Simulations of ENSO Phase-Locking in CMIP5 and CMIP6

The characteristics of El Nino-Southern Oscillation (ENSO) phase-locking in observations and CMIP5 and CMIP6 models are examined in this study. Two metrics based on the peaking month histogram for all El Nino and La Nina events are adopted to delineate the basic features of ENSO phase-locking in terms of the preferred calendar month and strength of this preference. It turns out that most models are poor at simulating the ENSO phase-locking, either showing little peak strength or peaking at the wrong seasons. By deriving ENSO's linear dynamics based on the conceptual recharge oscillator (RO) framework through the seasonal linear inverse model (sLIM) approach, various simulated phase-locking behaviors of CMIP models are systematically investigated in comparison with observations. In observations, phase-locking is mainly attributed to the seasonal modulation of ENSO's SST growth rate. In contrast, in a significant portion of CMIP models, phase-locking is codetermined by the seasonal modulations of both SST growth and phase transition rates. Further study of the joint effects of SST growth and phase transition rates suggests that for simulating realistic winter peak ENSO phase-locking with the right dynamics, climate models need to have four key factors in the right combination: 1) correct phase of SST growth rate modulation peaking at the fall, 2) large-enough amplitude for the annual cycle in growth rate, 3) small amplitude of semiannual cycle in growth rate, and 4) small amplitude of seasonal modulation in SST phase transition rate.