The prediction of the Madden-Julian Oscillation (MJO), a massive tropical weather event with vast global socio-economic impacts, has been infamously difficult with physics-based weather prediction models. Here we construct a machine learning model using reservoir computing technique that forecasts the real-time multivariate MJO index (RMM), a macroscopic variable that represents the state of the MJO. The training data was refined by developing a novel filter that extracts the recurrency of MJO signals from the raw atmospheric data and selecting a suitable time-delay coordinate of the RMM. The model demonstrated the skill to forecast the state of MJO events for a month from the pre-developmental stages. Best-performing cases predicted the RMM sequence over two months, which exceeds the expected inherent predictability limit of the MJO.
@article{arxiv.2301.01254,
title = {Machine learning prediction of the MJO extends beyond one month},
author = {Tamaki Suematsu and Kengo Nakai and Tsuyoshi Yoneda and Daisuke Takasuka and Takuya Jinno and Yoshitaka Saiki and Hiroaki Miura},
journal= {arXiv preprint arXiv:2301.01254},
year = {2023}
}