The El Niño-Southern Oscillation (ENSO) system is the most significant equatorial interannual climate variability pattern and has enormous global climate impacts. The ability to forecast ENSO accurately is crucially important to human livelihoods worldwide. ENSO is characterized by strong winter-peaking sea surface temperature (SST) anomalies in the central-eastern tropical Pacific during the mature phase. However, the bias in simulating ENSO phase-locking behavior persists in climate models. In this study, the ENSO seasonal phase-locking behaviors simulated in 42 Coupled Model Intercomparison Project Phase 6 (CMIP6) models are evaluated by comparing 43 CMIP5 models and observations. Only a few models (12 CMIP5 and 15 CMIP6) simulated ENSO with a majority proportion of winter-peaking events, which indicates that reasonable ENSO seasonal phase-locking is still a challenge for state-of-the-art climate models. Furthermore, the seasonal cycle of the zonal SST gradient along the equator can explain approximately 24% and 27% of the variance in the ENSO phase-locking for CMIP5 and CMIP6, respectively. In particular, the strengths of the zonal SST gradient in the central-eastern Pacific during boreal spring and autumn are crucial. The biases in simulating the seasonal changes in the zonal SST gradient influence the zonal advective feedback's strengths that respond to the anomalous SST of ENSO, therefore disturbing the simulation of ENSO phase-locking. Improvement of the simulated ENSO phase-locking should be realized by focusing on the zonal SST gradient's seasonal cycle.
