Abstract
Probability density distributions of each velocity component at pedestrian levels are required to predict extremely rare and strong wind events. Therefore, in this study, the PDFs were compared between the Gaussian and modified Gaussian distributions with the third-order and fourth-order statistics, as known as the Gram-Charlier Series (GCS), for the velocity components around a simplified urban array determined by a large-eddy simulation. The GCS with the third-order and fourth-order moments could predict the PDFs and percentile values of each velocity component more accurately than those determined by the Gaussian distributions. In addition, the comparison shows that it can be judged whether the PDFs are represented by the Gaussian distributions or not by confirming the third and fourth order statistics.
