Using Cloud-System Resolving Models with a Data-driven Strategy to Advance Theories of Convectively Coupled Waves and the Madden-Julian Oscillation

Satellite images of the tropics commonly show large regions of rain and deep convective clouds moving along the equator, crossing the Pacific over the course of a week or two. The size and propagation speed of the cloudy regions are linked to planetary-scale equatorial waves but the theory does not involve moisture, clouds, or precipitation. As a result, important questions remain about how large-scale wind patterns and gentle rising and subsiding wave motions produce deep convective clouds and rainfall. The extent to which clouds influence these waves remains uncertain. These systems, known as convectively coupled waves (CCWs), highlight the importance of interactions between large-scale atmospheric dynamics and convection. A related question is how convection interacts with the larger and more slowly evolving Madden–Julian Oscillation (MJO). Improving understanding of these processes will enhance our ability to anticipate weather patterns associated with these systems, including heavy rainfall, hurricanes, and other high-impact events. This project also advances the development of artificial intelligence and machine learning approaches to improve analysis and prediction of complex atmospheric processes. Work supported here takes a novel approach to convective coupling research by developing a three-part research methodology combining high-resolution simulations with AI/ML techniques. In the first step wave-convection interactions are simulated using two high-resolution atmospheric models: SAM, the System for Atmospheric Modeling (see AGS-2218827) and SPCAM, the superparameterized Community Atmosphere Model (see AGS-0425247). In the second step the model simulations are used to train a nonlinear neural network model which builds on an earlier linear model developed by the Prinicipal Investigator of this award. The third step uses model order reduction to distill the behavior of the neural network into a simpler model in which the variables are physically meaningful and the model can be used to develop and test hypotheses for wave-convection coupling. One question to be addressed is the role of convective "memory", meaning the extent to which the the slow propagation of CCWs and the MJO depends on the time evolution of organized convection over the course of a few hours. 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. NSF Award ID: 2516280 | Program: 01002627DB NSF RESEARCH & RELATED ACTIVIT | Principal Investigator: Zhiming Kuang | Institution: Harvard University, CAMBRIDGE, MA | Award Amount: $749,788 View on NSF Award Search: https://www.nsf.gov/awardsearch/show-award/?AWD_ID=2516280 View on Research.gov: https://www.research.gov/awardapi-service/v1/awards/2516280.html