Abstract: El Nino–Southern Oscillation (ENSO) is the dominant interannual variability on earth. It affects climate, ecosystems, and economies around the world. In the first part of this talk, I will develop a three-region multiscale stochastic model, which explains the generation of the ENSO diversity. The model perfectly reproduces the observed remarkable non-Gaussian statistics in both the central and eastern Pacific. Combining with a bivariate regression technique, the model can generate moderate and strong eastern Pacific events, including the delayed super El Nino, isolated and consecutive central Pacific events as well as single- and multi-year La Nina events with realistic amplitude, duration, irregularity and phase locking in the spatiotemporal patterns. Then in the second part of this talk, I will introduce a Bayesian machine learning forecast algorithm, which exploits only short observations for training the neural networks. It will be used for ENSO forecast.

Zoom Meeting number：690 1580 9116
Passcode：665779