Nonparametric sector dependence modelling for the directional synthesis of local wind climate and building aerodynamic responses: Adaptive kernel-based approach

• A nonparametric sector dependence modelling technique using Adaptive Kernel Density Estimators (Adaptive-KDE) is proposed to address challenges in sector-based directionality methods. • The applicability, accuracy, and robustness of the Adaptive-KDE framework have been demonstrated through application examples, in which local wind climate data from the three cities were synthesized with the wind tunnel test data of an 18-story prototype mass timber building. • The proposed Adaptive-KDE directionality analysis framework gives predictions that closely match the results of the direct calculation method for service-level acceleration responses. • For high MRI winds, where dependence between sectors weakens, the predictions of the Adaptive-KDE approach align more closely with the multi-sector method. • The new framework enables the use of narrow sectors, extending the application of sector-based directionality analysis to state-of-the-art performance-based wind design approaches. Accounting for the directionality of wind is crucial in estimating the response of buildings to wind load. Sector-based directionality techniques are widely used for analyzing directionality effects. In single- and multi-sector methods, directional sectors of the local wind climate and building aerodynamic responses are analyzed separately, while their statistical correlation is assumed to be fully dependent or independent, respectively. The multi-sector method, which is preferred for structural design due to its relative conservatism, requires the use of wide sectors to ensure the statistical independence assumption holds. This, in turn, requires interpolating aerodynamic response parameters, which is prone to errors due to rapid variations with small directional changes. Moreover, performance-based wind design (PBWD) approaches, as outlined in the American Society of Civil Engineers Prestandard for PBWD, require 10-degree or narrower sectors in aerodynamic response representation for detailed directional resolution. Narrow wind sectors often exhibit correlation, necessitating accurate dependence modelling. Parametric copula-based methods have been used to model sector correlations; however, they impose restrictive assumptions on dependence patterns. Therefore, this paper proposes a sector-based directionality technique with nonparametric dependence modelling using adaptive kernel density estimators. To demonstrate the applicability and accuracy of the method, wind responses of a prototype mass-timber building hypothetically located in three cities: i.e., Toronto (Canada), Melbourne (Australia), and Baltimore (USA), were predicted. The predictions were compared with responses empirically computed from historical records. The results demonstrated that the method extends the applicability of sector-based directionality analysis to narrow sectors, making it suitable for PBWD approaches.