Results of a Randomized Hail Suppression Experiment in Northeast Colorado. Part II: Surface Data Base and Primary Statistical Analysis

View More View Less
  • a National Center for Atmospheric Research, Boulder, CO 80307
  • | b Department of Statistics, Colorado State University, Fort Collins 80523
© Get Permissions Rent on DeepDyve
Restricted access

Abstract

An extensive statistical analysis is made of the precipitation data collected during the randomized seeding experiment conducted by the National Hail Research Experiment during 1972-74, aimed at testing the feasibility of diminishing hail by seeding with silver iodide. The major conclusion is that no effect of seeding is detected at the 10% significance level. This is true regardless of whether hail or rainfall response variables are considered, which of two methods of obtaining daily values for the response variables over the target area is used, or what distribution, if any, is assumed for the variables. Even though the ratios of hailfall or rainfall on seed days to those on control days are generally greater than 1, the confidence intervals attached to these ratios are so large, because of the large natural variance in each response variable and the small sample sizes, that the true underlying seeding effects could in every case have ranged from substantial decreases to large increases. The large confidence intervals emphasize the necessity of large sample sizes, large experimental areas or effective covariates for obtaining definitive results in precipitation modification experiments.

Abstract

An extensive statistical analysis is made of the precipitation data collected during the randomized seeding experiment conducted by the National Hail Research Experiment during 1972-74, aimed at testing the feasibility of diminishing hail by seeding with silver iodide. The major conclusion is that no effect of seeding is detected at the 10% significance level. This is true regardless of whether hail or rainfall response variables are considered, which of two methods of obtaining daily values for the response variables over the target area is used, or what distribution, if any, is assumed for the variables. Even though the ratios of hailfall or rainfall on seed days to those on control days are generally greater than 1, the confidence intervals attached to these ratios are so large, because of the large natural variance in each response variable and the small sample sizes, that the true underlying seeding effects could in every case have ranged from substantial decreases to large increases. The large confidence intervals emphasize the necessity of large sample sizes, large experimental areas or effective covariates for obtaining definitive results in precipitation modification experiments.

Save