Gravity waves (GWs) and their associated multi-scale dynamics are known to play fundamental roles in energy and momentum transport and deposition processes throughout the atmosphere. GWs are well described by the Navier-Stokes equations, but solving these equations extending to very small scales remains daunting, and is limited by the computational costs of resolving the smallest important spatio-temporal features in a large-scale environment. This has led to developments of a wide variety of GW parameterization schemes for regional and global atmospheric models. Traditionally, GW parameterizations are based on linear theory, and no parameterization of secondary GWs (SGWs) has been developed to date. In addition, there remain many aspects of GW dynamics (e.g., GW self-acceleration, instability, GW breaking, SGW generation, and multi-scale interactions) that are important to describe, but cannot be addressed by linear theory or existing schemes.
Here we describe an initial, two-dimensional (2-D), machine learning model – the Compressible Atmosphere Model Neural Network (CAMNet) - intended as a first step toward a more general, three-dimensional, highly-efficient, model for applications to nonlinear GW dynamics description. CAMNet employs a physics-informed neural operator and GPU hardware to dramatically accelerate GW and SGW simulations applied to two GW sources to date. CAMNet is trained on high-resolution simulations by the state-of-the-art model Complex Geometry Compressible Atmosphere Model (CGCAM). Two initial applications to a Kelvin-Helmholtz instability source and mountain wave generation, propagation, breaking, and SGW generation in two wind environments are described here. Results show that CAMNet can capture the key 2-D dynamics modeled by CGCAM with high precision. Spectral characteristics of primary and SGWs estimated by CAMNet agree well with those from CGCAM. Our results show that CAMNet can achieve a several order-of-magnitude acceleration relative to CGCAM without sacrificing accuracy and suggests a potential for machine learning to enable efficient and accurate descriptions of primary and secondary GWs in global atmospheric models.