Abstract By coupling large‐eddy simulation (LES) codes to weather data from large‐scale models, previous studies showed the viability of “real‐weather” LES. However, when simulating extended periods (up to one year) of weather, a number of them diagnosed an underestimation of the simulated temporal spectrum (of wind and solar irradiance) at timescales of a few hours (i.e., the atmospheric mesoscale). This study presents simulations aimed at reproducing the observed wind spectrum from timescales of one year to one minute, including the mesoscale. Reanalysis data (European Centre for Medium‐Range Weather Forecasts Reanalysis Version 5) are used as boundary conditions to a mesoscale simulation with either a local or a non‐local formulation of vertical diffusion, which then drives an LES (resolution of 50 m). Several domain sizes are used to simulate the weather during 2022 over a meteorological tower in The Netherlands. It is shown that, when increasing the size of the mesoscale simulation from 64 km to 1024 km, the LES wind spectrum at the mesoscale approaches the observed spectrum. The spectrum is also sensitive to the mesoscale diffusion formulation, which either resolves or suppresses explicit convection, resulting in a different LES wind spectrum. In addition, it is shown that the higher order statistics (structure functions) improve by using a large enough mesoscale simulation. The results indicate that LES can be used as a tool to simulate the temporal dynamics of the wind at all timescales between one minute and one year, if the atmospheric mesoscales are taken into account appropriately.
Postema et al. (Wed,) studied this question.