Probabilistic Multi Model Ensemble for S2S prediction: From Traditional to Machine Learning Approaches

The subseasonal-to-seasonal (S2S) predictive timescale, encompassing lead times ranging from 2 weeks to a season, is at the frontier of prediction science. However, the S2S climate predictions suffer from a significant lack of prediction skills. The probabilistic multi model ensemble (PMME) is a well-accepted approach to improve the skill of prediction from individual dynamical models. In this talk, I will first discuss my ongoing work on developing several schemes of PMME using traditional to sophisticated statistical modeling for both non-parametric and parametric approaches and its application in sub-seasonal to seasonal prediction. I will then discuss my most recent work on how machine learning models can be used for constructing PMME and whether they added any value over existing methods.