ESTIMATION OF EXTREME WINDS

The estimation of the magnitude of extreme winds, and of the risk of their being exceeded, is an essential part of the assessment of wind loads on structures. This paper surveys the various methods available for extreme value analysis which are directly applicable to wind data. These are of two types: those making use of a sampling procedure to derive the parameters of an extreme value distribution, and those which make estimates directly from the total or parent population of wind speeds. The former is a well established technique appropriate to the majority of applications, and will be described in greater detail. There is some disagreement among wind engineers as to the correct or most useful analytic form for the extreme value distribution of winds. In this paper it is argued that for temperate latitudes, where fully developed weather systems are the dominant cause of strong winds, the extreme value distribution is a Fisher-Tippett (Type I). There is a sound theoretical basis for the use of this distribution. In regions where there is a significant risk of strong winds originating from tropical cyclones and tornadoes a Fisher-Tippett (Type II) distribution may possess certain practical advantages, but a full solution of the problem of wind prediction in these areas should be based upon a different statistical approach, tailored to model these particular wind phenomena. This requires development and an accumulation of suitable meteorological data before it can be fully implemented. Brief mention is made of methods of estimating extreme winds from the parent population of wind speeds. This has application in cases where there is a very limited length of meteorological records, and also for the analysis of extremes on a directional basis where the more common methods are inapplicable. The paper is concerned with principles and does not include a detailed mathematical treatment of the topic, which should be south in the references. © 1979.