Efficient Graphics Processing Unit Modeling Of Street-Scale Weather Effects In Support Of Aerial Operations In The Urban Environment

Over the last few years, the concept of incorporating aerial vehicles into the urban environment for diverse purposes has attracted ample interest and investment. These purposes cover a broad spectrum of applications, from larger vehicles designed for passenger transport, to package delivery and inspection/surveillance missions performed by small unmanned drones. While these Advanced Air Mobility (AAM) operations have the potential to alleviate bottlenecks arising from saturated surface transportation networks, there are a number of challenges that need to be addressed to make these operations safe and viable. One challenge is predicting weather effects within the urban environment with the required level of spatiotemporal fidelity, which current operational weather models fail to provide due to the use of coarse grid spacings (a few kilometers) constrained by the predictive performance limitations of traditional computer architectures. Herein, we demonstrate how FastEddy (R), a microscale model that exploits the accelerated nature of graphics processing units for high-performance computing, can be used to understand and predict urban weather impacts from seasonal, day-to-day, diurnal, and sub-hourly scales. To that end, we efficiently perform more than 50 telescoped simulations of microscale urban effects at street-scale (5 m grid spacing) driven by realistic weather over a 20 km(2) region centered at the downtown area of Dallas, Texas. Our analyses demonstrate that urban-weather interactions at the street-scale are complex and tightly connected, which is of utmost relevance to AAM operations. These demonstrations reveal the capability of such models to provide real-time weather hazard avoidance products tailored to capture microscale urban effects.