Multiparameter probability distributions for at-site frequency analysis of annual maximum wind speed with L-Moments for parameter estimation

Abstract

Estimation of quantiles of Annual Maximum Wind Speed (AMWS) is needed in different environmental fields, engineering risk analysis, design of structures, renewable energy sources, agricultural operations, and climatology. Therefore, wind speed frequency analysis (WSFA) was carried out at nine stations from Pakistan. Multiparameter Probability Distributions (PDs), such as Generalized logistic (GLO), Generalized Extreme Value (GEV), Generalized Normal (GNO), Generalized Pareto (GPA), Weibull (WEI), Pearson type 3 (P3), Log Pearson type 3 (LP3); and two parameter PDs, such as Logistic (LOG), Normal (NOR), Gumbel (GUM), Exponential (EXP), and Uniform (UNI) were used to determine the most suitable distributions for the nine stations. The method of L-moments was used for estimating parameters of the distributions. The Kolmogorov-Smirnov (KS) test, Anderson-Darling (AD) test, Minimum L-Kurtosis (ML-K) Difference Criterion, and L-moment ratio diagram (L-ratio diagram) showed that four distributions, namely GEV, GNO, GPA, and GLO were the most suitable distributions for different stations and were superior to the two-parameter distributions. The quantile estimates (design estimates) from multiparameter PDs provide information on how fast the maximum wind will pass through a certain place and hence are important for policy makers and planners in the design and construction of different structures. The Multivariate Diebold–Mariano (DM) test was applied to check the accuracy of design estimates from the best fitted PDs and results indicated that they were significantly different.