This ESTO-funded project develops an AI-based Seasonal to Interannual Forecast Model to improve long-range weather and climate prediction by integrating Earth system observations with physics-guided machine learning. The research leverages Transformer architectures to capture the spatiotemporal correlations and memory associated with major climate variability modes, including the Madden–Julian Oscillation (MJO) and El Niño–Southern Oscillation (ENSO). The model combines atmospheric and oceanic reanalysis products with high-resolution satellite observations of air–sea–land interactions, including sea surface salinity, top-of-atmosphere radiation, surface energy fluxes, and greenhouse gas fluxes. By fusing these complementary datasets, the project aims to identify key observational variables that enhance seasonal forecasting skill while maintaining physical consistency. The resulting framework provides a scalable foundation for extending predictions to other climate modes such as the North Atlantic Oscillation (NAO) and MJO.
This work directly supports NASA’s Earth Science Division modeling strategy by advancing AI technologies for Earth system prediction, improving understanding of climate variability, and enabling more accurate long-range weather forecasts that benefit climate research, resource management, and hazard preparedness.