Impact of model mean sea surface temperature bias on the simulation of Pacific decadal variability

An unprecedented eastern Pacific cooling along with the strengthening of Pacific trade winds since the late 1990s is considered a pronounced contributor to the recent global surface warming slowdown. Recent studies have demonstrated that the Atlantic sea surface temperature (SST) warming trend and the corresponding trans-basin (Atlantic-Pacific) teleconnection are key drivers of the observed eastern Pacific cooling. Although the Atlantic warming is well captured in the CMIP5 models, most simulations failed to reproduce the eastern Pacific cooling or underestimated the magnitude. Here, we investigate a hypothesis that common model mean state (SST) bias of state-of-the-art models (the multi-model ensemble mean of CMIP5 models) may partly account for discrepancies between the observed and simulated Pacific La Niña-like cooling. Targeted model simulations under various mean SST backgrounds are then performed to investigate the eastern Pacific response to the observed Atlantic warming trend. Model results suggest that simulations include the mean SST bias in both Pacific and Atlantic basins leads to a pronounced underestimation of the eastern Pacific cooling. Assessed individually, the additional Pacific mean state bias has a more prominent (explains ~35% of the total tropical Pacific SST variance) impact on eastern Pacific cooling response to Atlantic warming than adding the Atlantic region mean state bias (explains ~25% variance). This result suggests that reducing model mean state bias in both the Pacific and Atlantic basins may significantly contribute towards improving simulations of the Pacific decadal variability and trans-basin teleconnections.