Extreme wind speeds from long-term synthetic records

This study proposes a novel methodology to create a large sized synthetic dataset of wind velocities and adopts this to discuss the probability distributions commonly used for extreme winds. A large number of long-term time series of mean wind speed are generated by a numerical procedure that faithfully reproduces the macro-meteorological component of wind velocity, while guaranteeing sample functions with random extremes. Through application of this technique, a large sized dataset of synthetic extreme wind observations has been extracted, of a size unprecedented in literature. Commonly applied extreme value (EV) methods are then used to process the dataset produced. In the first instance, the effectiveness of these models is tested to exclude any false effects due to the limited period covered by current wind measurements. Following this, interval estimations of design wind speeds are derived by analyzing EVs from records of different lengths in order to explore the applicability of EV distributions to real situations. The comparison between analytical and numerical results provides many interesting and intriguing points of discussion, and opens the way to new research horizons in EV analysis.

► The paper analyses the extreme wind speed models starting from long-term synthetic time series.
► More than 12,000 synthetic annual extreme values are adopted as a reference term.
► The most commonly adopted extreme value (EV) methods are applied on synthetic large database.
► The EV methods are discussed with reference to the complete database and with realistic-length sub-databases.
► Improvements in the applications of some EV methods are proposed.