Abstract With the growing demand for long-span bridges in mountainous canyons, wind resistance has become a critical design consideration. Accurately characterizing the complex, terrain-specific wind characteristics—through probabilistic modeling—is essential. However, conventional parametric distribution models often fail to capture the full complexity of wind behavior in such environments. This study addresses this challenge by analyzing one year of high-resolution wind speed and direction measurements collected at a representative mountainous site. We employ an adaptive bandwidth-optimized Gaussian kernel density estimation (KDE) method to construct marginal distribution models for wind characteristics—bypassing restrictive parametric assumptions and effectively capturing multimodal and asymmetric features inherent in the observed data. Building upon these nonparametric marginals, we further apply Copula theory to model the dependence structure between wind speed and direction, enabling a flexible and decoupled representation of their joint probabilistic behavior. This approach accurately captures the nonlinear interdependence between the two variables, which is often overlooked in traditional bivariate analyses. The key findings are as follows: (1) Wind fields in complex mountainous terrain exhibit pronounced directional confinement and “wind-locking” phenomena, with prevailing wind directions strongly aligned with local topographic orientation (e.g., valley axis). (2) The Gaussian KDE method demonstrates superior capability in representing the multimodal and skewed nature of wind data in such environments, outperforming conventional parametric fits. (3) Copula-based modeling effectively characterizes the nonlinear dependence between wind speed and direction, offering a robust and versatile framework for joint distribution modeling in complex terrain.
Chen et al. (Wed,) studied this question.