Laserfiche WebLink
<br />- <br /> <br />reduce the detectability of any seeding signal, presuming any seeding effect occurred in the <br />same direction (e.g., increased snowfall) in both target and controL <br /> <br />Choosing control measurement sites involves more work than simply picking locations as <br />near the target as possible while remaining reasonably certain that contamination of the <br />control sites is minimal or nonexistent. As will be discussed further, the most important <br />factor involves maximizing the degree of association between target and control <br />measurements so the latter can be used to predict thl:! former with good accuracy. Although <br />the degree of association tends to decrease with increasing distance, many other factors <br />influence target-control relationships in complex ways, particularly in mountainous terrain. <br />Some of the important factors include shape and orientation of the terrain for prevailing <br />storm tracks, elevation, blocking by nearby upwind mountain ranges, and local topography <br />and forest cover around the measurement sites. <br /> <br />3.2 Target-Control Relationships <br /> <br />Observations from nonseeded winters prior to the 19~92-1993 (hereafter 1993) seeded winter <br />were used to establish mathematical relationships between the target and control snowfall <br />observations. This procedure assumes that the relationship did not change during the seeded <br />period, which is probably a valid assumption but cannot be proven. Changes in such <br />relationships can occur over several years because of such factors as forest growth, which may <br />affect snowfall on the measurement sites. <br /> <br />The relationship between any given winter's target and control snowfall may depart from the <br />long-term relationship for completely natural reasons. For example, if the frequency of a <br />particular storm track is much higher or lower than usual during a particular winter, the <br />target-control relationship may differ markedly from more typical winters. One to a few <br />particularly wet storms, concentrated over either the target or control area but not both, can <br />produce a marked natural change in the relationship for a given winter. <br /> <br />Target-control relationships were established for a population of many winters. These <br />relationships were used with the control snowfall measurements from the 1993 seeded winter <br />to predict the target area's natural snowfall. Departures in seeded area snowfall from these <br />predictions were exam.ined. A large departure in target area snowfall from the predicted <br />amount might be related to seeding. However, as already pointed out, such a departure <br />might also be partially or totally a natural phenomena. <br /> <br />Examination of the variability ("scatter") of individual departures from many nonseeded <br />winters can provide some idea of the probability that a large seeded winter departure was <br />caused by seeding. As previously noted, claims of a specific level of statistical significance <br />for any departure are not valid without randomization. Unfortunately, such claims are often <br />made in analysis of operational projects. <br /> <br />3 <br />