Computational fluid dynamics modelling of wind conditions
In order to assess whether breeding distributions are affected by airflow characteristics, wind conditions around the cliffs of Skomer were simulated using the computational fluid dynamics (CFD) model OpenFoam (openfoam.org version 5.x). OpenFoam is widely used for modelling atmospheric boundary layer flows (e.g. in the wind energy industry) and has been extensively validated over a similarly steep island51.
The initial coarse model domain was 5300 x 5000 x 1000 m, with a horizontal resolution of 20 m and a vertical resolution of 10 m. The bottom boundary represented the surface of the island which was taken from a DEM of Skomer with 2 m resolution (Lle Geo-Portal http://lle.gov.wales). A simulation was also run using the DEM with 1 m resolution, however, the outputs were not significantly different. After establishing the initial mesh, the tool snappyHexMesh in OpenFoam was used to incorporate the DEM into the modelling process, refining initial mesh cells close to the surface up to 3 times. This resulted in a finer resolution close to surface of 2.5 m in the horizontal and 1.25 m in the vertical. Simulations were completed when convergence was achieved using a steady-state incompressible solver with a k-ε turbulence closure scheme.
Wind simulations were run for NW, NE, SE and SW wind directions to cover a representative sample of wind directions, including the prevailing SW direction (see Results). This allowed us to test whether birds selected sites that were predicted by a specific set of wind parameters across wind directions. The initial wind speed was set to 10 m s-1 at 20 m height. The following airflow characteristics were extracted from the model output at 2 m normal to the ground surface (this height was selected in order to estimate the airflow conditions that birds would be exposed to close to their breeding cliffs): The two horizontal and vertical wind vectors (U_0, U_1 and U_2 respectively), mean wind speed (MeanU), turbulent kinetic energy (TKE), pressure (P), kinematic viscosity (Nut) of the air medium and turbulence dissipation rate (ε). These outputs were further used to estimate horizontal wind speed; gustiness and turbulence intensity (TI) following31.