Building-Resolving LES within the GPU-Accelerated FastEddy Model: Toward Street-Scale Weather Forecasting

Fine-scale weather modeling in urban environments is critical for many applications. From an aviation standpoint, the use of small unmanned aerial systems (UAS), is growing at a fast pace, and is expected to play soon a critical role as a new emerging mode of transportation and package delivery, among other applications. In particular, such systems are more than an order of magnitude smaller in size and weight than typical commercial aircraft and therefore are expected to be sensitive to smaller-magnitude turbulence encounters that do not affect commercial aircraft operations at upper levels. Moreover, UAS will be frequently operating in urban environments, which present additional challenges in predicting flow and turbulence features, in turn requiring specific guidance to guarantee safe and efficient operation of UASs in these complex scenarios under realistic weather conditions.

As a first step toward achieving full physics urban capabilities within the GPU-resident large-eddy simulation (LES) FastEddy model, we have implemented and validated a method for explicit representation of building effects. Herein, we extend the immersed body force method (IBFM) from Chan et al. (2007) to: i) be scale-aware (can be applied at any grid resolution) and, ii) control building temperatures. The extended IBFM has the advantage of being computationally very efficient, while retaining the ability of immersed boundary methods to use structured grids. In addition, the extended IBFM has the potential to be coupled to building energy models. The adequacy and accuracy of the IBFM is demonstrated using different test cases, namely a laboratory scale experiment consisting of an array of staggered cubes of 2-cm side and an atmospheric scale test case through simulations of downtown Oklahoma City during the Joint Urban 2003 (JU03) field campaign. Our LES IBFM results for mean wind speed and turbulence kinetic energy compare well to observations, and produce turbulence spectra that are in good agreement with sonic anemometer observations. In addition, quantification of statistical performance metrics for the JU03 simulations are within the range of other LES models in the literature employing body fitted and immersed boundary approaches. All these features of the IBFM, combined with the GPU-enabled accelerated LES modeling in FastEddy will facilitate a path toward realistic street-scale operational weather forecasting in the near future.