Modelling of wind flows over realistic forests with LES

Abstract. A large-eddy simulation (LES)-based model for the representation of wind flows over realistic forests and topography is presented. Terrain elevation and forest density maps from airborne laser scans are employed to investigate the importance of specific model choices related to capturing upstream terrain effects on the wind resource. The study is divided into three parts. Firstly, an extended verification process under idealized conditions is carried out. Secondly, a validation is done where the model is compared to field measurements acquired in the southeast of Sweden, and, finally, an assessment of the forest and terrain footprint is carried out based on variations in the surface representation. The results show agreement of turbulence statistics compared to the literature when the forest is explicitly modelled, following expected trends as a function of the tree density. When the forest is explicitly modelled, the impact of the ground roughness becomes insignificant, even for an unrealistically sparse forest. The study also demonstrates that a model relying only on ground roughness yields notable differences in the turbulence characteristics. This is partly attributed to the inability of the model to reproduce sufficient drag for forest-equivalent values of roughness length z0 while maintaining the applicability of wall functions, which can impose strict limitations on the grid near the ground. This is further complicated by the problem of converting realistic, heterogeneous forest fields to z0. Moreover, turbulence statistics in the roughness sublayer are affected by the lack of vertical permeability. The validation shows that the model is able to capture the flow characteristics imprinted by different surface features on the wind along three distinctive wind directions. Vertically separated spectral coherence from the LES is slightly below that of the IEC standard, which can be attributed to the reference velocities used in the normalization of the frequency. The footprint study shows that the heterogeneity of a realistic forest produces higher drag in comparison with homogeneous conditions while also providing better agreement with observations. An analysis based on correlations of upstream forest drag with target wind statistics shows that a point above the terrain is most significantly influenced by the footprint of a forest area located at about 10 times upstream of its height above ground. When correlations are applied to turbulence, this separation increases five-fold. These findings provide valuable insight to determine the optimal domain size of a computational domain in forest simulations under neutral atmospheric stratification. Further comparisons of fully uniform vs. limited areas of realistic forest revealed that at heights above 100 m, no clear differences in the wind flow are seen. Conversely, comparing flat terrain with the actual topography – with a realistic forest distribution in both cases – demonstrated the clear importance of capturing small-scale terrain features.