed to develop an equation (typically a linear regression) that estimates the amount of target area precipitation. based on prccipitation obscrved in the control mea. This regression equation is thcn applied to the seeded period to estimate what the target area prccipitation would have been without sccding. bascd on that obser....ed in the control area(s). This allows a comparison bctween the prcdieted target area natural precipitation and that which actually oceurrcd during the scedcd pcriod to dctcrminc iftherc arc any diffcrcnces potcntially caused by cloud seeding activities. This targct and control technique viorks well where a good historical correlation can be found between target and control area precipitation. Generally. the c10scr the target and control arcas arc in tenns of elevation and topography. thc higher the correlation will be. Control sites that arc too close to the target arca. however. can be subject to contamination by the seeding activities. This can rcsult in an underestimate of the seeding effect. For precipitation and snowpack assessmcnts. a correlation coefficient (r) of 0.90 or better would be considered excellent. A correlation eoefficicnt of 0.90 would indicate that over 80 pcrccnt ofthc variance (~) in the historical data set would bc explained by the regrcssion cquation used to predict thc variablc (cxpected pn:cipitation or snowpack) in the seeded ycars. An equation indicating perfect correlation would havc an r value of 1.0. 13.2 Phnical Arlllroaches Thc rcsults from a statistical cvaluation. such as a target/control analysis. can bc strengthened through supporting physical studies. as recommended in a response to a National Research Council Report (2003) by thc Weather ~Iodification Association (WMA. 2004). One techniquc that has bccn employed by the Desert Research Institutc (DRI) in thc asscssment of the eflcctiveness of at least thc targeting (ifnot the magnitude) of seeding effects ofwintcr programs is that of analyzing samplcs of snow from the target area during seedcd periods to detennine whether silver is present in projects that use silver iodide as the seeding agent (\Varburton ct a\. 1995 and 1996). The following contains a summar)' of this technique.