Increased occurrence of Central-Pacific El Niño in ERSST v4

Interest in El Niño diversity has increased in the past decade, with much attention given to the hypothesis that El Niño comes in two distinct types: an Eastern-Pacific (EP) El Niño with temperature anomalies centered off the coast of Peru and a Central-Pacific (CP) El Niño with temperature anomalies centered near the International Date Line. It is widely believed that the CP type of El Niño has been increasing in frequency in recent decades, and some studies do not show CP El Niño appearing in the record at all until the 1960's. However, many findings regarding El Niño diversity are uncertain given the divergence of results from different classification schemes.

We test the idea that the CP type of El Niño is a recent and increasingly frequent phenomenon by using the newly released Version 4 of NOAA's Extended Reconstructed Sea Surface Temperatures (ERSST v4) dataset with one popular central-location classification method and a new method based on observed spatial patterns of sea surface temperature anomalies. The new method selects an EP and a CP El Niño “end-member” from events that are universally classified one way or the other by preexisting methods and uses pattern correlation with these end-members to sort all El Niño years into EP-dominated, CP-dominated, or mixed influence (M) classes. We selected the 1997-1998 El Niño as the EP El Niño end-member and the 1994-1995 El Niño as the CP El Niño end-member based on the consensus between papers as to their classification and their relatively high Oceanic Niño Index values for their type, under the assumption that a stronger event signal will better stand out against any background noise.

The advantage of this end-member correlation approach is its simplicity, relying only on observed data without the need for statistical regression, while taking into account a larger portion of the equatorial Pacific than the traditional Niño 3, Niño 4, and related indices. The downside of this approach is the somewhat arbitrary nature of end-member selection, although this disadvantage is mitigated through the use of transparent and meaningful selection criteria. Additionally, we tested our results using an alternate end-member selection of 1982-1983 for the EP El Niño end-member and 2009-2010 for the CP El Niño end-member and reached the same conclusions as the original designation.

The updated ERSST dataset identifies significantly more CP El Niños than two older versions and an independent dataset when using an identical central-location method, and the new pattern correlation method identifies more CP El Niños than EP El Niños for all datasets. The increase in CP El Niño identification in ERSST v4 is linked mostly to differences in the Niño 4 region (central Pacific) in the late nineteenth and early twentieth centuries, although differences between the datasets persist even up to the past two decades. The increase in CP El Niño identification when using the pattern correlation method over the central-location method is consistent with previous studies comparing different methods. Taken together, these results underscore the importance of careful selection of El Niño events when studying the difference between EP and CP types of El Niño and the necessity of confirming results with multiple methods and/or datasets. Further, the apparent instability of binary EP/CP classification systems may indicate that other frames, such as a spectrum of El Niño diversity, may be more appropriate.