摘要 : The zonal sea surface temperature (SST) gradient across the tropical Pacific is a pacemaker of the variable rates of global warming. In both historical simulations and future projections, the current state-of-the-art climate model... 展开 The zonal sea surface temperature (SST) gradient across the tropical Pacific is a pacemaker of the variable rates of global warming. In both historical simulations and future projections, the current state-of-the-art climate models show evident spread in the changes of zonal SST gradient, but the reasons remain unknown. Here, we quantify the contri-butions of internal variability and model spread to the uncertainty of zonal SST gradient changes by analyzing 342 realiza-tions from CMIP5 and CMIP6 models and several sets of large-ensemble simulations. We found that the internal variability dominates the total uncertainty at multidecadal time scales (;31-yr trends). Although the ratio of internal un-certainty to the total uncertainty declines along with higher emission of greenhouse gases under global warming, it is still over 80% at the multidecadal time scales in the future. The Pacific decadal oscillation is identified as the key internal mode responsible for the multidecadal uncertainty. For the future projections at centurial time scales, the uncertainty of zonal SST gradient changes is mainly from the intermodel spread in response to external forcing, accounting for about 70% of the uncertainty based on the difference between 2070-99 and 1970-99. The model spread in the cloud-shortwave radiation- SST feedback over the tropical Pacific is important in the uncertainty of zonal SST gradient changes. In particular, the intensity of negative convective cloud feedback in the western Pacific dominates the spread in CMIP5 models, while the intensity of stratocumulus cloud feedback over the southeastern Pacific is the primary process influencing the uncertainty in CMIP6 models. 收起