Suppressing Non-ENSO forcing with ensemble mean

An ensemble of forecasts is necessary to identify the uncertainty in prediction of a non-linear system like climate. Ensemble averages are often used to represent the seasonal mean state and diagnose underlying physical mechanisms. Here, using ensemble seasonal hindcasts made by the Climate Forecast System version-2 (CFSv2) model, we show that teleconnection patterns differ significantly between the ensemble mean and the model’s individual members. The strong relationship between El Nino Southern Oscillation (ENSO) and Indian summer monsoon rainfall (ISMR) is a distinctive feature of the ensemble mean, but not the individual member. The dominant ENSO forcing and its response in the ensemble mean arise due to the maximum number of members having the same sign of an anomaly. In contrast, this number is limited for non-ENSO forcing, leading to a suppressed impact and insufficient ISMR response in the ensemble mean. This study highlights the inadequacy of relying solely on the ensemble mean for analyzing the characteristics of the model and making forecasts.