Comparing RANS- and LES-based statistical methods for determining low-occurrence strong wind speeds in an actual urban area

Abstract

The low-occurrence strong wind speed (LOSWS) is a crucial factor in the urban wind environment. While several studies have estimated LOSWS based on high-order moments in statistical modelling, methods using commonly analyzed statistics in numerical simulations are more convenient but have not been thoroughly evaluated for urban cases. In this study, two statistical methods, KB method, which uses mean velocity components and turbulent kinetic energy, and Beta method, which additionally includes the integral time scale, were applied to estimate LOSWS using statistics from the Reynolds-averaged Navier–Stokes (RANS) simulations of an actual urban case. The accuracy of LOSWS estimation was also evaluated using statistics from large-eddy simulation (LES) to quantify potential error sources in the estimates derived from RANS statistics. Using LES statistics, both KB and Beta methods showed relative errors within ±10 % for LOSWSs at a 10 % exceedance probability and within ±25 % at 1 % and 0.1 % exceedance probabilities at most points. Although estimations based on RANS statistics showed larger deviations than those based on LES statistics, these two methods can still provide valuable a priori estimations, with most scatter points distributed along the 1:1 line, indicating acceptable agreement between the estimated and actual values. The main source of error for the two methods with RANS statistics is the numerical accuracy of turbulent kinetic energy. However, the significantly lower computational cost of RANS makes these estimations valuable for practical applications. The findings of this study provide valuable insights for estimating LOSWS using low-order statistics from LES or RANS simulations.