Abstract Global weather and climate forecasts hinge on the ability of coupled general circulation models (CGCMs) to simulate ocean‐atmosphere interactions in the tropical Pacific. Yet most CGCMs develop tropical Pacific biases, such as an overly‐intense equatorial cold tongue (ECT), which limit their simulation skill. Here we illuminate the sources of these biases in the GFDL SPEAR global CGCM, using bias‐corrected simulations with either surface nudging or flux adjustments. We find that the CGCM biases stem from (a) the atmosphere component's excessive convective rainfall and associated equatorial trade winds, which cause overly strong oceanic upwelling and meridional transport divergence in the ECT; (b) the ocean component's insufficient near‐surface vertical mixing, which shoals the tropical mixed layers and ECT thermocline; and (c) the ocean component's insufficient tropical instability wave stirring, especially during August‐November and La Niña, which weakens the meridional convergence of surface heat into the ECT. In the CGCM, coupled feedbacks further intensify and westward‐displace the equatorial climatological cold tongue, rainfall, trade winds, surface currents, Ekman upwelling, and thermocline tilt, which in turn lead to an overly‐semiannual cycle of equatorial sea surface temperature, and compensating biases among the upper‐ocean heat transports and air‐sea heat fluxes. We identify avenues to improve SPEAR and other CGCMs, via improved process representation informed by enhanced observations of the tropical Pacific upper‐ocean heat budget and air‐sea coupling.
Wu et al. (Sat,) studied this question.