WMOD00293 - Laserfiche WebLink

Home Browse Search

6. MISCELLANEOUS ASSOCIATED WORK . . A number of topics were explored during the Plateau program that do not fit under the above section headings. Consequently, these are included here. 6.1 Effect of Type II Statistical Errors on Experimental Duration Heimbach and Super (1992, 1996) explored the important problem of encountering type II statistical errors in past randomized weather modification experiments. A type II error occurs when an experiment fails to detect an actual response to seeding, usually because the experimental unit population is too limited. Many experiments did not estimate the duration (population size) needed to achieve an acceptably low probability of encountering a type II error, say 10 percent. If an experiment failed to indicate a seeding effect upon completion (usually determined by the sponsor's patience and available -resources), the "failure" may have been caused by a type II error. The only valid conclusion from such an experiment is that it failed to demonstrate anything about seeding effectiveness. Unfortunately, the incorrect interpretation is often given; that is, that the seeding approach did not produce the desired effect. ,- When attempts were made to estimate the experimental duration needed to demonstrate real seeding effects, it was usually assumed that each treated unit would respond in the same manner. But a "'. considerable body of physical evidence shows this assumption to be improbable. The effectiveness of seeding can be expected to vary widely depending upon cloud and airflow conditions. Heimbach and Super (1992, 1996) investigated the more likely possibility that different experimental units (storms or days) have different responses to treatment. They demonstrated that this more realistic response leads to much longer experimental durations than if every treated unit responded uniformly. This important finding raises the question of whether many of the past seeding experiments that were interpreted as failures were simply too brief to demonstrate real seeding responses. Of course, if their physical ~esign was flawed, they should have failed whatever their duration. But the point is that little can be gleaned . about the seeding effectiveness of many past statistical experiments because of the uncertainty of whether they had type II errors. While Heimback and Super (1996) made recommendations for possible improvements in future statistical experiments, their main recommendation was that such experiments be postponed until a much improved physical understanding emerges. An improved physical understanding was the main goal of the Pl;iteau work. 6.2 Runoff Increases Associated with Snowfall Enhancement . . Super and McPartland (1993) reported on an investigation of likely runoff increases from an assumed seasonal snowpack increase of 10 percent. Cloud seeding programs are usually evaluated in term of seasonal percentage increases of snow water equivalent, but water users are interested in streamflow enhancement. Historical snow water equivalent and streamflow measurements were used from high elevation watersheds in the Upper Colorado River Basin. Drainages were selected for-which streamflow measurements were not significantly affected by upstream diversions and were not regulated by upstream reservoirs. A simple linear regression analysis predicted seasonal streamflow increases between 6 and 21 percent. Ten percent or more additional runoffwas estimated for most drainages for the assumed 10 percent snow water equivalent increase, an encouraging result. Reasons for differing responses were discussed which included variations in geology, vegetation, drainage slope, and aspect. 25