Multiscale Large Eddy Simulations of a convective offshore boundary layer: towards a mesoscale-LES coupling

Large Eddy Simulation of atmospheric flows has the advantage of reproducing the most energetic turbulent structures present in the Atmospheric Boundary Layer (ABL), which determine mixing of momentum and scalars. However, grid resolution does not permit to cover large extensions while performing an accurate LES, due to the unaffordable computational cost associated.

Nesting techniques can provide a framework to alleviate such problem, by starting from a coarse resolution with which a large domain can be covered and progressively refining space and time turbulent scales as long as the nested domain sizes are decreased. At this point, there are two questions arising: is it possible to obtain similar results in terms of mean values and high order statistics through nesting than with a single domain with reduced dimensions?, can turbulence be progressively developed from a large scale motion?.

In order to address the previous questions we divide our study in two parts. Firstly, we focus on a single domain with 6 km x 6 km x 2km. A convective offshore boundary layer is simulated using WRF-LES with a single domain with periodic conditions applied in the horizontal directions. We use 48 h of quasi-steady conditions obtained from the FINO1 mast and LiDAR wind profiling, located 40 km off the German shore in the North Sea, to initialize the computations. The surface heat flux from FINO1 is applied as lower boundary condition to generate a similar convective stability regime. At this stage, an appropriate grid size is selected and the influence of the subgrid scale contribution is evaluated by confronting two SGS models: Smagorinsky 1.5 TKE and Nonlinear Backscatter and Anisotropy. Results are compared/validated with field measurements from FINO1.

The second part of our research is focused on the nesting methodology. We set multiple nests from a parent domain of 100 km x 100 km with a very coarse resolution of 540 m. This is done in order to generate a low frequency turbulent motion, from which the rest of nests are progressively fed with. We observe that results from nesting process are in a good agreement with the single domain LES for mean velocity profiles. Velocity variances produce similar pattern but there is a damping effect on the nesting case. We attribute it to the interpolation of the solution on the domain boundaries, which acts as a filter kind and the development of turbulence is delayed by each nested domain. An improvement is observed if the final resolution achieved is higher than the one of the single domain case. This works demonstrates the potential of nesting technique to build the transition between mesoscale and LES regimes. Similar methodology can be applied for wind energy purposes to properly include turbulence effects on mesoscale models and thus, obtain more accurate short-term wind and power forecasts.