Reconciling opposite trends in the observed and simulated equatorial Pacific zonal sea surface temperature gradient

The reasons for large discrepancies between observations and simulations, as well as for uncertainties in projections of the equatorial Pacific zonal sea surface temperature (SST) gradient, are controversial. We used CMIP6 models and large ensemble simulations to show that model bias and internal variabilities affected, i.e., strengthened, the SST gradient between 1981 and 2010. The underestimation of strengthened trends in the southeast trade wind belt, the insufficient cooling effect of eastern Pacific upwelling, and the excessive westward extension of the climatological cold tongue in models jointly caused a weaker SST gradient than the recent observations. The phase transformation of the Interdecadal Pacific Oscillation (IPO) could explain ~ 51% of the observed SST gradient strengthening. After adjusting the random IPO phase to the observed IPO change, the adjusted SST gradient trends were closer to observations. We further constrained the projection of SST gradient change by using climate models’ ability to reproduce the historical SST gradient intensification or the phase of the IPO. These models suggest a weakened SST gradient in the middle of the twenty-first century.