The Madden-Julian Oscillation (MJO) is a climate pattern of fluctuating rainfall and winds that influences weather across the tropics. Its behavior is less predictable in summer than in winter, complicating vital forecasts for agriculture and climate preparedness. A new npj Climate and Atmospheric Science study applied machine learning (ML) and eXplainable Artificial Intelligence (XAI) to identify key controls on the summer MJO’s predictability, focusing on how precipitable water and surface temperature impacts the unique northeastward propagation pattern from the Indian Ocean. Researchers developed an ML-based prediction model that integrates long-term climate simulations with observational data, achieving skillful results over a 24-day forecast period. These findings enhance our understanding of summer MJO dynamics and illustrate the potential of ML to improve climate predictions, which is crucial for addressing the impacts of climate variability on agriculture and water management.
Contributing author Daehyun Kim pursued this research with funding support from the Climate Program Office’s Modeling, Analysis, Predictions, and Projections (MAPP) and Climate Variability & Predictability (CVP) programs. Kim works toward a MAPP initiative to address key issues in representing Earth System processes to improve climate models, especially in the tropics. Kim also has a grant from CVP focused on connecting MJO processes with our ability to predict precipitation. The results from this study add to a growing body of CPO-supported work on improving MJO understanding and prediction.
For more information, contact Clara Deck.
