LEAP-HI: Advancing Precision Neighborhood Scale Weather Forecasting with Autonomous Aircraft Systems and Adaptive Microscale Models

About This Grant

The impact and severity of individual weather events can be shaped at the neighborhood scale by unique local environmental features such as buildings, tree cover, pavement or nearby bodies of water. These features influence temperature, humidity and wind, potentially amplifying weather effects and leading to localized extremes like strong winds, elevated surface temperatures, and poor air quality. This Leading Engineering for America's Prosperity, Health, and Infrastructure (LEAP-HI) award supports research investigating development of a new system that combines atmospheric measurements and simulation to deliver accurate and actionable weather forecasts at the neighborhood scale. The system looks to be designed to improve routine predictions by accounting for fine-scale environmental effects that are often unresolved in current weather forecasting models. Central to the measurement system are uncrewed aircraft systems (UAS), which offer a proven advantage in high-resolution sensing of atmospheric conditions. The UAS-based observations look to feed into a high-resolution nested numerical weather prediction model enhanced with model adaptation and machine learning. This approach should allow the model to continually adjust and minimize prediction errors, resulting in more accurate, fine-grained forecasts of localized weather variability. The project brings together a multidisciplinary team with expertise in fluid dynamics, computational science, machine learning, atmospheric science, microscale modeling and autonomous UAS operations. Once fully developed, the integrated system intends to equip decision-makers and emergency responders with neighborhood-level weather insights to better prepare for and respond to extreme events. The project emphasizes stakeholder engagement, workforce development, and engineering education and outreach to help train the next generation of engineers equipped to address climate-related challenges. This research seeks to address key challenges in producing accurate operational forecasts at micrometeorological scales. Specifically, inaccuracies in the boundary conditions that drive microscale predictions can lead to significant errors in simulating flow fields and turbulence in the lower atmosphere. To mitigate this, the project looks to initialize the microscale model with a high-resolution parameter map of boundary conditions constructed using offline machine learning. The initial focus will be on estimating surface roughness length scales (which governs surface fluxes) and inflow conditioning parameters (such as perturbation scale and magnitudes). These parameter maps will include localization functions that are adapted via retrospective cost adaptation using UAS measurements to optimize agreement between model predictions and physical observations. Since these maps and localization functions are expected to slowly vary over time, the system intends to reduce reliance on continuous UAS measurements. Anticipated scientific contributions include: (i) a comprehensive computational framework for high-fidelity microscale forecasts; (ii) a novel methodology for integrating UAS observations into microscale atmospheric models; and (iii) innovative use of machine learning to improve the representation of boundary conditions that drive microscale dynamics. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.