Evaluating the mixture Weibull distribution for modeling the pedestrian-level wind environment around an isolated building

Wind speed probability distribution functions (PDFs) are essential for assessing urban wind environments. While unimodal distribution functions have commonly been used in previous studies, urban wind exhibits bimodal PDF patterns at specific locations. This study examines the robustness of the mixture Weibull distribution (2W2W) in capturing unimodal and bimodal PDFs of wind speed around an isolated building, ensuring reliability even in the presence of rare bimodal features. The 2W2W model's performance is compared against two unimodal models of the two-parameter Weibull distribution (2W) and the three-parameter Weibull distribution (3W) using two parameter estimation methods: the method of moments (MM) and the maximum likelihood method (ML). MM utilizes higher-order moments, enabling convenient implementation with the statistics from simulations or experiments, while ML serves as a robust validation approach for MM. The findings indicate that 2W2W provides superior accuracy in modeling complex PDFs, particularly in representing bimodal patterns. However, numerical deviations were observed in 2W2W under MM, which are attributed to the parameter solution falling into local optima. To mitigate these issues and improve reliability, an adaptive strategy was introduced, enhancing the local accuracy of the model. This study successfully advanced PDF modeling, offering significant improvements in evaluating urban wind environment.