Statistical analysis of wind speeds

At the planning stage in any construction project, it is important for engineers to be able to assess accurately the design-strengths necessary for structures to withstand the wind climate. Codes of practice which supply wind-loading estimates, and regulate structural design standards accordingly, exist in many countries throughout the world. The aim of this study is the development of statistical methodology which will enable more precise and detailed information about the extremal behaviour of the wind climate to be supplied at any site at which wind records have been kept. Historically, almost all analyses of extreme wind speeds have been based, without strong justification, on fitting one narrow sub-class of limiting extreme value distributions to annual maximum speeds. The first model considered in this study utilizes the entire family of such distributions, while enabling a number of large gusts from each year to be fitted. Estimation precision is thereby improved, making it possible to derive reasonable estimates of extreme speeds from short runs of data. Generally, improvement of precision is gained by models which make a greater proportion of data available for fitting. Following this theme, a model for daily maximum gusts is explored in detail. A flexible technique, based on recently developed threshold methods is then applied to hourly maximum gusts. This leads to a considerable increase in the proportion of observations that may be modelled as extreme. Weather stations often keep records of hourly maximum gusts, mean speeds, and associated directional information. Models which jointly fit to mean speeds and maximum gusts potentially lead to improvement in estimation precision of both quantities. Two such models are studied, as well as a model which takes account of directional aspects of extreme velocities. The latter provides valuable information pertaining to choice of building orientation, potentially leading to considerable financial savings.