Wind loads are characterized by their spatial and temporal variability and are classified as variable actions according to EN 1990. The characteristic value Qk for variable actions is defined in Section 4.1.2, Paragraph (7) of the standard as the upper value of a specified exceedance probability for a given period. Specifically, reference is made to the 98% exceedance percentile of the extreme value distribution for a one-year reference period.
Determining extreme values thus requires a statistical analysis of pressure time series, wind loads, or their effects. It is essential to note that wind loads do not follow a Gaussian distribution, and the effects of wind loads do not correlate linearly with wind speeds. In standard engineering practice, design is based on normative specifications that define a gust speed with an associated pressure or force coefficient. These coefficients now implicitly include the extreme value statistics of wind loads, as opposed to earlier approaches based on averages. The exclusive use of extreme values in load determination leads to challenges regarding the simultaneity of applied wind loads (correlation effects). Alternative approaches, particularly in bridge construction, describe wind using mean values and variable components with different safety factors.
A direct transfer of extreme value statistics to different results such as wind pressure, wind suction, or sectional forces is generally not feasible in wind loads. Ideally, this method would need to be applied to each individual load combination. However, normative requirements are limited to the relative proof that the design effect value of all combined actions Ed is less than the design resistance value Rd of the components or structure:
Additionally, it is worth noting that modern CFD simulations and experimental wind tunnel tests are gaining importance for capturing complex wind load scenarios more accurately and analyzing the interaction between wind and structure in greater detail. A precise categorization of requirements for numerical investigations is therefore essential to meet various accuracy demands and to efficiently apply derived, specific procedures to different study objects and goals.