Diagnosing the Dynamics of ENSO Flavors using Linear Inverse Models

ENSO is a major climate phenomenon and has been extensively explored over the last three decades. Recent attention to ENSO diversity has generated a second wave of ENSO research. In particular, strong Eastern Pacific (EP) El Niños have been shown to be controlled largely by slow oceanic thermocline feedbacks, while weaker Central Pacific (CP) El Niños show greater influence from fast atmospheric feedbacks. Despite progress on characterizing ENSO diversity, there remain controversies about its causes. This paper explores the dynamics of ENSO diversity in a multidimensional phase space, using a Linear Inverse Model (LIM) applied to a 2000-year pre-industrial simulation from GFDL CM2.1 coupled GCM.

The dynamics of CM2.1's ENSO flavors are investigated by constructing two LIMs in the multivariate principal component space: one using tropical Pacific SST only (to capture fast ocean-atmosphere feedbacks), and another using tropical Pacific SST plus ocean heat content (OHC, to capture additional slower feedbacks from ocean dynamics). We then use these two LIMs to hindcast two ENSO events from the CM2.1 control run, which begin from similar La Niña states but evolve a year later into different flavors of El Niño (EP or CP). The SST-only LIM successfully forecasts the real CP El Niño, but under-forecasts the magnitude of the EP event. The SST+OHC LIM, on the other hand, successfully forecasts the EP El Niño but evolves the CP event into an incorrect EP event. This demonstrates that CM2.1's different ENSO flavors can be viewed as diverging flows from nearby initial states, dominated by different linearized dynamics. A more extensive suite of LIM forecasts shows that the GCM's ENSO flavors are potentially predictable at a range of lead-times. We further introduce a new measure, based on historical data, to facilitate probabilistic forecasts of ENSO flavors.