Predicting Spring El Niño Forecast Overconfidence in NMME

The prediction of El Niño is important for the North American region due to its large influence for seasonal forecasts. During the northern hemisphere spring, El Niño forecasts are particularly challenging. This is known as the spring predictability barrier. For forecasts initialized during the spring predictability barrier, dynamical seasonal forecast models produce too many confident but inaccurate predictions(False Alarms) of El Niño. Using the North American Multi-Model Ensemble, we explored seasonal forecasts for the past 30 years to identify these False Alarms and compare them to correct confident forecasts. We have noticed that for false alarms some regions have warmer sea surface temperature anomalies while other regions have cooler anomalies than in successful forecast years. We have created a False Alarm Test that looks at the average sea surface temperature anomalies for certain ocean regions outside of the equatorial Pacific that we have identified as important differentiators between false alarms and successful forecasts. If at initialization, a forecast has 4 or more regions that exceed a certain criteria then the forecast will likely produce a false alarm. This test is then incorporated into a statistical model to predict whether a confident forecast is a false alarm or not.