Downscaling by Assimilating NWP fields into a CFD Model

Understanding the details of locale-specific flow in the atmospheric boundary layer (ABL) is critical for a variety of applications, including aviation weather forecasting, siting wind power plants, and making short term predictions of wind variability. Although Numerical Weather Prediction (NWP) models provide a good coarse grid solution and incorporate information representing the outer scale geophysical variability through evolving boundary conditions and assimilated observations of the current state of the atmosphere to predict flow characteristics, they are not appropriate for fine scale detail. Turbulence features finer than about 1 km, however, are not well captured by the turbulence physics of such models. On the other hand, Computational Fluid Dynamics (CFD) is capable of modeling the details of flow around specific geographic and man-made features. Ideally we wish to combine the advantages of both types of models. Here we demonstrate that assimilating the output of a mesoscale NWP model into a CFD model can improve the fine scale structure that can be modeled in the ABL. This work describes how the output of the WRF NWP model with four dimensional data assimilation (FDDA) is used to initialize and assimilate into CFD simulations with much finer grid spacing. Assimilating the NWP model data is critical to obtaining the spatially varying outer scale patterns. The CFD model is then able to downscale by accurately modeling flow around fine scale topographic features. We specifically consider surface heating and buoyancy effects in the CFD model as well as in the NWP assimilated data. In addition, a porosity model simulates the effects of vegetation, causing a more turbulent local surface layer.

The technique is demonstrated with case studies in rolling topography of central Pennsylvania. Two cases are chosen to represent conditions on a cold blustery winter day as well as on a warm summer day with plenty of radiatively forced convection. The mesoscale model is run and the result assimilated into the CFD model. We compare profiles of both mean and fluctuating wind components between the mesoscale model alone, the CFD model alone, and the fully assimilated mesoscale/CFD solution. In addition, we do a spectral analysis of the wind fluctuations and determine the impact of the mesoscale assimilation into the CFD model.