Advanced Physics-AI Models for Rain Enhancement in Arid Regions

Abstract

Despite advances in numerical weather prediction, they have been limited in representing the details of significant precipitation events in arid regions with large, growing urban centers, where weather modification efforts such as cloud seeding are currently underway or being planned. Understanding these limitations is critical to the success of such activities. To address some of these issues, the development of a hybrid modeling framework is required to help predict and improve the effectiveness of weather modification experiments. We will present our conceptual approach and how it leverages our on-going activities. We start by building upon the community atmospheric model, Weather Research and Forecasting with Chemistry (WRF-Chem), since it represents coupled cloud and aerosol properties. However, we assert that urbanization is a major source of local convection, aerosol loads, and potentially of microphysics activation and invigoration in the nuclei, in arid regions with large urban centers. Hence, standard WRF-Chem may be too limited to predict strong precipitation events and assess areas for potential cloud seeding experiments in such regions. It requires a version that includes urban processes. Past studies have shown that typical configurations for aerosol specifications in WRF-Chem may not accurately represent their diversity that can be present in arid regions, especially along coastlines, including dust, urban and industrial pollution, particulates from wildfire and agricultural burning, and marine salt that can affect cloud condensation and ice nucleation. To reduce the uncertainty, we apply machine learning algorithms to historical aerosol observations. Such data are also used to validate models. The results of applying the machine learning could then lead to improved representations of aerosols in an urbanized WRF-Chem with the potential to enhance operational support of cloud-seeding exercises.