ABSTRACT
Wind loading estimation is one facet of the entire design process. Wind speed as a climatic process which is time-dependent requires proper assessment in order to predict its speed for specified return periods. This requires a good understanding of the wind process within the lower atmosphere (where the designer of civil engineering structures has interest) and then modelling it by applying stochastic methods. This is basically what this thesis addresses. The wind speed data for twenty (20) synoptic meteorological stations in Nigeria were collected and analyzed using statistical and probability approaches. The means, variances and coefficients of variation for all the stations were computed. An empirical selection of a most probable stochastic model based on the understanding of the underlying physical phenomenon and literature is done. The selected empirical and theoretical model turned out to be the extreme maximum value Type I Probability model (EV-I), also referred to as the Gumbel distribution Type I of the largest values. The model is tested using the acquired data. The parameters of the model are estimated using three methods of parameter estimation. These are then used in the statistical testing for the goodness-of-fit. The tests employed validate the selected model. The results of the predictions performed for twelve (12) return periods for all the stations are compared with those obtained by Soboyejo (1971) who worked on the distribution of extreme wind in Nigeria. The results from this present work are found to be more acceptable. It is concluded that the use of design wind speed estimates higher than the more realistic estimates implies allocating scarce resources to areas having no necessary need, and therefore does not give the appropriate balance between safety and economy. The use of the new wind speed estimates developed in this work is advocated and the need to undertake further research into the stochastic modelling of wind loads taking into consideration the subject of wind directionality is recommended.