Self-Organizing Maps Identify Windows of Opportunity for Seasonal European Summer Predictions

We combine a machine learning method and ensemble climate predictions to investigate windows of opportunity for seasonal predictability European summer associated with the North Atlantic jet stream. particularly focus on impact spring sea surface temperatures (SST) four dominant atmospheric teleconnections stream: Oscillation (NAO) in positive negative phases, Ridge (At. Ridge), Low Low). go beyond standard forecast practices by not only identifying these their SST precursors but making use identified analysis dynamical ensemble. Specifically, we train neural network-based classifier Self-Organizing Maps (SOM) ERA-20C reanalysis it model simulations from Max Planck Institute Earth System Model mixed resolution (MPI-ESM-MR). two different sets 30-member hindcast ensembles initialized every May, one training evaluation between 1902 2008, verification 1980–2016, respectively. Among analyzed here, find that At. simulated MPI-ESM-MR shows best agreement ERA-20C, thereby representing its occurrence skillful predictions. Conversely, lowest agreement, which might limit skill early warning warmer than average summers. In summary, patterns SOM can be used guess at initialization guide sub-selection potential members. This holds especially true is unclear NAO. show predictive selected exceeds full over regions Euro-Atlantic domain where significantly correlates level pressure (SLP). particular, significant improvement SLP, geopotential height 500 hPa, 2 m temperature 3–4 months lead time Scandinavia, robust among ensembles.