Sampling Errors in Estimation of Extreme Winds

Abstract

An investigation is presented into the question of whether series of extreme wind speed data in climates not subjected to hurricane winds are best fit by Type I or Type II extreme distributions. Statistical evidence is adduced supporting the assumption that the Type I distribution is an appropriate model of the extreme winds. The question is also examined of the magnitude of the sampling errors inherent in various methods of estimation of the N-year wind. It is shown that the method of moments, Lieblein's order statistics method, and the method of least-squares probability plot fitting are, for practical purposes, equally acceptable in engineering calculations. Numerical examples are presented illustrating the order of magnitude of the sampling errors in the estimation of the N-year wind for various confidence levels and for various values of the mean recurrence interval N.