Toward Turbulence Forecasting in the Atmospheric Boundary Layer at Large-Eddy Simulation Resolutions

The ability to accurately forecast turbulence within the atmospheric boundary layer (ABL) is critical to many applications including traffic management and operability of unmanned aerial systems, and wind farm power prediction and control, among others. Current turbulence forecast algorithms, mainly emergent from the aviation community, rely on the use of the so-called “turbulence indices” to diagnose mean turbulence levels utilizing operational numerical weather prediction (NWP) model output. Such NWP models typically have horizontal grid spacings of several kilometers, and therefore ABL turbulence effects are fully parameterized. In order to assess the potential for turbulence forecasts at eddy-resolving resolutions, we explore dynamic downscaling from the operational Rapid Refresh (RAP) forecast product, with horizontal grid spacing of ~13 km, down to 25 m grid spacing in large-eddy simulation (LES) mode using the Weather Research and Forecasting (WRF) model. The daytime portion of 10 selected days during the eXperimental Planetary boundary layer Instrumentation Assessment (XPIA) campaign was simulated, and compared to high-frequency sonic anemometer data from the Bolder Atmospheric Observatory tower. Our turbulence-resolving forecasts demonstrate the potential for improved turbulence predictions in the ABL. In particular, the LES forecasts better account for the observed spatiotemporal variability arising from the presence and evolution of turbulence structures. Also, we provide evidence of the value of LES-scale forecasts for probabilistic turbulence predictions. These promising findings further motivate ongoing efforts to develop accelerated model capabilities to enable turbulence-resolving forecasts in the near-future.