Approximating wind speed probability distributions around a building by mixture weibull distribution with the methods of moments and L-moments

Abstract

Wind speed probability distribution functions (PDFs) are crucial for evaluating urban wind environments. While previous studies have used unimodal distribution functions to model PDFs, bimodal patterns are also observed in urban areas. To more accurately model unimodal and bimodal PDFs, this study assessed the applicability of the mixture Weibull distribution (2W2W). The performance of the two-parameter Weibull distribution (2W) was also analyzed for comparison. Three parameter estimation methods (method of moments (MM), method of L-moments (LM), and maximum likelihood method (ML)) were applied to wind speed data of an isolated building case from a LES database. It was found that L-moments show non-linear relationships with moments, but with smaller magnitudes. 2W2W outperforms 2W in estimating both moments and L-moments, especially for higher-order statistics. 2W2W has the potential to better capture both unimodal and bimodal distributions compared to 2W. While 2W2W generally outperforms 2W under MM, noticeable oscillations were observed at some points. Although ML is the most accurate method at most points, LM still outperforms ML at specific locations based on both 2W and 2W2W. This study is expected to offer valuable insights into modeling PDFs for urban wind environments.