Determination of design wind speeds and the associated uncertainty is a crucial step in windresistant structural designs and in quantatative wind risk assessments. Design wind speeds are usually estimated by various statistical methods based on recorded field wind speeds. However, recorded field wind data are often quite limited and are not sufficient for reliable statistical analyses, especially when one estimates long-return-period (or directional) design wind speeds. Therefore, simulation of wind speeds with statistical properties compatible to those from observed wind speeds is a critical research topic. This paper respectively considers the prediction of wind speeds in well-behaved climates as well as those in typhoonprone regions. A non-stationary autoregressive model, with random coefficients modeled by a Hidden Markov Chain, is developed for simulating wind speeds in well-behaved climates. It is shown that the statistical properties of the simulated wind speeds are quite close to those of the field wind speeds. In addition, the simulated wind speeds exhibit the similar non-stationarity inherent in the field wind speeds. The second part of this paper deals with the prediction of surface wind speeds given typhoon parameters in typhoon-prone regions. A Gaussian process based regression model is used in which the prior joint distribution for surface wind speeds given typhoon parameters is assumed to be Gaussian with a covariance function containing independent uncertain hyper-parameters. The posterior distributions for hyper-parameters are first obtained by Bayesian analyses;the posterior distribution for the surface wind speed given a set of typhoon parameters is subsequently derived for predictions. The comparison between the predictions made by the proposed model and the observed wind speeds reflects that the predicted wind speeds exhibit the similar trend revealed by the observed wind speeds and the 95% confidence intervals for predictions envelop the respective observations.
展开▼
