Tracing the origins of equatorial Pacific biases in a coupled climate model

The mean climate of the tropical Pacific features a warm pool in the western Pacific and a cold tongue in the eastern equatorial Pacific, accompanied by easterly trade winds. This background state plays a critical role in regulating tropical climate variability and change across different time scales, including the El Niño-Southern Oscillation (ENSO), the most powerful driver of year-to-year climate variability. However, many state-of-the-art coupled climate models simulate an eastern equatorial cold tongue that is too strong and extends too far west, which limits their ability to accurately represent ENSO and other climate processes. Isolating the sources of these model biases has proven challenging due to strong ocean-atmosphere coupling. 

In a recent study, Wu and co-authors traced the origins of this cold tongue bias in the NOAA Geophysical Fluid Dynamics Laboratory (GFDL) SPEAR coupled climate model (SPEAR_LO; Fig. 1d-f) using a set of mean-state correction experiments. In these flux-adjusted simulations (SPEAR_LO_FA; Fig. 1a–c), the surface fluxes of heat, momentum, and freshwater were modified to bring the model’s time-mean sea surface temperatures (SSTs), surface wind stress, and sea surface salinity closer to observations. By examining how the model responded to these adjustments, the authors were able to determine whether the biases arise from the atmospheric component, the ocean component, or their coupled interactions.

The authors found that both the atmosphere and ocean contribute to the model’s equatorial Pacific biases. In the atmospheric model, excessive convective rainfall strengthens the equatorial trade winds (Fig. 1a). These stronger winds increase ocean upwelling and surface cooling, intensifying the equatorial cold tongue. Meanwhile, the ocean model exhibits insufficient near-surface vertical mixing and weak stirring by tropical instability waves (TIWs), which are 1,000–2,000 km scale ocean eddies that transport heat toward the equator (Fig. 1b). When the atmosphere and ocean interact in the fully coupled model, these individual errors amplify each other, through feedbacks that intensify and extend the cold tongue westward across the Pacific (Fig. 1d–f). The resulting biases alter ocean currents, rainfall patterns, and the seasonal cycle of SSTs, ultimately affecting the model’s simulation of ENSO.

By identifying the processes responsible for tropical Pacific mean-state biases, this study provides detailed guidance for improving the climate models used for seasonal-to-decadal simulations, reanalyses, predictions, and projections, and highlights the need for enhanced observations of upper-ocean heat transport, air-sea fluxes, and TIWs in the equatorial Pacific to better constrain model development.

This research was performed while the lead author was affiliated with Princeton University, through the generous support of NOAA GFDL and NOAA’s Climate Program Office (CPO) Climate Variability and Predictability (CVP) Program.