Using causal discovery to understand misrepresentations of boreal summer tropical – extratropical teleconnections in S2S forecasts
Improving the reliability of subseasonal to seasonal (S2S) prediction is challenged by model biases and misrepresentation of teleconnections in seasonal forecasts. Here, we apply a causal discovery technique developed to identify causal relationships on a 2D map (causal map) to identify tropical–extratropical teleconnections between tropical convective activity and mid-latitude circulation in the Northern Hemisphere during boreal summer in SEAS5 forecasts. We compare causal maps obtained in ERA5 with SEAS5 for the 1993-2016 period. Qualitatively, spatial patterns produced with SEAS5 seasonal forecasts show good agreement with ERA5 patterns. However, the magnitude of the causal links is generally weaker and a dependence on ENSO is missing. We use subsampling techniques and composite analysis to further investigate these issues and understanding the role of sea surface temperatures in determining the strength of these tropical–extratropical teleconnection in SEAS5.