Simulation of Coastal El Niño Events in CMIP6 Models and Its Relationship to Mean‐State and ENSO Biases

Abstract Coastal El Niño (COA) events follow two evolution pathways, remaining in the far eastern Pacific or expanding into basin‐wide El Niño events. Analyzing 55 CMIP6 models, we show that the two pathways are associated with distinct central and eastern Pacific ocean‐atmosphere conditions, and the simulated frequency of each type is linked to tropical Pacific mean‐state biases (zonal for standalone, meridional for spreading) while inversely related to ENSO simulation skill. Only 16% of models reproduce observed COA frequency. Models overproducing COA events do so because of mean‐state biases that make coastal warming too easy to trigger, amplified by an overly sensitive local atmospheric response. Spreading events are systematically under‐produced, but their occurrence increases in models with more realistic meridional precipitation structure and, among the better‐performing models, more efficient wind‐thermocline coupling. These findings identify physical constraints governing COA evolution and the key model biases limiting the fidelity of COA simulation in CMIP6.