Understanding urban turbulence at hectometric scale using WRF-LES and eddy covariance observations

Despite the growing application of hectometric (100-m) urban simulations, their ability to realistically represent urban turbulence remains uncertain. This study assesses the improvements offered by large eddy simulations (LES) over the commonly used Yonsei University (YSU) planetary boundary layer scheme (PBL) in urban WRF hectometric simulations at 100-m horizontal and 50-m vertical grid resolution. The results are evaluated against Eddy Covariance (EC) observations from Enschede city, The Netherlands. Two contrasting periods are examined: a windy, shear-driven event and a calm, hot, buoyancy-driven event. Both YSU and LES faithfully capture temporal patterns of near-surface meteorology. In addition, during convective conditions, LES captures realistic turbulent fluctuations indicated by turbulence spectra and resolved fluxes while YSU produces smooth fluxes. LES resolves significant portion of convective turbulence and structures (rolls during windy and cells during calm conditions), however it remains partially dependent on sub-grid scale model at first level. YSU fails to produce any convective turbulent structures during windy conditions, but succeeds in producing partially developed convective cells under calm conditions resolving upto 20% of turbulence. Partial resolving of turbulence with YSU violates the assumption of 1D PBL scheme causing potential double-counting of turbulence. Hectometric resolution is insufficient to resolve nighttime turbulence, and both schemes depend on parameterizations. The structures in LES also enhance the horizontal advection which redistribute the heat between urban and rural areas. Overall, hectometric WRF-LES generates realistic turbulence and influences boundary-layer dynamics during convective conditions, strengthening confidence in its use for urban climate studies, including heat stress assessments, and precipitation. • Hectometric (100 m) WRF-LES is evaluated against urban Eddy Covariance observations. • Performance compared with YSU 1D PBL scheme at the same resolution. • LES resolves turbulent structures and flux variability whereas YSU relies on parameterization. • LES partially resolves turbulence and spectra at 50 m, YSU produces overly smooth fluxes.