ENSO Teleconnections: How Well Do We Know Them And How Do We Evaluate Models Accordingly?

How do we determine extra-tropical ENSO teleconnections? Typically, one forms composites over many events in order to separate the component due to ENSO from that due to other sources of internal variability. Here we show through random sampling techniques applied to composite differences between 18 El Niño and 14 La Niña events observed since 1920 that considerable uncertainty exists in both the pattern and amplitude of the winter SLP response over the extra-tropical Northern Hemisphere, with associated effects on terrestrial climate impacts. This uncertainty poses considerable challenge for the evaluation of ENSO teleconnections in models. An approach is proposed that incorporates both pattern and amplitude uncertainty in the observational target, allowing for discrimination between true model biases in the forced ENSO response and apparent model biases that arise from limited sampling of non-ENSO-related internal variability. Large initial-condition coupled model ensembles with realistic tropical Pacific sea surface temperature anomaly evolution during 1920–2013 show similar levels of uncertainty in their ENSO teleconnections as found in observations. Because the set of ENSO events in each of the model composites is the same (and identical to that in observations), these uncertainties are entirely attributable to sampling fluctuations arising from internal variability, which is shown to originate from atmospheric processes. The initial-condition model ensembles thus inform the interpretation of the single observed ENSO composite and vice versa.