A Bayesian Machine Learning Forecast of the Nino 3 Sea Surface Temperature Index

A Bayesian Machine Learning Algorithm for Predicting ENSO Using Short Observational Time Series

A simple and efficient bayesian machine learning (bml) training and forecasting algorithm is developed to predict the ni\~no 3 sea surface temperature (sst)index.The algorithm exploits only a 20-year short observational timeseries and an approximate prior model, which exploits only a 20-year short observational timeseries and an approximate prior model.The bml forecast significantly outperforms model-based ensemble predictions and standard machine learning forecasts.The forecast starting from spring remains skillful for nearly 10 months.The bml algorithm can also effectively utilize multiscale features : the forecast of sst using sst, thermocline, and wind burstimproves on the forecast using just sst by at least 2 months.The bml algorithm also reduces the forecast uncertainty of neural networks and is robust to input perturbations.